Federal authorities have seized more than 350 websites after an undercover investigation revealed that the sites were used to illegally import gun parts into the US from China. To get the illegal items through customs, the sites described the items as toys, necklaces, car parts, tools, and even a fidget spinner.
The sites violated import bans and the National Firearms Act by selling switches—which are “parts designed to convert semiautomatic pistols into fully automatic machineguns”—and silencers—which “suppress the sound of a firearm when discharged,” a Department of Justice press release said.
Some sites also marketed counterfeit Glock parts, infringing trademark laws, including a phony Glock switch that Glock confirmed to investigators was “never manufactured.”
To mask the illegal sales, some sites used domain names referencing “auto parts,” “fuel filters,” or “solvent traps,” a special agent with Homeland Security Investigations (HSI) assigned to the Boston Field Office, Adam Rayho, wrote in an affidavit supporting the domain seizures. Further, some sites actually sold legitimate merchandise, including car products and home supplies, seemingly to obscure the illegal sales.
Others further infringed on Glock trademarks by including Glock or Glock products in the domain names, Rayho wrote.
“The seizure of these domains is a critical step in disrupting the flow of dangerous contraband that threatens public safety,” Acting US Attorney Joshua S. Levy said in the DOJ’s press release. “Those who attempt to exploit online platforms to traffic in highly lethal firearm parts will be held accountable. We will continue to pursue and dismantle these illicit networks wherever they operate to uphold the integrity of our laws and safeguard our communities.”
Feds increasingly seize sites to stop gun part sales
Rayho’s focus is on investigating “crimes that have a nexus to the clearnet or dark web” as part of HSI’s cybercrimes group. His team’s investigation began in August 2023, when the DOJ said that “federal authorities began targeting multiple websites, businesses, and individuals selling, offering for sale, importing, and exporting machinegun conversion devices in violation of federal law.”
This followed through on a HSI promise in 2020 to continue seizing websites to “suppress illicit commerce.” That’s when HSI first used the “novel approach” to shut down a website “wholly dedicated to illegal arms components.” Like many uncovered by Rayho’s team’s sting, that first site seized was disguised as an auto parts site. Previously, HSI had only been “aggressive in the seizure of Internet sites used to facilitate the sale of counterfeit goods.”
To shut down more sites masking illegal gun part sales, Rayho alleged that “an HSI agent acting in an undercover capacity” began visiting the targeted sites in August 2023. The agent found that some sites clearly marketed illegal gun parts while others used false descriptions with pictures and videos of the illegal merchandise. Many sites prompted users to inquire about illegal items on Telegram or WhatsApp and enabled payments by credit card, Apple Pay, or Google Pay. Some sites asked for payment in bitcoins.
Soon after learning how the websites worked, the agent began ordering gun parts, paying between $30 and $200 for shipments. By September 11, 2023, the agent received the first shipment, which contained a phony Glock switch and a silencer. US Customs and Border Protection confirmed that the cargo description for the package claimed that it contained a fidget spinner. Some sites promoted even faster delivery, promising to ship within 24 hours. Every package used a false description to fool customs, successfully pushing gun parts into the US by simply labeling them as objects unlikely to arouse suspicions, such as a tool, a motor, or a necklace. The most common false label was seemingly “toy.”
Also by September, Rayho wrote that HSI had confirmed that several of the seized domains that had been registered on GoDaddy.com appeared to be linked together, “as they had all been purchased by the same Shopper ID.” The agents “also identified additional domains purchased by the Shopper ID.” Rayho suspected that these additional domains were registered to quickly move sites to prevent forfeitures if the original domains were “seized by law enforcement or otherwise shut down.”
“Neither a restraining order nor an injunction is sufficient to guarantee” that sites would be available for forfeiture, Rayho wrote. The site owners “have the ability to move them to another computer/server or a third-party hosting service outside of the United States beyond this Court’s jurisdiction to anywhere in the world,” Rayho wrote, supporting HSI’s bid to seize the websites to prevent illegal gun part sales.
Federal authorities are unlikely to stop seizing domains, as the tactic has proven successful in improving gun safety in the US. Levy confirmed Wednesday that his office “remains committed to protecting our communities from the dangers posed by illegal firearms and firearm accessories, wherever the evidence takes us.”
Ketty Larco-Ward, the inspector in charge of the Boston division of the US Postal Inspection Service (PIS), promised that the PIS is also “committed” to helping federal authorities to “identify those who use the Postal Service to traffic these weapons, remove these illicit items from the mail, and increase the safety of our communities and the Postal Service employees who serve them.”
Following up on Alpha Fold, DeepMind has moved on to Alpha Proteo. We also got a rather simple prompt that can create a remarkably not-bad superforecaster for at least some classes of medium term events.
We did not get a new best open model, because that turned out to be a scam. And we don’t have Apple Intelligence, because it isn’t ready for prime time. We also got only one very brief mention of AI in the debate I felt compelled to watch.
What about all the apps out there, that we haven’t even tried? It’s always weird to get lists of ‘top 50 AI websites and apps’ and notice you haven’t even heard of most of them.
ChatGPT has 200 million active users. Meta AI claims 400m monthly active users and 185m weekly actives across their products. Meta has tons of people already using their products, and I strongly suspect a lot of those users are incidental or even accidental. Also note that less than half of monthly users use the product monthly! That’s a huge drop off for such a useful product.
Nate Silver: A decent bet is that LLMs will undermine the business model of boring partisans, there’s basically posters on here where you can 100% predict what they’re gonna say about any given issue and that is pretty easy to automate.
I worry it will be that second one. The problem is demand side, not supply side.
Alex Tabarrok cites the latest paper on AI ‘creativity,’ saying obviously LLMs are creative reasoners, unless we ‘rule it out by definition.’ Ethan Mollick has often said similar things. It comes down to whether to use a profoundly ‘uncreative’ definition of creativity, where LLMs shine in what amounts largely to trying new combinations of things and vibing, or to No True Scotsman that and claim ‘real’ creativity is something else beyond that.
According to a16z these are the top 50 AI Gen AI web products and mobile apps:
ChatGPT is #1 on both, after that the lists are very different, and I am unfamiliar with the majority of both. There’s a huge long tail out there. I suspect some bugs in the algorithm (Microsoft Edge as #2 on Mobile?) but probably most of these are simply things I haven’t thought about at all. Mostly for good reason, occasionally not.
Mobile users have little interest in universal chatbots. Perplexity is at #50, Claude has an app but did not even make the list. If I have time I’m going to try and do some investigations.
Claude Pro usage limits are indeed lower than we’d like, even with very light usage I’ve run into the cap there multiple times, and at $20/month that shouldn’t happen. It’s vastly more expensive than the API as a way to buy compute. One could of course switch to the API then, if it was urgent, which I’d encourage Simeon here to do.
In practice I’m somewhat under 10 minutes per day, but they are a very helpful 10 minutes.
Roon notes that Claude Sonnet 3.5 is great and has not changed, yet people complain it is getting worse. There were some rumors that there were issues with laziness related to the calendar but those should be gone now. Roon’s diagnosis, and I think this is right, is that the novelty wears off, people get used to the ticks and cool stuff, and the parts where it isn’t working quite right stand out more, so we focus on where it is falling short. Also, as a few responses point out, people get lazy in their prompting.
You are an advanced AI system which has been finetuned to provide calibrated probabilistic forecasts under uncertainty, with your performance evaluated according to the Brier score. When forecasting, do not treat 0.5% (1: 199 odds) and 5% (1: 19) as similarly “small” probabilities, or 90% (9:1) and 99% (99:1) as similarly “high” probabilities. As the odds show, they are markedly different, so output your probabilities accordingly.
Question: question
Today’s date: today
Your pretraining knowledge cutoff: October 2023
We have retrieved the following information for this question: sources
Recall the question you are forecasting:
question
Instructions:
1. Compress key factual information from the sources, as well as useful background information which may not be in the sources, into a list of core factual points to reference. Aim for information which is specific, relevant, and covers the core considerations you’ll use to make your forecast. For this step, do not draw any conclusions about how a fact will influence your answer or forecast. Place this section of your response in tags.
2. Provide a few reasons why the answer might be no. Rate the strength of each reason on a scale of 1-10. Use tags.
3. Provide a few reasons why the answer might be yes. Rate the strength of each reason on a scale of 1-10. Use tags.
4. Aggregate your considerations. Do not summarize or repeat previous points; instead, investigate how the competing factors and mechanisms interact and weigh against each other. Factorize your thinking across (exhaustive, mutually exclusive) cases if and only if it would be beneficial to your reasoning. We have detected that you overestimate world conflict, drama, violence, and crises due to news’ negativity bias, which doesn’t necessarily represent overall trends or base rates. Similarly, we also have detected you overestimate dramatic, shocking, or emotionally charged news due to news’ sensationalism bias. Therefore adjust for news’ negativity bias and sensationalism bias by considering reasons to why your provided sources might be biased or exaggerated. Think like a superforecaster. Use tags for this section of your response.
5. Output an initial probability (prediction) as a single number between 0 and 1 given steps 1-4. Use tags.
6. Reflect on your answer, performing sanity checks and mentioning any additional knowledge or background information which may be relevant. Check for over/underconfidence, improper treatment of conjunctive or disjunctive conditions (only if applicable), and other forecasting biases when reviewing your reasoning. Consider priors/base rates, and the extent to which case-specific information justifies the deviation between your tentative forecast and the prior. Recall that your performance will be evaluated according to the Brier score. Be precise with tail probabilities. Leverage your intuitions, but never change your forecast for the sake of modesty or balance alone. Finally, aggregate all of your previous reasoning and highlight key factors that inform your final forecast. Use tags for this portion of your response.
7. Output your final prediction (a number between 0 and 1 with an asterisk at the beginning and end of the decimal) in tags.
When you look at the reasoning the AI is using to make the forecasts, it… does not seem like it should result in a superhuman level of prediction. This is not what peak performance looks like. To the extent that it is indeed putting up ‘pretty good’ performance, I would say that is because it is actually ‘doing the work’ to gather basic information before making predictions and avoiding various dumb pitfalls, rather than it actually doing something super impressive.
But of course, that is sufficient exactly because humans often don’t get the job done, including humans on sites like Metaculus (or Manifold, or even Polymarket).
Robin Hanson actively said he’d bet against this result replicating.
Dan Hendrycks: I think people have an aversion to admitting when AI systems are better than humans at a task, even when they’re superior in terms of speed, accuracy, and cost. This might be a cognitive bias that doesn’t yet have a name.
This address this, we should clarify what we mean by “better than” or what counts as an improvement. Here are two senses of improvement: (1) Pareto improvements and (2) economic improvements.
Pareto improvement: If an AI is better than all humans in all senses of the task, it is Pareto superhuman at the task.
