Zuckerberg’s “Scorched Earth” LLaMA AI Strategy – Zuck just messed with the deep state on AI and the AI Manhattan Project

First, I would like to link you to a video I wanted to make before this Zuck gem showed up: 

From The Washington Post:

1:24

But let’s continue. Donald Trump’s allies are drafting a sweeping AI executive order that would launch a series of Manhattan Projects to develop military technology and immediately review unnecessary and burdensome regulations. Seeing how a second Trump term may affect AI policies, certainly, these would be favorable to Silicon Valley investors and companies. They would create industry-led agencies to evaluate AI models and secure systems from foreign adversaries.

For example, if we’re talking about China, it has a long history of successfully reproducing or copying some of the tech developed in the United States. Many people have concerns about what happens if a powerful AI developed in the United States gets copied, stolen, or transferred abroad to a power that may either be unfriendly to the United States or, worse, leak to some rogue nation. The idea is to secure systems from foreign adversaries.

Zuck’s Counter Move

While ChatGPT and other AI companies were lobbying to make AI a closed source and to initiate Manhattan Projects for AI, Zuck created models as good as top models and made them open source. As a result, Chinese entities are now running these models for free. However, he also defended his decision.

20:36

One of the questions that you sometimes hear debated is, if you’re open-sourcing the really advanced models, how do you make sure that it doesn’t get to China or they don’t use it against us? That’s sometimes an argument people have for locking down development. But I think that’s missing a few things.

First, for this to work, the US or the West needs to have an advantage in the first place. Our advantage is basically open and decentralized innovation. It’s not just a small number of big companies or labs; it’s startups, universities, and individuals creating things which aren’t even part of companies. That’s a big part of it, and you don’t want to shut that down. If you do, you increase the chance that we stop leading.

Then, there’s the issue of any government, like China, and the risks of stealing models and espionage. Many models fit on a hard drive that you can quickly put in your backpack. We need to be realistic about how likely it is that any tech company can secure these models long-term against sophisticated efforts to steal them. The situational awareness paper by Leopold Ashenbrener specifically calls out China. The chances of protecting a closed-source model’s weights indefinitely are essentially zero. US companies have been hacked, specifically by China, and their IP stolen. It only needs to happen once; one minor lapse in security, and you lose the model weights.

Let’s assume state actors are going to have these AI models, with the likelihood of a private company like OpenAI getting hacked and having their model weights stolen. So, let’s make sure everybody has them. Let’s ensure all the good guys and bad guys have the same capabilities. Thus, it’s one AI versus another AI. My fear is that if we lock down development, we end up in a world where a small number of companies, plus all the adversaries who can steal the model, are the only ones with access. All the startups, universities, and individual creators are left out and don’t have the ability to do anything.

A realistic aim should be to use open source to develop the leading and most robust ecosystem in the world. We should expect our companies to work closely with our government and allied governments on national security, so our governments can persistently integrate the latest technology and have a six to eight month advantage over our adversaries. In this world, you might not get a 10-year permanent advantage, but a perpetual lead will make us safer than having a small number of closed labs locking down development. We probably risk not being in the lead at all, with other governments getting access to it.

Spreading prosperity more evenly around the world and making progress on safety will likely show that open source leads bring opportunities. There will be issues, and we’ll have to mitigate them by testing everything rigorously and working with governments.

Deep State vs. Open Source

So, the deep state wanted to make Manhattan Projects for AI and make AI closed source, as did most AI companies. Here you have Zuck doing something different. I am asking myself if I agree with him. If we consider what Leopold Aschenbrener wrote in his manifesto, which I wrote about in a post: Ex-OpenAI Employee Reveals Terrifying Future of AI.

Here is a fragment of his long interview:

5:50

They will, and at some point, they will realize that this will be utterly decisive—not just for some proxy war somewhere, but for whether liberal democracy can continue to thrive, and whether the Chinese Communist Party will continue existing. I think that will activate forces that we haven’t seen in a long time.

Zuck’s argument makes some sense, but his coziness to China is a little worrying and explains his position.

Here you have a good historical anecdote from a full interview with Leopold Aschenbrenner:

1:03:29

“Germany during World War II went down the wrong path with heavy water. There’s an amazing anecdote in The Making of the Atomic Bomb about this. Secrecy was one of the most contentious issues early on. Leo Szilard really thought a nuclear chain reaction and an atomic bomb were possible. He went around saying, ‘This is going to be of enormous strategic and military importance.’

A lot of people didn’t believe it or thought, ‘Maybe this is possible, but I’m going to act as though it’s not, and science should be open.’ In the early days, there had been some incorrect measurements made on graphite as a moderator. Germany thought graphite wasn’t going to work, so they had to use heavy water. But then Enrico Fermi made new measurements indicating that graphite would work. This was really important. Szilard assaulted Fermi with another secrecy appeal, and Fermi was angry, throwing a temper tantrum. He thought it was absurd, saying, ‘Come on, this is crazy.’ But Szilard persisted, and they roped in another guy, George Pegram.

In the end, Fermi didn’t publish it. That was just in time. Fermi not publishing meant that the Nazis didn’t figure out graphite would work. They went down the path of heavy water, which was the wrong path. This is a key reason why the German WWII nuclear bomb project didn’t work out. They were way behind. We face a similar situation now. Are we just going to instantly leak how to get past the data wall and what the next paradigm is? Or are we not?”