In his post, Brooks recounts being “way too close” to an Agility Robotics Digit humanoid when it fell several years ago. He has not dared approach a walking one since. Even in promotional videos from humanoid companies, Brooks notes, humans are never shown close to moving humanoid robots unless separated by furniture, and even then, the robots only shuffle minimally.
This safety problem extends beyond accidental falls. For humanoids to fulfill their promised role in health care and factory settings, they need certification to operate in zones shared with humans. Current walking mechanisms make such certification virtually impossible under existing safety standards in most parts of the world.
The humanoid Apollo robot. Credit: Google
Brooks predicts that within 15 years, there will indeed be many robots called “humanoids” performing various tasks. But ironically, they will look nothing like today’s bipedal machines. They will have wheels instead of feet, varying numbers of arms, and specialized sensors that bear no resemblance to human eyes. Some will have cameras in their hands or looking down from their midsections. The definition of “humanoid” will shift, just as “flying cars” now means electric helicopters rather than road-capable aircraft, and “self-driving cars” means vehicles with remote human monitors rather than truly autonomous systems.
The billions currently being invested in forcing today’s rigid, vision-only humanoids to learn dexterity will largely disappear, Brooks argues. Academic researchers are making more progress with systems that incorporate touch feedback, like MIT’s approach using a glove that transmits sensations between human operators and robot hands. But even these advances remain far from the comprehensive touch sensing that enables human dexterity.
Today, few people spend their days near humanoid robots, but Brooks’ 3-meter rule stands as a practical warning of challenges ahead from someone who has spent decades building these machines. The gap between promotional videos and deployable reality remains large, measured not just in years but in fundamental unsolved problems of physics, sensing, and safety.
Still, the coincidence between Altman’s statement and the MIT report reportedly spooked tech stock investors earlier in the week, who have already been watching AI valuations climb to extraordinary heights. Palantir trades at 280 times forward earnings. During the dot-com peak, ratios of 30 to 40 times earnings marked bubble territory.
The apparent contradiction in Altman’s overall message is notable. This isn’t how you’d expect a tech executive to talk when they believe their industry faces imminent collapse. While warning about a bubble, he’s simultaneously seeking a valuation that would make OpenAI worth more than Walmart or ExxonMobil—companies with actual profits. OpenAI hit $1 billion in monthly revenue in July but is reportedly heading toward a $5 billion annual loss. So what’s going on here?
Looking at Altman’s statements over time reveals a potential multi-level strategy. He likes to talk big. In February 2024, he reportedly sought an audacious $5 trillion–7 trillion for AI chip fabrication—larger than the entire semiconductor industry—effectively normalizing astronomical numbers in AI discussions.
By August 2025, while warning of a bubble where someone will lose a “phenomenal amount of money,” he casually mentioned that OpenAI would “spend trillions on datacenter construction” and serve “billions daily.” This creates urgency while potentially insulating OpenAI from criticism—acknowledging the bubble exists while positioning his company’s infrastructure spending as different and necessary. When economists raised concerns, Altman dismissed them by saying, “Let us do our thing,” framing trillion-dollar investments as inevitable for human progress while making OpenAI’s $500 billion valuation seem almost small by comparison.
This dual messaging—catastrophic warnings paired with trillion-dollar ambitions—might seem contradictory, but it makes more sense when you consider the unique structure of today’s AI market, which is absolutely flush with cash.
A different kind of bubble
The current AI investment cycle differs from previous technology bubbles. Unlike dot-com era startups that burned through venture capital with no path to profitability, the largest AI investors—Microsoft, Google, Meta, and Amazon—generate hundreds of billions of dollars in annual profits from their core businesses.
MIT graduate student Alex Kachkine once spent nine months meticulously restoring a damaged baroque Italian painting, which left him plenty of time to wonder if technology could speed things up. Last week, MIT News announced his solution: a technique that uses AI-generated polymer films to physically restore damaged paintings in hours rather than months. The research appears in Nature.
Kachkine’s method works by printing a transparent “mask” containing thousands of precisely color-matched regions that conservators can apply directly to an original artwork. Unlike traditional restoration, which permanently alters the painting, these masks can reportedly be removed whenever needed. So it’s a reversible process that does not permanently change a painting.
“Because there’s a digital record of what mask was used, in 100 years, the next time someone is working with this, they’ll have an extremely clear understanding of what was done to the painting,” Kachkine told MIT News. “And that’s never really been possible in conservation before.”
Figure 1 from the paper. Credit: MIT
Nature reports that up to 70 percent of institutional art collections remain hidden from public view due to damage—a large amount of cultural heritage sitting unseen in storage. Traditional restoration methods, where conservators painstakingly fill damaged areas one at a time while mixing exact color matches for each region, can take weeks to decades for a single painting. It’s skilled work that requires both artistic talent and deep technical knowledge, but there simply aren’t enough conservators to tackle the backlog.
The mechanical engineering student conceived the idea during a 2021 cross-country drive to MIT, when gallery visits revealed how much art remains hidden due to damage and restoration backlogs. As someone who restores paintings as a hobby, he understood both the problem and the potential for a technological solution.
To demonstrate his method, Kachkine chose a challenging test case: a 15th-century oil painting requiring repairs in 5,612 separate regions. An AI model identified damage patterns and generated 57,314 different colors to match the original work. The entire restoration process reportedly took 3.5 hours—about 66 times faster than traditional hand-painting methods.
