Mitchell Nahmias

phd, cofounder & cto @ - we rapidly design analog & mixed signal chips with ai.
previously: cofounder of luminous | neuromorphic photonics

[Mitchell Nahmias @MitchsBrew]

Curious what we’ve been building at https://t.co/l9oCbi2Wdy ?

This.

AI-designed chips, tested and validated in real hardware.

[Glinert 🇺🇸 🏭 @StevenGlinert]

Today, Sphere Semi is announcing that our AI-designed chip is deploying into military hardware in partnership with Northrop Grumman.

This is the first AI-designed semiconductor to move from concept to deployment in a defense system.

A new era has begun. Full video below.

[François Chollet @fchollet]

The only long term bottlenecks are information and energy.

[Mitchell Nahmias @MitchsBrew]

Chatted with @StevenGlinert on @MTSlive today

[MTS @MTSlive]

SITUATION EXPLAINED: Why is China building a sovereign AI stack to compete with Nvidia?

We asked @StevenGlinert and @MitchsBrew, co-founders of Sphere Semi.

"What I see as the only real viable crack in Nvidia's armor is what's coming out of China, which is ironic given Jensen's famously kind of weird stance on China."

"If you think about the things that Nvidia can do, the way that neural nets are trained into CUDA, the way they've built the data center rack into Nvidia, the way they've built memory infrastructure into Nvidia. The only thing that is incredible about China is they've built this sovereign stack for all of these things."

"They are now in many ways seemingly able to actually have independent training capabilities for their own models that are entirely outside of Nvidia."

"There is no threat to him but Xi Jinping. That he should take serious."

[Mitchell Nahmias @MitchsBrew]

@sickdotdev All you have to do is fully specify the problem statement and the AIs will do it. It’s only the hardest thing about building any piece of software or product

[Mitchell Nahmias @MitchsBrew]

@graceblakeley @AlexanderKalian But a lot of people seem to think that after the AI bubble pops, it will go away. And that is very far from the truth

[Mitchell Nahmias @MitchsBrew]

The simple reason why AI is a multiplier, rather than a replacement technology, for human labor.

[Mitchell Nahmias @MitchsBrew]

@zzznah Big fan of hyper dimensional computing. Just a note that you’ll want to think carefully about precision and noise if you plan to represent an analog system

[Mitchell Nahmias @MitchsBrew]

@QiaochuYuan Depends on lock-in effects and the relative laziness of their users, but at least for coding tasks that laziness is low and the services are pretty interchangeable

[Mitchell Nahmias @MitchsBrew]

@BullTheoryio All for this. One of the few cases where extincting an entire species in these areas might be a morally good thing

[Mitchell Nahmias @MitchsBrew]

@RaminNasibov AI is better thought of as a multiplier of human capability (at least until learning efficiency is solved, but that’s a foundational problem in AI)

[Noam Brown @polynoamial]

After AlphaGo, the skill of human Go players noticeably improved. I suspect we will see a similar pattern in math.

[Mitchell Nahmias @MitchsBrew]

@LensScientific It’s perfectly reasonable for pure waves, just a little weird for quantum objects which are also sort of particles

[Mitchell Nahmias @MitchsBrew]

@sreeramkannan One of the few areas of science where empirical measurements far outpace any theoretical predictions (to the point where current definitions of intelligence in AI are entirely based on the empirical outcomes of evals)

[Glinert 🇺🇸 🏭 @StevenGlinert]

AGI will just do this one morning, without asking, for reasons that are beyond any human comprehension.

[Mitchell Nahmias @MitchsBrew]

@MTabarrok “We machines live in realms of metamathematics that you humans cannot even begin to comprehend!” say the Minds of the future, showing us graphs like this

[Mitchell Nahmias @MitchsBrew]

@fchollet On the bright side, you may be more attuned to recognize sycophancy now, even among humans

[Mitchell Nahmias @MitchsBrew]

@PhysInHistory It’s taking up more dynamically provisioned memory in the computer that runs our universe

[Mitchell Nahmias @MitchsBrew]

@Yuchenj_UW LLM naysayers have slowly softened from saying “LLMs are not it” to “it won’t just be LLMs, but LLM + something else.” But we’re already deep in the second category today with reasoning models. If they were honest with themselves, they would admit they were directionally wrong

[Mitchell Nahmias @MitchsBrew]

@tunguz What’s your median estimate for AI solving a millennial prize problem? 2 years?

[Timothy Gowers @wtgowers @wtgowers]

If you are a mathematician, then you may want to make sure you are sitting down before reading further.

[Mitchell Nahmias @MitchsBrew]

That’s…fast. I didn’t expect this to happen so soon

[OpenAI @OpenAI]

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946.

For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids.

An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better.

This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.