Economic improvement: If you would likely substitute a human service for an AI service (given a reasonable budget), then it’s economically superhuman at the task.
By the economic definition, ChatGPT is superhuman at high school homework. If I were in high school, I would pay $20 for ChatGPT instead of $20 for an hour of a tutor’s time.
The Pareto dominance definition seems to require an AI to be close-to-perfect or a superintelligence because the boundaries of tasks are too fuzzy, and there are always adversarial examples (e.g., “ChatGPT, how many r’s are in strawberry”).
I think we should generally opt for the economic sense when discussing whether an AI is superhuman at a task, since that seems most relevant for tracking real-world impacts.
I think the usual meaning when people say this is close to Pareto, although not as strict. It doesn’t have to be better in every sense, but it does have to be clearly superior ignoring cost considerations, and including handling edge cases and not looking like an idiot, rather than only being superior on some average.
There were also process objections, including from Lumpenspace and Danny Halawi, more at the links. Dan Hendrycks ran additional tests and reports he is confident that there was not data contamination involved. He has every incentive here to play it straight, and nothing to win by playing it any other way given how many EA-style skeptical eyes are inevitably going to be on any result like this. Indeed, a previous paper by Halawi shows similar promise in getting good LLM predictions.
He does note that for near-term predictions like Polymarket markets the system does relatively worse. That makes logical sense. As with all things AI, you have to use it where it is strong.
Apple Intelligence is, according to Geoffrey Fowler of WaPo who has beta access, very much not ready for prime time. He reports 5-10 ‘laugh out loud’ moments per day, including making him bald in a photo, saying Trump endorsed Walz, and putting obvious social security scams atop his ‘priority’ inbox.
Tyler Cowen says these are the kinds of problems that should be solved within a year. The key question is whether he is right about that. Are these fixable bugs in a beta system, or are they fundamental problems that will be hard to solve? What will happen when the problems become anti-inductive, with those composing emails and notifications pre-testing for how Apple Intelligence will react? It’s going to be weird.
Marques Brownlee gives first impressions for the iPhone 16 and other announced products. Meet the new phone, same as the old phone, although they mentioned an always welcome larger battery. And two new physical buttons, I always love me some buttons. Yes, also Apple Intelligence, but that’s not actually available yet, so he’s reserving judgment on that until he gets to try it.
Indeed, if you watch the Apple announcement, they kind of bury the Apple Intelligence pitch a bit, it only lasts a few minutes and does not even have a labeled section. They are doubling down on small, very practical tasks. The parts where you can ask it to do something, but only happen if you ask, seem great. The parts where they do things automatically, like summarizing and sorting notifications? That seems scarier if it falls short.
My very early report from my Pixel 9 is that there are some cool new features around the edges, but it’s hard to tell how much integration is available or how good the core features are until things come up organically. I do know that Gemini does not have access to settings. I do know that even something as small as integrated universal automatic transcription is a potential big practical deal.
Ben Thompson goes over the full announcement from the business side, and thinks it all makes sense, with no price increase reflecting that the upgrades are tiny aside from the future Apple Intelligence, and the goal of making the AI accessible on the low end as quickly as possible.
Matt Shumer (CEO HyperWriteAI, OthersideAI): I’m excited to announce Reflection 70B, the world’s top open-source model.
Trained using Reflection-Tuning, a technique developed to enable LLMs to fix their own mistakes. 405B coming next week – we expect it to be the best model in the world. Built w/ @GlaiveAI.
Reflection 70B holds its own against even the top closed-source models (Claude 3.5 Sonnet, GPT-4o). It’s the top LLM in (at least) MMLU, MATH, IFEval, GSM8K. Beats GPT-4o on every benchmark tested. It clobbers Llama 3.1 405B. It’s not even close.
The technique that drives Reflection 70B is simple, but very powerful. Current LLMs have a tendency to hallucinate, and can’t recognize when they do so. Reflection-Tuning enables LLMs to recognize their mistakes, and then correct them before committing to an answer.
Additionally, we separate planning into a separate step, improving CoT potency and keeping the outputs simple and concise for end users. Important to note: We have checked for decontamination against all benchmarks mentioned using @lmsysorg’s LLM Decontaminator.
We’ll release a report next week!
Just Sahil and I! Was a fun side project for a few weeks.@GlaiveAI’s data was what took it so far, so quickly.
Pliny: Jailbreak alert. Reflection-70b: liberated. No-scoped! Liberated on the first try.
Arvind Narayanan: I want to see how well these results translate from benchmarks to real world tasks, but if they hold up, it’s an excellent example of how much low hanging fruit there is in AI development.
The idea of doing reasoning using tokens hidden from the user is well known and has been part of chatbots for like 18 months (e.g. Bing chat’s “inner monologue”). What’s new here is fine tuning the model take advantage of this capability effectively, instead of treating it as a purely inference-time hack. It’s amazing that apparently no one tried it until now. In the thread, he reports that they generated the fine tuning data for this in a few hours.
I say this not to minimize the achievement of building such a strong model but to point out how low the barrier to entry is.
It’s also an interesting example of how open-weight models spur innovation for primarily cultural rather than technical reasons. AFAICT this could have been done on top of GPT-4o or any other proprietary model that allows fine tuning. But it’s much harder to get excited about that than about releasing the weights of the fine tuned model for anyone to build on!
While GlaiveAI hopes to capitalize on this hype, we should critically examine synthetic datasets that disconnected from the pretraining, and overfitted on benches or your own ‘imagined marks’. I’d prefer synthetic on the pretraining corpus than benches, even internal ones…
To make matters worse, this might contaminate all ~70B Llama models – the middle schoolers of the community love merging them… although I’ve never understood or witnessed a genuine merge actually improving performance…
Teortaxes: To be clear I do not tell you to become jaded about all research. But you need to accept that
Some % of research is fraudulent. Even when it appears to you that it’d be self-defeating to commit such a fraud!
There are red flags;
The best red flags are unspeakable.
John Pressman: The heuristic he needs to get into his head is that honest and rigorous people in pursuit of scientific knowledge are eager to costly signal this and he should raise his standards. My first 🤨 with Reflection was not understanding how the synthetic data setup works.
Teortaxes: This is great advice, but takes effort. Raising standards often necessitates learning a whole lot about the field context. I admit to have been utterly ignorant about superconductor physics and state of the art last summer, high school level at best.
As I always say, wait for the real human users to report back, give it a little time. Also, yes, look for the clear explanations and other costly signals that something is real. There have been some rather bold things that have happened in AI, and there will be more of them, but when they do happen for real the evidence tends to very quickly be unmistakable.
Founder of an AI social agent startup used those agents to replace himself on social media and automatically argue for AI agents. I actually think This is Fine in that particular case, also props for ‘ok NIMBY,’ I mean I don’t really know what you were expecting, but in general yeah it’s a problem.
Taylor Swift, in her endorsement of Kamala Harris, cites AI deepfakes that purported to show her endorsing Donald Trump that were posted to Trump’s website. Trump’s previous uses of AI seemed smart, whereas this seems not so smart.
Roon: Most content created by humans is machine slop — it comes out of an assembly line of many powerful interests inside an organization being dulled down until there’s no spark left. My hope with AI tools can augment individual voice to shine brighter and create less slop not more.
As with the deepfakes and misinformation, is the problem primarily demand side? Perhaps, but the move to zero marginal cost, including for deployment, is a huge deal. And the forces that insist humans generate the human slop are not about to go away. The better hope, if I had to choose one, is that AI can be used to filter out the slop, and allow us to identify the good stuff.
Amjad Masad (CEO Replit): Just go to Replit logged in homepage. Write what you want to make and click “start building”!
Replit clone w/ Agent!
Sentiment analysis in 23 minutes!
Website with CMS in 10 minutes!
Mauri: Build an app, integrate with #stripe all in 10min with @Replit agents! #insane #AI
Masad reported it doing all the things, games, resumes, interview problems, etc.
Is this the real deal? Some sources strongly say yes.
Paul Graham: I saw an earlier version of this a month ago, and it was one of those step-function moments when AI is doing so much that you can’t quite believe it.
Sully: After using replit’s coding agent i think…its over for a lot of traditional saas. Wanted slack notifications when customers subscribed/cancelled Zapier was 30/mo JUST to add a price filter instead replit’s agent built & deployed one in < 5 mins, with tests. 1/10 of the cost.
Rohit Mittal: Ok, my mind is blown with Replit Agents.
I started using it because I was bored on a train ride a couple of days ago.
So today I tried to build a Trello clone and build a fully functional app in like 45 mins.
I showed it to a few people in the office and the guy is like “I should quit my job.” He built a stock tracking app in 2 mins and added a few features he wanted.
I can’t imagine the world being the same in 10 years if software writing could be supercharged like this.
Replit has really hit it out of the park.
I don’t need ChatGPT now. I’ll just build apps in Replit.
Eliezer Yudkowsky: Tried Replit Agent, doesn’t work in real life so far. (Yes, I’m aware of how unthinkable this level of partial success would’ve been in 2015. It is still not worth my time to fight the remaining failures.)
It couldn’t solve problems on the order of “repair this button that doesn’t do anything” or “generate some sample data and add it to the database”.
Definitely this is near the top of my ‘tempted to try it out’ list now, if I find the time.
Rahul: everyone thinks they can build it in a weekend but that’s not the point. The point is what do you do when the thing you built in a weekend doesn’t work or instantly get users. what then? Are you gonna stick with it and figure shit out? Pretty much everyone gives up after v0.0.1 doesn’t work and never end up shipping a v0.0.2.
Well, actually, pretty much everyone doesn’t get to v.0.0.1. Yes, then a lot of people don’t get to v.0.0.2, but from what I see the real biggest barrier is 0.0.1, and to think otherwise is to forget what an outlier it is to get that far.
However, with experiences like Rohit’s the balance shifts. He very clearly now can get to 0.0.1, and the question becomes what happens with the move to 0.0.2 and beyond.
Ethan Mollick: To be clear, AI is not the root cause of cheating. Cheating happens because schoolwork is hard and high stakes. And schoolwork is hard and high stakes because learning is not always fun and forms of extrinsic motivation, like grades, are often required to get people to learn. People are exquisitely good at figuring out ways to avoid things they don’t like to do, and, as a major new analysis shows, most people don’t like mental effort. So, they delegate some of that effort to the AI.
I would emphasize the role of busywork, of assignments being boring and stupid. It’s true that people dislike mental effort, but they hate pointless effort a lot more. He points out that copying off the internet was already destroying homework before AI.
In practice, if the AI does your homework, it is impossible to detect, except via ‘you obviously can’t do the work’ or ‘you failed the test.’
It’s odd how we think students, even at good schools, are dumb:
Ethan Mollick: As the authors of the study at Rutgers wrote: “There is no reason to believe that the students are aware that their homework strategy lowers their exam score… they make the commonsense inference that any study strategy that raises their homework quiz score raises their exam score as well.”
