←
Ben Brooks801 posts
Ben Brooks avatar

Ben Brooks

@opensauceAI

Policy @bfl_ai. Affiliate @BKCHarvard. ex-Stability AI (weights), GoogleX (drones), Uber (rides), Coinbase (magic beans). Views my own

◷ Joined April 2023▧ 241 Photos and videos
Ben Brooks avatar
Ben Brooks@opensauceAI·2 Sep 2025

Nearly every AI firm is invoking frontier risk to justify a persistent gap between open and closed models. But this application of the precautionary principle deserves more scrutiny than it gets. Restricting access to useful technology—models that will, in their developers' own words, transform the economy—shouldn't be our primary response to uncertain risks. Check it out at @aif_media! There are already a bunch of reasons firms might not release their best models openly: cost recovery, competitive pressure, anxious investors. But we should be skeptical of efforts to freeze open-source behind the frontier under the guise of risk.

Post attachment

AI Frontiers @ai_frontiers_ 2 Sep 2025 Article Precaution Shouldn't Keep Open-Source AI Behind the Frontier Invoking speculative risks to keep our most capable models behind paywalls could create a new form of digital feudalism. Ben Brooks — August 31, 2025 This article originally appeared in AI Frontiers.

Ben Brooks avatar
Ben Brooks@opensauceAI·4h

Lots of zesty recommendations... - "The state should bear the final responsibility for determining whether a developer has met its standards and has mitigated risk acceptably." - "More critically, the legislature should provide the state with standby authority to cap automated AI R&D when the level of risk is unacceptable." - "To the extent that open-weight model providers are relying on automating AI research to improve their models, they must meet the same level of transparency and reporting requirements as other developers, including meaningful access for IVOs." - "Specifically, open-weight frontier models should be subject to the same set of evaluation, verification, and auditing as closed-weight models." ...but generally promising: - "In general, the state should take care not to impose obligations that are impractical for open-weight developers, but where appropriate, maintain basic and foundational transparency and accountability requirements without drawing distinctions on the basis of model release." govops.ca.gov/wp-content/upl…

Post attachmentPost attachment
Ben Brooks avatar
Ben Brooks@opensauceAI·5h

The California AI working group on kill switches has finally acknowledged what open advocates have been saying for years! "For open-weight developers, [shutdown] requirements should only apply to model deployments within their control."

Post attachment
Ben Brooks avatar
Ben Brooks@opensauceAI·Nov 16

Today's rule on remote access to advanced chips (1) formally revokes the Biden Admin's controls for model weights under the Diffusion Rule, incl. the Foreign Direct Product rule for models 👍 but (2) alludes to future destination-based or end-use controls on models 👎. My monthly reminder that controls for intangible technology are still in flux; the Biden rule was actually quite favorable to open weights; and there is a real risk the Megaspeed panic produces a framework that makes capable open weights impossible.

Post attachment
Robin Rombach avatar
⟳ Ben Brooks retweeted
Robin Rombach@robrombach·Nov 16

SOTA, again 😍

Post attachment

Today we are previewing FLUX-mimic-x, our new frontier Video-Action Model. We achieve the #1 spot on the RoboLab benchmark among real-time "system 1" models, as well as impressive zero-shot generalization across embodiments including grippers and dexterous hands. FLUX-mimic-x is a 7B-parameter model based on our updated mimic-video recipe and trained on top of FLUX 3 by @bfl_ai. The model is never trained on simulation data or any other data from the RoboLab embodiment beyond that of the original DROID dataset. Our model can be deployed on any robot and any task with minimal adaptation, running on-device and in real-time.

1:03
Ben Brooks avatar
Ben Brooks@opensauceAI·Nov 14

I can't stress enough how important it is that open weights are understood as instruments of the First Amendment, not the Second.

Open-source AI is like the Second Amendment. It protects everyone's last line of defense in the AGI era. Don't be fooled by cult-like narratives: "We're the good god, and only we should be allowed to develop AI." @JensenHuang has done his best to protect everyone's freedom and right to choose. Just as the Second Amendment serves as a safeguard against government tyranny, open-source AI can serve as a safeguard against the concentration and abuse of AI power.

