America's Los Alamos Lab Is Now Investing Heavily In AI For Science

Established in 1943 to coordinate America's building of the first atomic bomb, the Los Alamos National Lab in New Mexico is still "one of the world's largest and most advanced scientific institutions" notes Wikipedia.

And it now has a "National Security AI Office," where senior director Jason Pruet is working to help "prepare for a future in which AI will reshape the landscape of science and security," according to the lab's science and technology magazine 1663. "This year, the Lab invested more in AI-related work than at any point in history..." Pruet: AI is starting to feel like the next great foundation for scientific progress. Big companies are spending billions on large machines, but the buy-in costs of working at the frontiers of AI are so high that no university has the exascale-class machines needed to run the latest AI models. We're at a place now where we, meaning the government, can revitalize that pact by investing in the infrastructure to study AI for the public good... Part of what we're doing with the Lab's machines, like Venado — which has 2500 GPUs — is giving universities access to that scale of computing. The scale is just completely different. A typical university might have 50 or 100 GPUs.

Right now, for example, we have partnerships with the University of California, the University of Michigan, and many other universities where researchers can tap into this infrastructure. That's something we want to expand on. Having university collaboration will be critical if the Department of Energy is going to have a comprehensive AI program at scale that is focused on national security and energy dominance...

There was a time when I wouldn't have advocated for government investment in AI at the scale we're seeing now. But the weight of the evidence has become overwhelming. Large models — "frontier models" — have shown such extraordinary capabilities with recent advances in areas as diverse as hypothesis generation, mathematics, biological design, and complex multiphysics simulations. The potential for transformative impact is too significant to ignore.
"He no longer views the technology as just a tool, but as a fundamental shift in how scientists approach problems and make discoveries," the article concludes.

"The global race humanity is now in... is about how to harness the technology's potential while mitigating its harms."

Thanks to Slashdot reader rabbitface25 — also a Los Alamo Lab science writer — for sharing his article.

I'm leaning to prefer bubble, you?

[Tablizer]

We are either headed face first into a 1930's style bubble, or the Singularity.

Re:I'm leaning to prefer bubble, you?

[rahmrh]

A bubble/hangover from all the money wasted on AI (basically a giant correlation engine) seems more likely than the Singularity.

I don't see any evidence that current AI is much more than the original Eliza program from the 60's on steroids.

And AI is making people stupid because they believe they can rely on it even though any answer it provides is just an answer provided by some other person someplace on the web. At best it is a better search engine except for the fact that it fails to provide a right answer often enough to make is useless.

Re:

[dinfinity]

I don't see any evidence that current AI is much more than the original Eliza program from the 60's on steroids.

You need to open your eyes.

Do this: Try to have a conversation with Eliza about a reasonably advanced topic. Then do it with any of the current frontier models.
Or try this: Do a "Deep Research" request on Google's Gemini (it's free), then let it generate a website about the results for you.

Willful ignorance is not a virtue.

Re: I'm leaning to prefer bubble, you?

[SirSlud]

Calling AI like 60s AI on steroids is like calling a rocketship just a basketball on steroids. The only possible explanation for your post is that your feelings prevent you from discussing the topic in good faith.

Re: I'm leaning to prefer bubble, you?

[Plugh]

Go look up de novo protein design. It has totally transformed over the last 24 months. Same for climate and weather modeling. These sciences are speedrunning from a relative stoneage to advanced predictive ability.

Re:

[Tablizer]

I don't see any evidence that current AI is much more than the original Eliza program from the 60's on steroids.

One of these coming days somebody will figure out how to hook it up to a logic engine, something like CYC, and it will then be able to plot and plan.

Re:

[registrations_suck]

What does this have to do with Los Alamos?

Re:

[ArchieBunker]

I guess that means Obama is eligible in 2028.

Re:

[registrations_suck]

Why not?

AI waffle ..

[Mirnotoriety]

Going on the current versions - I am not impressed.

Re: AI waffle ..

[Plugh]

Have you used claude code? o3? If not then you have not used the latest models. It is like judging a sophomore versus a seasoned professional.

Not LLMs

[q_e_t]

AI is already used successfully in many areas of science, including quantum chemistry. The AI isn't LLMs.

Re: Not LLMs

[Plugh]

LLMs however are proving much more successful than quantum mechanics computations at accurately creating designer proteins

Re:

[q_e_t]

That's a sloppy use of the term LLM. You can train AI (deep networks) on DNA sequences, but they are not language. People should not be using the term LLM so broadly when deep network is more appropriate.