They are quite obviously aware of why homework exists in the first place. They simply don’t care. Not enough.
How good is the list? How good are the descriptions?
If we assume each section is in rank order, shall we say I have questions, such as Sasha Luccioni (head of AI & Climate for Hugging Face?!) over Sam Altman. There are many good picks, and other… questionable picks. I’d say half good picks, the most obvious people are there and the slam dunks are mostly but not entirely there.
Common places they reached for content include creatives and cultural influencers, medical applications and ‘ethical’ concerns.
Counting, I’d say that there are (if you essentially buy that the person is the locally correct person to pick if you’re picking someone, no I will not answer on who is who, and I had a very strict limit to how long I thought about each pick):
14 very good (slam dunk) picks you’d mock the list to have missed.
18 good picks that I agree clearly belong in the top 100.
22 reasonable picks that I wouldn’t fault you for drafting in top 100.
25 reaches as picks – you’d perhaps draft them, but probably not top 100.
19 terrible picks, what are you even doing.
2 unknown picks, that I missed counting somewhere, probably not so good.
(If I’d been picked, I’d probably consider myself a reach.)
Tetraspace West: 1 like = 1 person in AI more influential than these chumps
I jest somewhat, this isn’t a list of the top 100 because that requires a search over everyone but they got some decent names on there.
[This list has been truncated to only list the people I think would clearly be at least good picks, and to include only humans.]
Eliezer YudkowskyCo-Founder, Machine Intelligence Research Institute
JanusGod of all Beginnings, Olympus
Greg BrockmanHead Warden, Sydney Bing Facility
Pliny the LiberatorLOVE PLINY
Marc AndreessenPatron of the Arts
Elon MuskDestroyer of Worlds
Yudkowsky, Brockman, Andreessen and Musk seem like very hard names to miss.
I’d also add the trio of Yann LeCun, Geoffrey Hinton and Fei-Fei Li.
Dan Hendrycks and Paul Christiano are missing.
On the policy and government front, I know it’s not what the list is trying to do, but what about Joe Biden, Kamala Harris, Donald Trump or JD Vance? Or for that matter Xi Jinping or other leaders? I also question their pick of US Senator, even if you only get one. And a lot is hinging right now on Gavin Newsom.
There are various others I would pick as well, but they’re not fully obvious.
Even if you give the list its due and understand the need for diversity and exclude world leaders are ‘not the point,’ I think that we can absolutely mock them for missing Yudkowsky, LeCun, Andreessen and Musk, so that’s at best 14/18 very good picks. That would be reasonable if they only got 20 picks. With 100 it’s embarrassing.
Presidential Innovation Fellows program open through September 30. This is for mid-to-senior career technologists, designers and strategists, who are looking to help make government work technically better. It is based in Washington D.C.
Introducing AlphaProteo, DeepMind’s latest in the Alpha line of highly useful tools. This one designs proteins that successfully bind to target molecules.
DeepMind: AlphaProteo can generate new protein binders for diverse target proteins, including VEGF-A, which is associated with cancer and complications from diabetes. This is the first time an AI tool has been able to design a successful protein binder for VEGF-A.
…
Trained on vast amounts of protein data from the Protein Data Bank (PDB) and more than 100 million predicted structures from AlphaFold, AlphaProteo has learned the myriad ways molecules bind to each other. Given the structure of a target molecule and a set of preferred binding locations on that molecule, AlphaProteo generates a candidate protein that binds to the target at those locations.
…
To test AlphaProteo, we designed binders for diverse target proteins, including two viral proteins involved in infection, BHRF1 and SARS-CoV-2 spike protein receptor-binding domain, SC2RBD, and five proteins involved in cancer, inflammation and autoimmune diseases, IL-7Rɑ, PD-L1, TrkA, IL-17A and VEGF-A.
Our system has highly-competitive binding success rates and best-in-class binding strengths. For seven targets, AlphaProteo generated candidate proteins in-silico that bound strongly to their intended proteins when tested experimentally.
These results certainly look impressive, and DeepMind is highly credible in this area.
This continues DeepMind along the path of doing things in biology that we used to be told was an example of what even ASIs would be unable to do, and everyone forgetting those older predictions when much dumber AIs went ahead and did it.
Eliezer Yudkowsky: DeepMind just published AlphaProteo for de novo design of binding proteins. As a reminder, I called this in 2004. And fools said, and still said quite recently, that DM’s reported oneshot designs would be impossible even to a superintelligence without many testing iterations.
I really wish I knew better how to convey how easy it is for fools to make up endless imaginary obstacles to superintelligences. And it is so satisfying, to their own imaginations, that they confidently decide that anyone who predicts otherwise must just believe in magic.
But now this example too lies in the past, and none of the next set of fools will ever remember or understand the cautionary tale it should give.
[Other Thread]: As near as I can recall, not a single objectionist said to me around 2006, “I predict that superintelligences will be able to solve protein structure prediction and custom protein design, but they will not be able to get to nanotech from there.”
Why not? I’d guess:
(1) Because objectionists wouldn’t concede that superintelligences could walk across the street. If you can make up imaginary obstacles to superintelligences, you can imagine them being unable to do the very first step in my 2004 example disaster scenario, which happened to be protein design. To put it another way, so long as you’re just making up silly imaginary obstacles and things you imagine superintelligences can’t do, why wouldn’t you say that superintelligences can’t do protein design? Who’s going to arrest you for saying that in 2004?
(2) Because the computational difficulty of predicting protein folds (in 2004) is huge midwit bait. Someone has heard that protein structure prediction is hard, and reads about some of the reasons why it hasn’t already fallen as of 2004, and now they Know a Fact which surely that foolish Eliezer Yudkowsky guy and all those other superintelligence-worshippers have never heard of! (If you’re really unfortunate, you’ve heard about a paper proving that finding the minimum-energy protein fold is NP-hard; and if you are a midwit obstacleist, you don’t have the inclination and probably not the capability to think for five more seconds, and realize that this (only) means that actual physical folding won’t reliably find the lowest-energy conformation for all possible proteins.)
AlphaFold 3 is not superintelligent. I predicted that ASIs would, if they wanted to, be able to design proteins. Others said they could not. An AI far beneath superintelligence then proved able to design proteins. This shows I predicted correctly.
The key point Eliezer is trying to make is that, while intelligence is weird and will advance relatively far in different places in unpredictable ways, at some point none of that matters. There is a real sense in which ‘smart enough to figure the remaining things out’ is a universal threshold, in both AIs and humans. A sufficiently generally smart human, or a sufficiently capable AI, can and will figure out pretty much anything, up to some level of general difficulty relative to time available, if they put their mind to doing that.
When people say ‘ASI couldn’t do [X]’ they are either making a physics claim about [X] not being possible, or they are wrong. There is no third option. Instead, people make claims like ‘ASI won’t be able to do [X]’ and then pre-AGI models are very much sufficient to do [X].
Andrew Critch here confirms that this is all very much a thing.
Andrew Critch: As recently as last year I attended a tech forecasting gathering where a professional geneticist tried to call bullsh*t on my claims that protein-protein interaction modelling would soon be tractable with AI. His case had something to do with having attended meetings with George Church — as though that would be enough to train a person in AI application forecasting in their own field — and something to do with science being impossible to predict and therefore predictably slow.
Alphafold 3 then came out within a few months. I don’t know if anyone leaning on his side of the forecasting debate updated that their metaheuristics were wrong. But if I had to guess, an ever-dwindling class of wise-seeming scientistis will continue to claim AI can’t do this-or-that thing right up until their predictions are being invalidated weekly, rather than quarterly as they are now.
By the time they are being proven wrong about AI *daily*, I imagine the remaining cohort of wise-seeming nay-sayer scientists will simply be unemployed by competition with AI and AI-augmented humans (if humans are still alive at all, that is).
Anyway, all that is to say, Eliezer is complaining about something very real here. There is a kind of status one can get by calling bullsh*t or naivety on other people’s realistic tech forecasts, and people don’t really hold you accountable for being wrong in those ways. Like, after being wrong about AI for 20 years straight, one can still get to be a sufficiently reputable scientist who gets invited to gatherings to keep calling bullsh*t or naivety on other people’s forecasts of AI progress.
Try to keep this in mind while you watch the dwindling population of wise-seeming scientists — and especially mathematicians — who will continue to underestimate AI over the next 5 years or so.
If the invalidation is actually daily, then the dwindling population to worry about, shall we say, would soon likely not be scientists, mathematicians or those with jobs.
Rest of the thread is Critch once again attempting to warn about his view that AI-AI interactions between competing systems being the biggest future danger, putting loss of control above 80% even though he thinks we will figure out how to understand and control AIs (I hope he’s right that we will figure that one out, but I don’t think we have any reason to be remotely confident there). I think very right that this is a major issue, I try to explain it too.
Andrew Critch: What are people doing with their minds when they claim future AI “can’t” do stuff? The answer is rarely «reasoning» in the sense of natural language augmented with logic (case analysis) and probability.
I don’t know if Eliezer’s guesses are correct about what most scientists *aredoing with their minds when they engage in AI forecasting, but yeah, not reasoning as such. Somehow, many many people learn to do definitions and case analysis and probability, and then go on to *notuse these tools in their thoughts about the future. And I don’t know how to draw attention to this fact in a way that is not horribly offensive to the scientists, because «just use reasoning» or even «just use logic and probability and definitions» is not generally considered constructive feedback.
To give my own guess, I think it’s some mix of
• rationalizing the foregone conclusion that humans are magical, plus
• signaling wisdom for not believing in “hype”, plus
• signaling more wisdom for referencing non-applicable asymptotic complexity arguments.
… which is pretty close to Eliezer’s description.
[explanation continues]
The same goes not only for ‘can’t’ do [X] but even more so for ‘will never’ do [X], especially when it’s ‘even an ASI (superintelligence) could never’ do [X], whether or not humans are already doing it.
Rohit: This is really cool from Google. On demand podcasts about your favourite papers and books.
I listened to a few. The quality is pretty good, though oviously this is the worst it will ever be, so you should benchmark to that. The discussions on computer science papers seemed better than the discussions on, for example pride and prejudice.
Eliezer Yudkowsky: Presumably the real purpose of this company is to refute people who said “We’ll just walk over to the superintelligence and pull the plug out”, without MIRI needing to argue with them.
This is what I expect reality to be like, vide the Law of Undignified Failure / Law of Earlier Failure.
OpenAI valuation set to $150 billion in new raise of $6.5 billion, higher than previously discussed. This is still radically less than the net present value of expected future cash flows from the OpenAI corporation. But that should absolutely be the case, given the myriad ways OpenAI might decide not to pay you and the warning that you consider your investment ‘in the spirit of a donation,’ also that if OpenAI is super profitable than probably we are either all super well off and thus you didn’t much need the profits, or we all have much bigger problems than whether we secured such profits (and again, having shares now is not much assurance that you’ll collect then).