Ben Brooks avatar
Ben Brooks@opensauceAI·Nov 13

I appreciate the catastrophic risk lawsuit filed today in SF Superior Court. "They say they might kill everyone, and nothing can be worth that risk." A fair and internally consistent argument.

Post attachment
Ben Brooks avatar
Ben Brooks@opensauceAI·Nov 12

The real question is: - if good open-weight models rely on SOTA chips, and - if we agree unauthorized distillation is increasingly unimportant / never really mattered for these firms - why haven't US labs with ample compute pushed the global open-weight frontier since ~2024? We need to spend more time on policy that incentivizes US labs to share good models openly, not just policy that hobbles competitors.

I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo…. There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.

Post attachment
Ben Brooks avatar
Ben Brooks@opensauceAI·Nov 11

Each year, Americans spend 4,300 human lifetimes stuck on hold. The idea that quasi-utilities like airlines, telcos, and hospitals can refuse to interact with agents while wasting our lives on terrible UX will be outrageous in the near future. Compelling access for agents to these platforms will be one of the most important AI reforms for everyday consumers. You solve the obvious issues with good design, remediation, and insurance.

All personal AI assistants have been banned from United Airlines systems, and from booking flights.

Post attachment
Ben Brooks avatar
Ben Brooks@opensauceAI·Nov 11

We were glad to contribute to Open Future's discussion paper today on a framework for sharing the value of AI training in Europe, including through a statutory levy. These are sensitive issues that engage deeply held feelings about AI. But it's important that we're able to work through these issues together. Our @bfl_ai goal is to find a path forward that supports both competitive AI development and the long-term sustainability of the cultural commons. Check out the full paper below as the starting point for a much bigger conversation ⬇

Post attachment
Ben Brooks avatar
Ben Brooks@opensauceAI·Nov 4

The spluttering indignation over the data center backlash is bewildering to me. Frontier labs and their thinkfluencers have spent ~2 electoral cycles telling anyone who cares to listen that frontier capability will at best upend the economy and at worst kill everyone.

Ben Brooks avatar
Ben Brooks@opensauceAI·Oct 26

If this technology will "rapidly transform every sector of the global economy in ways greater than the impact of industrialization, electricity, semiconductors, and the internet"... then the availability and adoption of open alternatives will matter more than ever before.

Ben Brooks avatar
Ben Brooks@opensauceAI·Oct 26

Open-weight deployers value cost efficiency *and* intelligence. They also value transparency, composability, modifiability, security, privacy, and operational independence.

Post attachment
Ben Brooks avatar
Ben Brooks@opensauceAI·Oct 22

A fun night at @github HQ fighting for the proposition that democratizing AI means owning, not just renting, intelligence. In a ¡shocking! turn of events, open source ultimately prevailed in the Open Source AI Week Debate 😂 Thanks for bringing us together @matthew_d_white, @ekang426, Margaret Tucker, and crew. Glad to link arms with @istoica05 and @percyliang to defend openness—and big kudos to Cheryl Cheng @M12vc, @m_bourgon, and @DavidRBellamy for taking on the GH audience with aplomb!

Post attachmentPost attachmentPost attachment
Arvind Narayanan avatar
⟳ Ben Brooks retweeted
Arvind Narayanan@random_walker·Oct 20

Is it "weird" or does he maybe know what he's talking about? Anyway, for those who are open to the idea that there's a well-thought-out AI-as-normal-technology perspective on AI safety and the recent loss-of-control incidents, see my essay with @sayashk, or my interview with Ezra Klein, or the writing of many others with similar views including @joshua_saxe and @natolambert. normaltech.ai/p/the-ai-as-no…

It’s weird that the CEO of Hugging Face continues to be an “AI is normal technology” guy despite being the victim of an autonomous cyberattack by a swarm of rogue agents, arguably the least normal thing AI has done so far.