Alex Kachkine, who developed the AI-printed film technique. Credit: MIT
Notably, Kachkine avoided using generative AI models like Stable Diffusion or the “full-area application” of generative adversarial networks (GANs) for the digital restoration step. According to the Nature paper, these models cause “spatial distortion” that would prevent proper alignment between the restored image and the damaged original.
Enlarge/ Nathan Benaich of Air Street Capital delivers the opening presentation on the state of AI at EmTech Digital 2024 on May 22, 2024.
Benj Edwards
CAMBRIDGE, Massachusetts—On Wednesday, AI enthusiasts and experts gathered to hear a series of presentations about the state of AI at EmTech Digital 2024 on the Massachusetts Institute of Technology’s campus. The event was hosted by the publication MIT Technology Review. The overall consensus is that generative AI is still in its very early stages—with policy, regulations, and social norms still being established—and its growth is likely to continue into the future.
I was there to check the event out. MIT is the birthplace of many tech innovations—including the first action-oriented computer video game—among others, so it felt fitting to hear talks about the latest tech craze in the same building that hosts MIT’s Media Lab on its sprawling and lush campus.
EmTech’s speakers included AI researchers, policy experts, critics, and company spokespeople. A corporate feel pervaded the event due to strategic sponsorships, but it was handled in a low-key way that matches the level-headed tech coverage coming out of MIT Technology Review. After each presentation, MIT Technology Review staff—such as Editor-in-Chief Mat Honan and Senior Reporter Melissa Heikkilä—did a brief sit-down interview with the speaker, pushing back on some points and emphasizing others. Then the speaker took a few audience questions if time allowed.
Enlarge/ EmTech Digital 2024 took place in building E14 on MIT’s Campus in Cambridge, MA.
Benj Edwards
The conference kicked off with an overview of the state of AI by Nathan Benaich, founder and general partner of Air Street Capital, who rounded up news headlines about AI and several times expressed a favorable view toward defense spending on AI, making a few people visibly shift in their seats. Next up, Asu Ozdaglar, deputy dean of Academics at MIT’s Schwarzman College of Computing, spoke about the potential for “human flourishing” through AI-human symbiosis and the importance of AI regulation.
Kari Ann Briski, VP of AI Models, Software, and Services at Nvidia, highlighted the exponential growth of AI model complexity. She shared a prediction from consulting firm Gartner research that by 2026, 50 percent of customer service organizations will have customer-facing AI agents. Of course, Nvidia’s job is to drive demand for its chips, so in her presentation, Briski painted the AI space as an unqualified rosy situation, assuming that all LLMs are (and will be) useful and reliable, despite what we know about their tendencies to make things up.
The conference also addressed the legal and policy aspects of AI. Christabel Randolph from the Center for AI and Digital Policy—an organization that spearheaded a complaint about ChatGPT to the FTC last year—gave a compelling presentation about the need for AI systems to be human-centered and aligned, warning about the potential for anthropomorphic models to manipulate human behavior. She emphasized the importance of demanding accountability from those designing and deploying AI systems.
Asu Ozdaglar, deputy dean of Academics at MIT’s Schwarzman College of Computing, spoke about the potential for “human flourishing” through AI-human symbiosis at EmTech Digital on May 22, 2024.
Benj Edwards
Asu Ozdaglar, deputy dean of Academics at MIT’s Schwarzman College of Computing spoke with MIT Technology Review Editor-in-Chief Mat Honan at EmTech Digital on May 22, 2024.
Benj Edwards
Kari Ann Briski, VP of AI Models, Software, and Services at NVIDIA, highlighted the exponential growth of AI model complexity at EmTech Digital on May 22, 2024.
Benj Edwards
MIT Technology Review Senior Reporter Melissa Heikkilä introduces a speaker at EmTech Digital on May 22, 2024.
Benj Edwards
After her presentation, Christabel Randolph from the Center for AI and Digital Policy sat with MIT Technology Review Senior Reporter Melissa Heikkilä at EmTech Digital on May 22, 2024.
Benj Edwards
Lawyer Amir Ghavi provided an overview of the current legal landscape surrounding AI at EmTech Digital on May 22, 2024.
Benj Edwards
Lawyer Amir Ghavi provided an overview of the current legal landscape surrounding AI at EmTech Digital on May 22, 2024.
Benj Edwards
Amir Ghavi, an AI, Tech, Transactions, and IP partner at Fried Frank LLP, who has defended AI companies like Stability AI in court, provided an overview of the current legal landscape surrounding AI, noting that there have been 24 lawsuits related to AI so far in 2024. He predicted that IP lawsuits would eventually diminish, and he claimed that legal scholars believe that using training data constitutes fair use. He also talked about legal precedents with photocopiers and VCRs, which were both technologies demonized by IP holders until courts decided they constituted fair use. He pointed out that the entertainment industry’s loss on the VCR case ended up benefiting it by opening up the VHS and DVD markets, providing a brand new revenue channel that was valuable to those same companies.
In one of the higher-profile discussions, Meta President of Global Affairs Nick Clegg sat down with MIT Technology Review Executive Editor Amy Nordrum to discuss the role of social media in elections and the spread of misinformation, arguing that research suggests social media’s influence on elections is not as significant as many believe. He acknowledged the “whack-a-mole” nature of banning extremist groups on Facebook and emphasized the changes Meta has undergone since 2016, increasing fact-checkers and removing bad actors.