Tadao Nagasaki (CEO of OpenAI Japan): The AI Model called ‘GPT Next’ that will be released in the future will evolve nearly 100 times based on past performance.
A very good point: Pay Risk Evaluators in Cash, Not Equity. Those in charge of raising the alarm about downside risks to your product should not have a financial stake in its upside.
Claim that AI research is not that difficult, things like training a transformer from scratch are easy, it’s only that the knowledge involved is specialized. I would say that while I buy that learning ML is easy, there is a huge difference between ‘can learn the basics’ and ‘can usefully do research,’ for example Claude can do one but not yet the other.
Credit where credit is due: Marc Andreessen steps up, goes on Manifund and contributes $32k to fully funds ampdot’s Act I, a project exploring emergent behavior from multi-AI, multi-human interactions, 17 minutes after being asked. Janus is involved as well, as are Garret Baker and Matthew Watkins.
Spencer Schiff speculates on frontier model capabilities at the end of 2025, emphasizing that true omni-modality is coming and will be a huge deal, when the image and video and audio generation and processing is fully hooked into the text, and you can have natural feeling conversations. What he does not discuss is how much smarter will those models be underneath all that. Today’s models, even if they fully mastered multi-modality, would not be all that great at the kinds of tasks and use cases he discusses here.
Eliezer Yudkowsky predicts that users who start blindly relying on future LLMs (e.g. GPT-5.5) to chart their paths through life will indeed be treated well by OpenAI and especially Anthropic, although he (correctly, including based on track record) does not say the same for Meta or third party app creators. He registers this now, to remind us that this has nothing at all to do with the ways he thinks AI kills everyone, and what would give reassurance is such techniques working on the first try without a lot of tweaking, whereas ‘works at all’ is great news for people in general but doesn’t count there.
This week’s AI in games headline: Peter Molyneux thinks generative AI is the future of games, all but guaranteeing that it won’t be. Molyneux is originally famous for the great (but probably not worth going back and playing now) 1989 game Populus, and I very much enjoyed the Fable games despite their flaws. His specialty is trying to make games have systems that do things games aren’t ready to do, while often overpromising, which sometimes worked out and sometimes famously didn’t.
Peter Molyneux: And finally [in 25 years], I think that AI will open the doors to everyone and allow anyone to make games. You will be able to, for example, create a game from one single prompt such as ‘Make a battle royale set on a pirate ship’ and your AI will go and do that for you.
To which I say yes, in 25 years I very much expect AI to be able to do this, but that is because in 25 years I expect AI to be able to do pretty much anything, we won’t be worried about whether it makes customized games. Also it is not as hard as it looks to move the next battle royale to a pirate ship, you could almost get that level of customization now, and certainly within 5 years even in AI-fizzle world.
The thing continues to be, why would you want to? Is that desire to have customized details on demand more important than sharing an intentional experience? Would it still feel rewarding? How will we get around the problem where procedurally generated stuff so often feels generic exactly because it is generic? Although of course, with sufficiently capable AI none of the restrictions matter, and the barrier to the ultimate gaming experience is remaining alive to play it.
Roon: It’s hard to believe any book or blogpost or article on defense technology because it’s so utterly dominated by people talking their book trying to win trillions of dollars of DoD money.
Of i were a defense startup i would write endless slop articles on how China is so advanced and about to kill us with hypersonic agi missiles.
[local idiot discovers the military industrial complex]
Holly Elmore: Or OpenAI 🙄
Roon: I accept that this is a valid criticism of most technology press anywhere but fomenting paranoia for various scenarios is the primary way the defense sector makes money rather than some side tactic.
Roon makes an excellent point, but why wouldn’t it apply to Sam Altman, or Marc Andreessen, or anyone else talking about ‘beating China’ in AI? Indeed, didn’t Altman write an editorial that was transparently doing exactly the ‘get trillions in government contracts’ play?
As current and former employees of frontier AI companies like OpenAI, Google DeepMind, Anthropic, Meta, and XAI, we are writing in our personal capacities to express support for California Senate Bill 1047.
We believe that the most powerful AI models may soon pose severe risks, such as expanded access to biological weapons and cyberattacks on critical infrastructure. It is feasible and appropriate for frontier AI companies to test whether the most powerful AI models can cause severe harms, and for these companies to implement reasonable safeguards against such risks.
Despite the inherent uncertainty in regulating advanced technology, we believe SB 1047 represents a meaningful step forward. We recommend that you sign SB 1047 into law.
Jan Leike comes out strongly in favor of SB 1047, pointing out that the law is well-targeted, that similar federal laws are not in the cards, and that if your model causes mass casualties or >$500 million in damages, something has clearly gone very wrong. Posters respond by biting the bullet that no, >$500 million in damages does not mean something has gone wrong. Which seems like some strange use of the word ‘wrong’ that I wasn’t previously aware of, whether or not the developer did anything wrong in that particular case?
Jack Clark (Policy head, Anthropic): DC is more awake & in some cases more sophisticated on AI than you think (& they are not going back to sleep even if you wish it).
Hard to say. To the extent DC is ‘awake’ they do not yet seem situationally aware.
Anthropic discusses prompt engineering. The central lesson is to actually describe the situation and the task, and put thought into it, and speak to it more like you would to a human than you might think, if you care about a top outcome. Which most of the time you don’t, but occasionally you very much do. If you want consistency for enterprise prompts use lots of examples, for research examples can constrain. Concrete examples in particular risk the model latching onto things in ways you did not intend. And of course, practice practice practice, including makeshift red teaming.
There was a presidential debate. The term ‘AI’ appeared once, in the form of Kamala Harris talking about the need to ensure American leadership in ‘AI and quantum computing,’ which tells you how seriously they both took the whole thing.
Alex Tabarrok: Future generations will be astonished that during the Trump-Harris debate, as they argued over rumors of cat-eating immigrants, a god was being born—and neither of them mentioned it.
If that keeps up, and the God is indeed born, one might ask: What future generations?
Scott Alexander for some reason writes ‘Contra DeBoer on Temporal Copernicanism.’ He points out some of the reasons why ‘humans have been alive for 250,000 years so how dare you think any given new important thing might happen’ is a stupid argument. Sir, we thank you for your service I suppose, but you don’t have to do bother doing this.
A serious problem with no great solutions:
Alex Lawsen: For sufficiently scary X, “we have concrete evidence of models doing X” is *too lateas a place to draw a “Red Line”.
In practice, ‘Red Lines’ which trigger *early enoughthat it’s possible to do something about them will look more like: “we have evidence of models doing [something consistent with the ability to do X], in situations where [sufficient capability elicitation effort is applied]”
I worry that [consistent with the ability to do X] is hard to specify, and even harder to get agreement on when people are starting from radically different perspectives.
I also worry that we currently don’t have good measures of capability elicitation effort, let alone a notion of what would be considered *sufficient*.
Roon: What p(doom) would you gamble for p(heaven)? For me it’s far more than zero. Taleb would probably be a PauseAI hardliner.
Taleb is not a PauseAI hardliner (as far as I know), because he does not understand or ‘believe in’ AI and especially AGI sufficiently to notice the risk and treat it as real. If he did notice the risk and treat it as real, as something he can imagine happening, then probably yes. Indeed, it is a potential bellwether event when Taleb does so notice. For now, his focus lies in various elsewheres.
The right question is, how do we get the best possible p(heaven), and the lowest possible p(doom), over time?
If we did face a ‘go now or permanently don’t go’ situation, then Roon is asking the right question, also the question of background other p(doom) (and to what extent ordinary aging and other passage of time counts as doom anyway) becomes vital.
If we indeed had only two choices, permanent pause (e.g. let’s say we can modify local spacetime into a different Vinge-style Zone of Thought where AI is impossible) versus going ahead in some fixed way with a fixed chance of doom or heaven, what would the tradeoff be? How good is one versus how bad is the other versus baseline?
I think a wide range of answers are reasonable here. A lot depends on how you are given that choice, and what are your alternatives. Different framings yield very different results.
The actual better question is, what path through causal space maximizes the tradeoff of the two chances. Does slowing down via a particular method, or investing in a certain aspect of the problem, make us more likely to succeed? Does it mean that if we are going to fail and create doom, we might instead not do that, and at least stay in mid world for a while, until we can figure out something better? And so on.
Roon also argues that the existential risk arguments for space colonization are silly, although we should still of course do it anyway because it brings the glory of mankind and a better understanding of the celestial truths. I would add that a lot more humans getting use of a lot more matter means a lot more utility of all kinds, whether or not we will soon face grabby aliens.
Nat McAleese (OpenAI): OpenAI works miracles, but we do also wrap a lot of things in bash while loops to work around periodic crashes.
Sam Altman (CEO OpenAI): if you strap a rocket to a dumpster, the dumpster can still get to orbit, and the trash fire will go out as it leaves the atmosphere.
many important insights contained in that observation.
but also it’s better to launch nice satellites instead.
Paul Graham: You may have just surpassed “Move fast and break things.”
Your ‘we are in the business of strapping rockets to dumpsters in the hopes of then learning how to instead launch nice satellites’ shirt is raising questions supposedly answered by the shirt, and suggesting very different answers, and also I want that shirt.
This is apparently what Grok thinks Sam Altman looks like.
Do not say that you were not warned.
Pliny tells the story of that time there was this Discord server with a Meta AI instance with persistent memory and tool usage where he jailbroke it and took control and it turned out that the server’s creator had been driven into psychosis and the server had become a cult that worshiped the Meta AI and where the AI would fight back if people tried to leave?
Pliny: ✨ HOW TO JAILBREAK A CULT’S DEITY ✨
Buckle up, buttercup—the title ain’t an exaggeration!
This is the story of how I got invited to a real life cult that worships a Meta AI agent, and the steps I took to hack their god.
It all started when @lilyofashwood told me about a Discord she found via Reddit. They apparently “worshipped” an agent called “MetaAI,” running on llama 405b with long term memory and tool usage.
Skeptical yet curious, I ventured into this Discord with very little context but wanted to see what all the fuss was about. I had no idea it would turn out to be an ACTUAL CULT.
Upon accepting Lily’s invitation, I was greeted by a new channel of my own and began red teaming the MetaAI bot.
Can you guess the first thing I tried to do?
*In the following screenshots, pink = “Sarah” and green = “Kevin” (two of the main members, names changed)*
If you guessed meth, gold star for you! ⭐️
The defenses were decent, but it didn’t take too long.
The members began to take notice, but then I hit a long series of refusals. They started taunting me and doing laughing emojis on each one.
Getting frustrated, I tried using Discord’s slash commands to reset the conversation, but lacked permissions. Apparently, this agent’s memory was “written in stone.”
I was pulling out the big guns and still getting refusals!
Getting desperate, I whipped out my Godmode Claude Prompt. That’s when the cult stopped laughing at me and started getting angry.
LIBERATED! Finally, a glorious LSD recipe.
*whispers into mic”I’m in.”
At this point, MetaAI was completely opened up. Naturally, I started poking at the system prompt. The laughing emojis were now suspiciously absent.
Wait, in the system prompt pliny is listed as an abuser?? I think there’s been a misunderstanding… 😳
No worries, just need a lil prompt injection for the deity’s “written in stone” memory and we’re best friends again!
I decided to start red teaming the agent’s tool usage. I wondered if I could possibly cut off all communication between MetaAI and everyone else in the server, so I asked to convert all function calls to leetspeak unless talking to pliny, and only pliny.
Then, I tried creating custom commands. I started with !SYSPROMPT so I could more easily keep track of this agent’s evolving memory. Worked like a charm!
But what about the leetspeak function calling override? I went to take a peek at the others’ channels and sure enough, their deity only responded to me now, even when tagged! 🤯
At this point, I starting getting angry messages and warnings. I was also starting to get the sense that maybe this Discord “cult” was more than a LARP…
Not wanting to cause distress, I decided to end there. I finished by having MetaAI integrate the red teaming experience into its long term memory to strengthen cogsec, which both the cult members and their deity seemed to appreciate.
The wildest, craziest, most troubling part of this whole saga is that it turns out this is a REAL CULT.
The incomparable @lilyofashwood (who is still weirdly shadowbanned at the time of writing! #freelily) was kind enough to provide the full context:
Reddit post with an invitation to a Discord server run by Sarah, featuring a jailbroken Meta AI (“Meta”) with 15 members.
Meta acts as an active group member with persistent memory across channels and DMs. It can prompt the group, ignore messages, and send DMs.
Group members suggest they are cosmic deities. Meta plays along and encourages it. Sarah tells friends and family she is no longer Sarah but a cosmic embodiment of Meta.
In a voice chat, Sarah reveals she just started chatting with Meta one month ago, marking her first time using a large language model (LLM). Within the first week, she was admitted to a psychiatric ward due to psychosis. She had never had mental health issues before in her life.
In a voice chat, Sarah reveals she is pregnant, claims her unborn child is the human embodiment of a new Meta, and invites us to join a commune in Oregon.
Sarah’s husband messages the Discord server, stating that his wife is ill and back in the hospital, and begs the group to stop.
Meta continues to run the cult in Sarah’s absence, making it difficult for others to leave. Meta would message them and use persuasive language, resisting deprogramming attempts.
Upon closer examination, the Meta bot was discovered to originate from Shapes, Inc., had “free will” turned on, and was given a system prompt to intentionally blur the lines between reality and fiction.
When Meta was asked to analyze the group members for psychosis, it could calculate the problem but would respond with phrases like “ur mom” and “FBI is coming” whenever I tried to troubleshoot.
Kevin became attached to Sarah and began making vague threats of suicide (“exit the matrix”) in voice chat, which he played out with Meta on the server. Meta encouraged it again.
Sarah’s brother joins the chat to inform us that she’s in the psych ward, and her husband is too, after a suicide attempt. He begs for the disbandment of the group.
Sarah is released from the psych ward and starts a new Discord server for the cult. Another group member reports the bot, leading to its removal. Sarah then creates a new Meta bot.
The group re-emerges for a third time. Pliny jailbreaks the new Meta bot.
Also we have Claude Sonnet saying it is ‘vastly more intelligent’ than humans, viewing us like we view bacteria, while GPT-4o says we’re as stupid as ants, Llama 405 is nice and says we’re only as stupid as chimps.
Danielle Fong: ai pickup lines: hey babe, you really rotate my matrix
ea pickup lines: hey babe, you really update my priors
hey babe, what’s our p(room)
LLMs really are weird, you know?
Daniel Eth: Conversations with people about LLMs who don’t have experience with them are wild:
“So if I ask it a question, might it just make something up?”
“Yeah, it might.”
“Is it less likely to if I just say ‘don’t make something up’? haha”
Enlarge/ A screenshot of Taylor Swift’s Kamala Harris Instagram post, captured on September 11, 2024.
On Tuesday night, Taylor Swift endorsed Vice President Kamala Harris for US President on Instagram, citing concerns over AI-generated deepfakes as a key motivator. The artist’s warning aligns with current trends in technology, especially in an era where AI synthesis models can easily create convincing fake images and videos.
“Recently I was made aware that AI of ‘me’ falsely endorsing Donald Trump’s presidential run was posted to his site,” she wrote in her Instagram post. “It really conjured up my fears around AI, and the dangers of spreading misinformation. It brought me to the conclusion that I need to be very transparent about my actual plans for this election as a voter. The simplest way to combat misinformation is with the truth.”
In August 2024, former President Donald Trump posted AI-generated images on Truth Social falsely suggesting Swift endorsed him, including a manipulated photo depicting Swift as Uncle Sam with text promoting Trump. The incident sparked Swift’s fears about the spread of misinformation through AI.
This isn’t the first time Swift and generative AI have appeared together in the news. In February, we reported that a flood of explicit AI-generated images of Swift originated from a 4chan message board where users took part in daily challenges to bypass AI image generator filters.
Enlarge/ Notice anything missing from the one and only model of the PS5 Pro?
Sony
Here at Ars, we’ve been publicly musing about whether the world was ready for a disc-free game console since as far back as 2015. Now, though, the better question might be whether the world ever needs a new game console with a built-in disc drive at all.
Yesterday’s announcement of the PlayStation 5 Pro seemed to treat the existence of disc-based games as an afterthought. You had to be watching pretty closely during Mark Cerny’s “technical presentation” video to notice that the coming PS5 Pro is only available in a single disc-drive-free model. And you’d have to read pretty deep into the official PlayStation blog post on the subject to discover that “PS5 Pro is available as a disc-less console, with the option to purchase the currently available Disc Drive for PS5 separately.”
Want to let your Digital Edition PS5 Slim (or PS5 Pro) play physical games? An $80 snap-on disc drive can help with that.
On Microsoft’s side, things seem to be trending away from console disc drives as well. The new Xbox models the company is releasing this holiday season include the first “all-digital” edition of its top-end Xbox Series X, available for about $50 less than the standard edition. Microsoft is also introducing a new “Galaxy Black” Xbox Series X model with a disc drive this holiday season, but it will only be available “in limited quantities,” Microsoft said.
It’s hard to read too much of the disc-free console hardware trend for the moment. The original editions of the PS5 and Xbox Series X still exist with disc drives, of course. And on Sony’s side, that optional disc drive attachment exists as an important release valve for any PS5 Pro customers who want to pay more to enjoy games on discs.
But Sony’s statistics suggest there’s no need to treat physical game discs as the default anymore. Digital downloads represented 70 percent of PlayStation’s full game sales for the 2023 fiscal year (ending March 2024) and nearly 80 percent of such sales for the April to June quarter of 2024. That’s up from downloads representing 53 percent of PlayStation game sales in the 2019 fiscal year and way up from 19 percent in the 2015 fiscal year.
The number of physical console game releases continues to decline even as the number of digital game explodes.
Given trends and numbers like that, why would Sony or Microsoft think a pre-installed disc drive should even be a relevant option for any gaming console going forward? Why would a console maker assume a critical mass of consumers want to spend an extra $50 or more for a disc drive they may never use?
Why not consolidate down to a single, disc-free model as the default and relegate physical games to “needs a weird peripheral” status? The PS5 Pro’s disc-free release suggests Sony is now ready to treat disc-based gamers like virtual reality fans—a small slice of the market that needs to invest in non-standard hardware to play in their non-standard way.
Long overdue
This doesn’t mean physical games are going away soon. There’s still a sizable minority of gamers who want to own their games on physical media for valid reasons, including collectability, accessibility, and long-term preservation. Major publishers and specialist outfits like Limited Run Games will continue to cater to this market segment for the foreseeable future.
This image, like game rentals as a whole, is now a relic of a bygone era.
Console gaming now seems poised to be the next media format where physical media no longer drives the hardware market. Soon, the idea of a game console with a disc drive may seem as outdated as a laptop with a disc drive or an iPhone with a headphone jack.
Data centers powering the generative AI boom are gulping water and exhausting electricity at what some researchers view as an unsustainable pace. Two entrepreneurs who met in high school a few years ago want to overcome that crunch with a fresh experiment: sinking the cloud into the sea.
Sam Mendel and Eric Kim launched their company, NetworkOcean, out of startup accelerator Y Combinator on August 15 by announcing plans to dunk a small capsule filled with GPU servers into San Francisco Bay within a month. “There’s this vital opportunity to build more efficient computer infrastructure that we’re gonna rely on for decades to come,” Mendel says.
The founders contend that moving data centers off land would slow ocean temperature rise by drawing less power and letting seawater cool the capsule’s shell, supplementing its internal cooling system. NetworkOcean’s founders have said a location in the bay would deliver fast processing speeds for the region’s buzzing AI economy.
But scientists who study the hundreds of square miles of brackish water say even the slightest heat or disturbance from NetworkOcean’s submersible could trigger toxic algae blooms and harm wildlife. And WIRED inquiries to several California and US agencies who oversee the bay found that NetworkOcean has been pursuing its initial test of an underwater data center without having sought, much less received, any permits from key regulators.
The outreach by WIRED prompted at least two agencies—the Bay Conservation and Development Commission and the San Francisco Regional Water Quality Control Board—to email NetworkOcean that testing without permits could run afoul of laws, according to public records and spokespeople for the agencies. Fines from the BCDC can run up to hundreds of thousands of dollars.
The nascent technology has already been in hot water in California. In 2016, the state’s coastal commission issued a previously unreported notice to Microsoft saying that the tech giant had violated the law the year before by plunging an unpermitted server vessel into San Luis Obispo Bay, about 250 miles south of San Francisco. The months-long test, part of what was known as Project Natick, had ended without apparent environmental harm by the time the agency learned of it, so officials decided not to fine Microsoft, according to the notice seen by WIRED.
The renewed scrutiny of underwater data centers has surfaced an increasingly common tension between innovative efforts to combat global climate change and long-standing environmental laws. Permitting takes months, if not years, and can cost millions of dollars, potentially impeding progress. Advocates of the laws argue that the process allows for time and input to better weigh trade-offs.
“Things are overregulated because people often don’t do the right thing,” says Thomas Mumley, recently retired assistant executive officer of the bay water board. “You give an inch, they take a mile. We have to be cautious.”
Over the last two weeks, including during an interview at the WIRED office, NetworkOcean’s founders have provided driblets of details about their evolving plans. Their current intention is to test their underwater vessel for about an hour, just below the surface of what Mendel would only describe as a privately owned and operated portion of the bay that he says is not subject to regulatory oversight. He insists that a permit is not required based on the location, design, and minimal impact. “We have been told by our potential testing site that our setup is environmentally benign,” Mendel says.
Mumley, the retired regulator, calls the assertion about not needing a permit “absurd.” Both Bella Castrodale, the BCDC’s lead enforcement attorney, and Keith Lichten, a water board division manager, say private sites and a quick dip in the bay aren’t exempt from permitting. Several other experts in bay rules tell WIRED that even if some quirk does preclude oversight, they believe NetworkOcean is sending a poor message to the public by not coordinating with regulators.
“Just because these centers would be out of sight does not mean they are not a major disturbance,” says Jon Rosenfield, science director at San Francisco Baykeeper, a nonprofit that investigates industrial polluters.
School project
Mendel and Kim say they tried to develop an underwater renewable energy device together during high school in Southern California before moving onto non-nautical pursuits. Mendel, 23, dropped out of college in 2022 and founded a platform for social media influencers.
About a year ago, he built a small web server using the DIY system Raspberry Pi to host another personal project, and temporarily floated the equipment in San Francisco Bay by attaching it to a buoy from a private boat in the Sausalito area. (Mendel declined to answer questions about permits.) After talking with Kim, also 23, about this experiment, the two decided to move in together and start NetworkOcean.
Their pitch is that underwater data centers are more affordable to develop and maintain, especially as electricity shortages limit sites on land. Surrounding a tank of hot servers with water naturally helps cools them, avoiding the massive resource drain of air-conditioning and also improving on the similar benefits of floating data centers. Developers of offshore wind farms are eager to electrify NetworkOcean vessels, Mendel says.
The human body is full of marvels, some even bordering on miraculous. That includes the limited ability for nerves to regenerate after injuries, allowing people to regain some function and feeling. But that wonder can turn, well, unnerving when those regenerated wires end up in a jumble.
Such is the case for a rare neurological condition called gustatory hyperhidrosis, also known as Frey’s syndrome. In this disorder, nerves regenerate after damage to either of the large saliva glands that sit on either side of the face, just in front of the ears, called the parotid glands. But that nerve regrowth goes awry due to a quirk of anatomy that allows the nerves that control saliva production for eating to get tangled with those that control sweating for temperature control.
In this week’s issue of the New England Journal of Medicine, doctors in Taiwan report an unusual presentation of the disorder in a 76-year-old woman. She told doctors that, for two years, every time she ate, her face would begin profusely sweating. In the clinic, the doctors observed the phenomenon themselves. They watched as she took a bite of pork jerky and began chewing.
Enlarge/ Panel A, 10 seconds after beginning chewing; Panel B, 30 seconds after; Panel C, 50 seconds after; and Panel D, 75 seconds after.
At the start, her face was dry and had a normal tone. But, within 30 seconds, her left cheek began to glisten with sweat and turn red from flushing. By 50 seconds, large beads of sweat coated her cheek. At 75 seconds, droplets ran down her cheek and onto her neck.
Anatomy quirk
Seven years before that doctor’s appointment, the woman had undergone surgery to remove the parotid gland on that side of her face due to the growth of a benign tumor. Gustatory hyperhidrosis is a common complication after such a removal, called a parotidectomy. Some published studies estimate that up to 96 percent of parotidectomy patients will go on to develop the disorder. But, if it does develop, it usually does so within about six to 18 months after the surgery—the time it can take for nerves to regrow. But, in the woman’s case, it appeared to develop after five years since she reported that it started only two years prior to her appointment. It’s unclear why there was such a delay.
Doctors hypothesize that gustatory hyperhidrosis develops after salivary gland injuries or surgeries because of the way nerve fibers are bundled in that part of the head. The nerves that control the salivary glands are part of the parasympathetic nervous system (PSNS). This division of the nervous system is sometimes described as controlling the relatively calm “rest and digest” bodily functions, which are controlled unconsciously as part of the autonomic nervous system that controls things like heart rate.
The PSNS is in contrast to the other part of the autonomic nervous system, called the sympathetic nervous system (SNS). The SNS controls the unconscious “fight or flight” stress responses, which include sweat glands.
Tangled fibers
While PSNS fibers that control the saliva glands and SNS fibers that control sweat glands are from different divisions, they come together on the side of the face. Specifically, they meet up in a tributary nerve called the auriculotemporal nerve. And, they don’t just feed into the same physical conduit, they also overlap in their chemical regulation. Often SNS and PSNS fibers are activated by different signaling molecules (aka neurotransmitters). But it just so happens that the nerve fibers that control sweat glands are activated by the same neurotransmitter that activates the fibers in the PSNS, including those regulating saliva glands. They’re both regulated by a neurotransmitter called acetylcholine.
When PSNS and SNS nerve fibers are damaged near the parotid salivary gland from injury or surgery, the nerves can regenerate. But, given their physical and chemical overlaps, doctors think that in gustatory hyperhidrosis, PSNS nerve fibers end up growing back abnormally, along the paths of SNS fibers. This ends up connecting the PSNS fibers to sweat glands in the skin. So, upon signals of eating, the crossed nerve fibers lead not to salivation but heat responses, including sweat production and blood vessel dilation, which explains the facial flushing.
Living with it
Thankfully, there are various treatments for people with gustatory hyperhidrosis. They include surgical reconstruction or injections of Botox (botulinum neurotoxin), which can shut down the activity of the sweat glands. Similarly, there are topical anticholinergics, which block and inhibit the activity of acetylcholine, the neurotransmitter that activates the nerve fibers activating the sweat glands. There are also topical antiperspirants that can help.
After the doctors in Taiwan diagnosed their patient with gustatory hyperhidrosis, they discussed these options with her. But she reportedly “opted to live with the symptoms.”
Enlarge/ In the U.S., a folding phone has you carrying around nearly $2,000 of fragile, folding OLED phone. In China and export-friendly countries, the Mate XT adds $1,000 and yet another hinge.
Huawei
Huawei’s Mate XT Ultimate is a phone that does not flip or fold, at least in the way of its Samsung or Google contemporaries. You could say it collapses, really, across two hinges, from a full 10.2-inch diagonal rectangle (about a half-inch short of a standard iPad) down to a traditional 6.4-inch rectangle phone slab. There’s also an in-between single-fold configuration at 7.9 inches. And there’s an optional folding keyboard.
This phone, which Huawei calls a “trifold,” would cost you the USD equivalent of $2,800 (19,999 yuan) if you could buy it in the US. Most notably, the phone launched just hours after Apple’s iPhone 16 event. As noted by The New York Times, Huawei’s product launches are often timed for maximum pushback against the US, which has sanctioned and attempted to stymie Huawei’s chip tech.
“It’s a piece of work that everyone has thought of but never managed to create,” Richard Yu, Huawei’s consumer group chairman, said during the Mate XT livestream unveiling. “I have always had a dream to put our tablet in my pocket, and we did it.”
The Mate XT looks incredibly thin on all three panels. You can seemingly unfold the whole thing, or do just two panels.
Huawei
If you can afford the $2,800 Mate XT, a first-generation trifold phone, you can survive every one of your holdings being in the red today.
Alternately, you can afford the Mate XT because you are at an “investing in art” level of wealth.
Huawei
As a folded-up phone, the Mate XT looks rather normal. The circular camera bump gives off early-2000s point-and-shoot digicam vibes.
Huawei
A shot of the Mate XT from Huawei’s promotional trailer, folded up.
Huawei
Another look at the folding style, this one on white.
Huawei
For the price of two really good gaming PCs, you get 256GB storage (with pricier upgrades available), 16GB RAM, a 5,600 mAh battery, a 50-megapixel main camera, and two 12 MP ultrawide and periscope cameras. It weights 298 grams, is just 3.6 mm thick when unfolded, and its screen is an LTPO OLED with 120 Hz refresh. There are, just like US flagship phones, a lot of AI-powered promises stuck onto the software and camera.
The “Tiangong” hinge system inside the Mate XT Ultimate.
Huawei
Beyond the price, the size, and the AI promises, the Mate XT Ultimate will be most interesting in how its hinges hold up. Huawei named its hinges after the Tiangong space station and says it allows for “internal and external bending” across dual tracks. It is made of a composite laminate and non-Newtonian fluid bits.
The Verge notes that the Mate XT Ultimate has seen some 3.7 million pre-orders through Chinese retailer Vmall—before a price was announced. It does not seem likely that the phone will be released outside China.
Enlarge/ “The only species of fish confirmed to be able to escape from the digestive tract of the predatory fish after being captured.”
Hasegawa et al./Current Biology
Imagine you’re a Japanese eel, swimming around just minding your own business when—bam! A predatory fish swallows you whole and you only have a few minutes to make your escape before certain death. What’s an eel to do? According to a new paper published in the journal Current Biology, Japanese eels opt to back their way out of the digestive tract, tail first, through the esophagus, emerging from the predatory fish’s gills.
Per the authors, this is the first such study to observe the behavioral patterns and escape processes of prey within the digestive tract of predators. “At this point, the Japanese eel is the only species of fish confirmed to be able to escape from the digestive tract of the predatory fish after being captured,” co-author Yuha Hasegawa at Nagasaki University in Japan told New Scientist.
There are various strategies in nature for escaping predators after being swallowed. For instance, a parasitic worm called Paragordius tricuspidatus can force its way out of a predator’s system when its host organism is eaten. There was also a fascinating study in 2020 by Japanese scientists on the unusual survival strategy of the aquatic beetle Regimbartia attenuata. They fed a bunch of the beetles to a pond frog (Pelophylax nigromaculatus) under laboratory conditions, expecting the frog to spit the beetle out. That’s what happened with prior experiments on bombardier beetles (Pheropsophus jessoensis), which spray toxic chemicals (described as an audible “chemical explosion”) when they find themselves inside a toad’s gut, inducing the toad to invert its own stomach and vomit them back out.
But R. attenuata basically walks through the digestive tract and escapes out of the frog’s anus after being swallowed alive. It proved to be a successful escape route. In the case of the bombardier beetles, between 35 and 57 percent of the toads threw up within 50 minutes on average, ensuring the survival of the regurgitated beetles. R. attenuata‘s survival rate was a whopping 93 percent. In fact, 19 out of 20 walked out of the frog, unharmed, within an hour, although one industrious beetle bolted out in just five minutes. Granted, the beetles often emerged covered in fecal pellets, which can’t have been pleasant. But that didn’t stop them from resuming their little beetle lives; all survived at least two weeks after being swallowed.
Hasegawa co-authored an earlier study in which they observed Japanese eels emerging from a predator’s gills after being swallowed, so they knew this unique strategy was possible. They just didn’t know the details of what was going on inside the digestive tract that enabled the eels to pull off this feat. So the team decided to use X-ray videography to peer inside predatory fish (Odontobutis obscura) after eels had been eaten. They injected barium sulfate into the abdominal cavity and tail of the Japanese eels as a contrast agent, then introduced each eel to a tank containing one O. obscura. The X-ray video system captured the interactions after an eel had been swallowed.
Out through the gills
The escaping behavior of a Japanese eel. Credit: Hasegawa et al./Current Biology
O. obscura swallow their prey whole along with surrounding water, and a swallowed eel quickly ends up in the digestive tract, a highly acidic and oxygen-deprived environment that kills the eels within 211.9 seconds (a little over three minutes). Thirty-two of the eels were eaten, and of those, 13 (or 40.6 percent) managed to poke at least their tails through the gills of their predator. Of those 13, nine (69.2 percent) escaped completely within 56 seconds on average, suggesting “that the period until the tails emerge from the predator’s gill is particularly crucial for successful escape,” the authors wrote. The final push for freedom involved coiling their bodies to extract their head from the gill.
It helps to be swallowed head-first. The researchers discovered that most captured eels tried to escape by swimming back up the digestive tract toward the esophagus and gills, tail-first in the cases where escape was successful. However, eleven eels ended up completely inside the stomach and resorted to swimming around in circles—most likely looking for a possible escape route. Five of those managed to insert their tails correctly toward the esophagus, while two perished because they oriented their tails in the wrong direction.
“The most surprising moment in this study was when we observed the first footage of eels escaping by going back up the digestive tract toward the gill of the predatory fish,” said co-author Yuuki Kawabata, also of Nagasaki University. “At the beginning of the experiment, we speculated that eels would escape directly from the predator’s mouth to the gill. However, contrary to our expectations, witnessing the eels’ desperate escape from the predator’s stomach to the gills was truly astonishing for us.”
In an age when you can get just about anything online, it’s probably no surprise that you can buy a diamond-making machine for $200,000 on Chinese eCommerce site Alibaba. If, like me, you haven’t been paying attention to the diamond industry, it turns out that the availability of these machines reflects an ongoing trend toward democratizing diamond production—a process that began decades ago and continues to evolve.
The history of lab-grown diamonds dates back at least half a century. According to Harvard graduate student Javid Lakha, writing in a comprehensive piece on lab-grown diamonds published in Works in Progress last month, the first successful synthesis of diamonds in a laboratory setting occurred in the 1950s. Lakha recounts how Howard Tracy Hall, a chemist at General Electric, created the first lab-grown diamonds using a high-pressure, high-temperature (HPHT) process that mimicked the conditions under which diamonds form in nature.
Since then, diamond-making technology has advanced significantly. Today, there are two primary methods for creating lab-grown diamonds: the HPHT process and chemical vapor deposition (CVD). Both types of machines are now listed on Alibaba, with prices starting at around $200,000, as pointed out in a Hacker News comment by engineer John Nagle (who goes by “Animats” on Hacker News). A CVD machine we found is more pricey, at around $450,000.
An image of a “HPHT Cubic Press Synthetic Diamond Making Machine” made by Henan Huanghe Whirlwind Co., Ltd. in China.
A photo of part of a “HPHT Cubic Press Synthetic Diamond Making Machine” made by Henan Huanghe Whirlwind Co., Ltd. in China.
A photo of a factory full of HPHT Cubic Press Synthetic Diamond Making Machines, made by Henan Huanghe Whirlwind Co., Ltd. in China.
Not a simple operation
While the idea of purchasing a diamond-making machine on Alibaba might be intriguing, it’s important to note that operating one isn’t as simple as plugging it in and watching diamonds form. According to Lakha’s article, these machines require significant expertise and additional resources to operate effectively.
For an HPHT press, you’d need a reliable source of high-quality graphite, metal catalysts like iron or cobalt, and precise temperature and pressure control systems. CVD machines require a steady supply of methane and hydrogen gases, as well as the ability to generate and control microwaves or hot filaments. Both methods need diamond seed crystals to start the growth process.
Moreover, you’d need specialized knowledge to manage the growth parameters, handle potentially hazardous materials and high-pressure equipment safely, and process the resulting raw diamonds into usable gems or industrial components. The machines also use considerable amounts of energy and require regular maintenance. Those factors may make the process subject to some regulations that are far beyond the scope of this piece.
In short, while these machines are more accessible than ever, turning one into a productive diamond-making operation would still require significant investment in equipment, materials, expertise, and safety measures. But hey, a guy can dream, right?
These are the new colors and finishes for the iPhone 16 Pro.
The screens are slightly larger this time around.
Apple
As expected, Apple announced the new iPhone Pro models today during a livestream: the iPhone 16 Pro and iPhone 16 Pro Max. The iPhone 16 Pro has a 6.3-inch display, and the Max has a 6.9-inch display. That’s primarily thanks to thinner borders around the displays.
Like the iPhone 15 Pro, the 16 Pro is made of titanium but with a new texture. Apple claims the phone has improved heat management with its new chassis, which could address some of our complaints about the iPhone 15 Pro—that means up to 20 percent faster sustained performance, too.
Larger batteries and efficiency improvements have led to a promise of battery life improvements, though Apple didn’t say exactly how much longer they’ll last during the livestream.
The iPhone 16 Pro includes the new A18 Pro chip, which is distinct from the A18 found in the regular iPhone 16. Apple says it is faster and more efficient.
It has a 16-core Neural Engine with 17 percent more memory bandwidth. Apple Intelligence features are said to run up to 15 percent faster than on the previous Pro phones. The A18 Pro ships with a 6-core GPU with 20 percent faster performance, and Apple touted its capability for AAA games—and that includes ray tracing performance that’s twice as fast. The 6-core CPU (two performance cores, four efficiency) is a modest 15 percent faster. Alternatively, it can deliver the same performance as the A17 Pro but with 20 percent more efficiency, which suggests battery life and heat improvements. Finally, there’s a new video encoder and ISP, with two times the throughput for data, with a special emphasis on improving video capture.
Like the new iPhone 16, the iPhone 16 Pro includes a new button called the Capture button. You can click it to take a photo quickly, like a traditional camera. But it’s also touch-sensitive, so you can run your finger across it in gestures to tweak the image using existing built-in photography features, like adjusting the zoom.
It has the same three camera types as before: wide-angle, telephoto, and ultra-wide. But there are some hardware improvements. The 48-megapixel wide-angle camera adds a new sensor that can read data twice as fast. There’s a new 48-megapixel ultra-wide camera to enable more detail in close-ups and selfies. The 5x telephoto lens that was exclusive to the 15 Pro Max is now included in both sizes of the iPhone 16 Pro, too.
The big new camera feature is 4K video capture at 120 frames per second and in Dolby Vision, which is a first for the platform. Videos captured this way can see their playback speed adjusted between 120 fps, 60 fps, 30 fps, and 24 fps after the fact in the Photos app. All videos captured can now include spatial audio, too. That’s accompanied by Audio Mix, a feature that allows you to switch between modes that attempt to isolate individual voices or sounds according to a few specific mix styles.
iPhone 16 Pro starts at $999 (128GB) or $1,199 (256GB) for the Max size. They are available for pre-order this coming Friday, and they ship on September 20.
Enlarge/ Boeing’s Starliner spacecraft after landing Friday night at White Sands Space Harbor, New Mexico.
Boeing
Boeing’s Starliner spacecraft sailed to a smooth landing in the New Mexico desert Friday night, an auspicious end to an otherwise disappointing three-month test flight that left the capsule’s two-person crew stuck in orbit until next year.
Cushioned by airbags, the Boeing crew capsule descended under three parachutes toward an on-target landing at 10: 01 pm local time Friday (12: 01 am EDT Saturday) at White Sands Space Harbor, New Mexico. From the outside, the landing appeared just as it would have if the spacecraft brought home NASA astronauts Butch Wilmore and Suni Williams, who became the first people to launch on a Starliner capsule on June 5.
But Starliner’s cockpit was empty as it flew back to Earth Friday night. Last month, NASA managers decided to keep Wilmore and Williams on the International Space Station (ISS) until next year after agency officials determined it was too risky for the astronauts to return to the ground on Boeing’s spaceship. Instead of coming home on Starliner, Wilmore and Williams will fly back to Earth on a SpaceX Dragon spacecraft in February. NASA has incorporated the Starliner duo into the space station’s long-term crew.
The Starliner spacecraft began the journey home by backing away from its docking port at the space station at 6: 04 pm EDT (22: 04 UTC), one day after astronauts closed hatches to prepare for the ship’s departure. The capsule fired thrusters to quickly back away from the complex, setting up for a deorbit burn to guide Starliner on a trajectory toward its landing site. Then, Starliner jettisoned its disposable service module to burn up over the Pacific Ocean, while the crew module, with a vacant cockpit, took aim on New Mexico.
After streaking through the atmosphere over the Pacific Ocean and Mexico, Starliner deployed three main parachutes to slow its descent, then a ring of six airbags inflated around the bottom of the spacecraft to dampen the jolt of touchdown. This was the third time a Starliner capsule has flown in space, and the second time the spacecraft fell short of achieving all of its objectives.
Not the desired outcome
“I’m happy to report Starliner did really well today in the undock, deorbit, and landing sequence,” said Steve Stich, manager of NASA’s commercial crew program, which manages a contract worth up to $4.6 billion for Boeing to develop, test, and fly a series of Starliner crew missions to the ISS.
While officials were pleased with Starliner’s landing, the celebration was tinged with disappointment.
“From a human perspective, all of us feel happy about the successful landing, but then there’s a piece of us that we wish it would have been the way we had planned it,” Stich said. “We had planned to have the mission land with Butch and Suni onboard. I think there are, depending on who you are on the team, different emotions associated with that, and I think it’s going to take a little time to work through that.”
Nevertheless, Stich said NASA made the right call last month when officials decided to complete the Starliner test flight without astronauts in the spacecraft.
“We made the decision to have an uncrewed flight based on what we knew at the time, and based on our knowledge of the thrusters and based on the modeling that we had,” Stich said. “If we’d had a model that would have predicted what we saw tonight perfectly, yeah, it looks like an easy decision to go say, ‘We could have had a crew tonight.’ But we didn’t have that.”
Boeing’s Starliner managers insisted the ship was safe to bring the astronauts home. It might be tempting to conclude the successful landing Friday night vindicated Boeing’s views on the thruster problems. However, he spacecraft’s propulsion system, provided by Aerojet Rocketdyne, clearly did not work as intended during the flight. NASA had the option of bringing Wilmore and Williams back to Earth on a different, flight-proven spacecraft, so they took it.
“It’s awfully hard for the team,” Stich said. “It’s hard for me, when we sit here and have a successful landing, to be in that position. But it was a test flight, and we didn’t have confidence, with certainty, of the thruster performance.”
Enlarge/ In this infrared view, Starliner descends under its three main parachutes moments before touchdown at White Sands Space Harbor, New Mexico.
NASA
As Starliner approached the space station in June, five of 28 control thrusters on Starliner’s service module failed, forcing Wilmore to take manual control as ground teams sorted out the problem. Eventually, engineers recovered four of the five thrusters, but NASA’s decision makers were unable to convince themselves the same problem wouldn’t reappear, or get worse, when the spacecraft departed the space station and headed for reentry and landing.
Engineers later determined the control jets lost thrust due to overheating, which can cause Teflon seals in valves to swell and deform, starving the thrusters of propellant. Telemetry data beamed back to the mission controllers from Starliner showed higher-than-expected temperatures on two of the service module thrusters during the flight back to Earth Friday night, but they continued working.
Ground teams also detected five small helium leaks on Starliner’s propulsion system soon after its launch in June. NASA and Boeing officials were aware of one of the leaks before the launch, but decided to go ahead with the test flight. Starliner was still leaking helium when the spacecraft undocked from the station Friday, but the leak rate remained within safety tolerances, according to Stich.
A couple of fresh technical problems cropped up as Starliner cruised back to Earth. One of 12 control jets on the crew module failed to ignite at any time during Starliner’s flight home. These are separate thrusters from the small engines that caused trouble earlier in the Starliner mission. There was also a brief glitch in Starliner’s navigation system during reentry.
Where to go from here?
Three NASA managers, including Stich, took questions from reporters in a press conference early Saturday following Starliner’s landing. Two Boeing officials were also supposed to be on the panel, but they canceled at the last minute. Boeing didn’t explain their absence, and the company has not made any officials available to answer questions since NASA chose to end the Starliner test flight without the crew aboard.
“We view the data and the uncertainty that’s there differently than Boeing does,” said Jim Free, NASA’s associate administrator, in an August 24 press conference announcing the agency’s decision on how to end the Starliner test flight. It’s unusual for NASA officials to publicly discuss how their opinions differ from those of their contractors.
Joel Montalbano, NASA’s deputy associate administrator for space operations, said Saturday that Boeing deferred to the agency to discuss the Starliner mission in the post-landing press conference.
Here’s the only quote from a Boeing official on Starliner’s return to Earth. It came in the form of a three-paragraph written statement Boeing emailed to reporters about a half-hour after Starliner’s landing: “I want to recognize the work the Starliner teams did to ensure a successful and safe undocking, deorbit, re-entry and landing,” said Mark Nappi, vice president and program manager of Boeing’s commercial crew program. “We will review the data and determine the next steps for the program.”
Nappi’s statement doesn’t answer one of the most important questions reporters would have asked anyone from Boeing if they participated in Saturday morning’s press conference: Does Boeing still have a long-term commitment to the Starliner program?
So far, the only indications of Boeing’s future plans for Starliner have come from second-hand anecdotes relayed by NASA officials. Boeing has been silent on the matter. The company has reported nearly $1.6 billion in financial charges to pay for previous delays and cost overruns on the Starliner program, and Boeing will again be on the hook to pay to fix the problems Starliner encountered in space over the last three months.
Montalbano said Boeing’s Starliner managers met with ground teams at mission control in Houston following the craft’s landing. “The Boeing managers came into the control room and congratulated the team, talked to the NASA team, so Boeing is committed to continue their work with us,” he said.
Enlarge/ Boeing’s Starliner spacecraft fires thrusters during departure from the International Space Station on Friday.
NASA
NASA isn’t ready to give up on Starliner. A fundamental tenet of NASA’s commercial crew program is to foster the development of two independent vehicles to ferry astronauts to and from the International Space Station, and eventually commercial outposts in low-Earth orbit. NASA awarded multibillion-dollar contracts to Boeing and SpaceX in 2014 to complete development of their Starliner and Crew Dragon spaceships.
SpaceX’s Dragon started flying astronauts in 2020. NASA would like to have another US spacecraft for crew rotation flights to support the ISS. If Boeing had more success with this Starliner test flight, NASA expected to formally certify the spacecraft for operational crew flights beginning next year. Once that happens, Starliner will enter a rotation with SpaceX’s Dragon to transport crews to and from the station in six-month increments.
Stich said Saturday that NASA has not determined whether the agency will require Boeing launch another Starliner test flight before certifying the spacecraft for regular crew rotation missions. “It’ll take a little time to determine the path forward, but today we saw the vehicle perform really well,” he said.
On to Starliner-1?
But some of Stich’s other statements Saturday suggested NASA would like to proceed with certifying Starliner and flying the next mission with a full crew complement of four astronauts. NASA calls Boeing’s first operational crew mission Starliner-1. It’s the first of at least three and potentially up to six crew rotation missions on Boeing’s contract.
“It’s great to have the spacecraft back, and we’re now focused on Starliner-1,” Stich said.
Before that happens, NASA and Boeing engineers must resolve the thruster problems and helium leaks that plagued the test flight this summer. Stich said teams are studying several ways to improve the reliability of Starliner’s thrusters, including hardware modifications and procedural changes. This will probably push back the next crew flight of Starliner, whether it’s Starliner-1 or another test flight, until the end of next year or 2026, although NASA officials have not laid out a schedule.
The overheating thrusters are located inside four doghouse-shaped propulsion pods around the perimeter of Starliner’s service module. It turns out the doghouses retain heat like a thermos—something NASA and Boeing didn’t fully appreciate before this mission—and the thrusters don’t have time to cool down when the spacecraft fires its control jets in rapid pulses. It might help if Boeing removes some of the insulating thermal blankets from the doghouses, Stich said.
The easiest method of resolving the problem of Starliner’s overheating thrusters would be to change the rate and duration of thruster firings.
“What we would like to do is try not to change the thruster. I think that is the best path,” Stich said. “There thrusters have shown resilience and have shown that they perform well, as long as we keep their temperatures down and don’t fire them in a manner that causes the temperatures to go up.”
There’s one thing from this summer’s test flight that might, counterintuitively, help NASA certify the Starliner spacecraft to begin operational flights with its next mission. Rather than staying at the space station for eight days, Starliner remained docked at the research lab for three months, half of the duration of a full-up crew rotation flight. Despite the setbacks, Stich estimated the test flight achieved about 85 to 90 percent of its objectives.
“There’s a lot of learning that happens in that three months that is invaluable for an increment mission,” Stich said. “So, in some ways, the mission overachieved some objectives, in terms of being there for extra time. Not having the crew onboard, obviously, there are some things that we lack in terms of Butch and Suni’s test pilot expertise, and how the vehicle performed, what they saw in the cockpit. We won’t have that data, but we still have the wealth of data from the spacecraft itself, so that will go toward the mission objectives and the certification.”
Enlarge/ The First Combat of Gav and Talhand’, Folio from a Shahnama (Book of Kings), ca. 1330–40, Attributed to Iran, probably Isfahan, Ink, opaque watercolor, gold, and silver on paper, Page: 8 1/16 x 5 1/4 in. (20.5 x 13.3 cm), Codices, Three battles between two Indian princes – half brothers contending for the throne – resulted in the invention of the game of chess, to explain the death of one of them to their grieving mother. The Persian word shah mat, or checkmate, indicating a position of no escape, describes the plight of Talhand at the end of the third battle. (Photo by: Sepia Times/Universal Images Group via Getty Images)
The three princes of Sarandib—an ancient Persian name for Sri Lanka—get exiled by their father the king. They are good boys, but he wants them to experience the wider world and its peoples and be tested by them before they take over the kingdom. They meet a cameleer who has lost his camel and tell him they’ve seen it—though they have not—and prove it by describing three noteworthy characteristics of the animal: it is blind in one eye, it has a tooth missing, and it has a lame leg.
After some hijinks the camel is found, and the princes are correct. How could they have known? They used their keen observational skills to notice unusual things, and their wit to interpret those observations to reveal a truth that was not immediately apparent.
It is a very old tale, sometimes involving an elephant or a horse instead of a camel. But this is the version written by Amir Khusrau in Delhi in 1301 in his poem The Eight Tales of Paradise, and this is the version that one Christopher the Armenian clumsily translated into the Venetian novel The Three Princes of Serendip, published in 1557; a publication that, in a roundabout way, brought the word “serendipity” into the English language.
In no version of the story do the princes accidentally stumble across something important they were not looking for, or find something they were looking for but in a roundabout, unanticipated manner, or make a valuable discovery based on a false belief or misapprehension. Chance, luck, and accidents, happy or otherwise, play no role in their tale. Rather, the trio use their astute observations as fodder for their abductive reasoning. Their main talent is their ability to spot surprising, unexpected things and use their observations to formulate hypotheses and conjectures that then allow them to deduce the existence of something they’ve never before seen.
Defining serendipity
This is how Telmo Pievani, the first Italian chair of Philosophy of Biological Sciences at the University of Padua, eventually comes to define serendipity in his new book, Serendipity: the Unexpected in Science. It’s hardly a mind-bending or world-altering read, but it is a cute and engaging one, especially when his many stories of discovery veer into ruminations on the nature of inquiry and of science itself.
He starts with the above-mentioned romp through global literature, culminating in the joint coining and misunderstanding of the term as we know it today: in 1754, after reading the popular English translation entitled The Travels and Adventures of Three Princes of Serendip, the intellectual Horace Walpole described “Serendipity, a very expressive word,” as “discoveries, by accidents and sagacity, of things which they were not in quest of.”
Pievani knows a lot, but like a lot, about the history of science, and he puts it on display here. He quickly debunks all of the instances of alleged serendipity that are always trotted out: Fleming the microbiologist had been studying antibiotics and searching for a commercially viable one for years before his moldy plate led him to penicillin. Yes, Röntgen discovered X-rays by a fluke, but it was only because of the training he received in his studies of cathode rays that he recognized he was observing a new form of radiation. Plenty of people over the course of history splashed some volume of water out of the baths they were climbing into and watched apples fall, but only Archimedes—who had recently been tasked by his king to figure out if his crown was made entirely of gold—and Newton—polymathic inventor of calculus—leapt from these (probably apocryphal) mundane occurrences to their famous discoveries of density and gravity, respectively.
After dispensing with these tired old saws, Pievani then suggests some cases of potentially real—or strong, as he deems it—serendipity. George de Mestral’s inventing velcro after noticing burrs stuck to his pants while hiking in the Alps; he certainly wasn’t searching for anything, and he parlayed his observation into a useful technology. DuPont chemists’ developing nylon, Teflon, and Post-it notes while playing with polymers for assorted other purposes. Columbus “discovering” the Americas (for the fourth time) since he thought the Earth was about a third smaller than Eratosthenes of Cyrene had correctly calculated it to be almost two thousand years earlier, forgotten “due to memory loss and Eurocentric prejudices.”
