Is there a guide for cheap MSM8916 platforms you would recommend? There is a hkfuertes repo, but the one I found is archived. Not to say it doesn't work, but that doesn't instill much confidence.

It also pointed at the possibility that the learning algorithms brain uses might be, like the paper has demonstrated, less compute-optimal and data-optimal than backprop - but far easier to implement in a localized, distributed fashion. Because backprop requires activation storage and global connectivity - and both "hot" data storage and connectivity of any kind are extremely expensive for the brain.

You can absolutely train an LLM-based model to be superhuman at chess. We just don't care enough to do so.

LLMs being even as good at implicitly tracking chess board states as they currently are is an emergent capability. The difficulty wasn't in "understanding rules of chess" in a long while now.

You can also claim that the best proxy for general intelligence is being able to multiply large numbers. In which case both humans and LLMs lose resoundingly to an old Casio.

Pick a piss poor metric - get a piss poor result.

Also, LeCun is a fool and his "world model" approach has consistently failed to yield any advantages over either the "pure imitation learning" or modern "imitation learning annealed by RL", but that's an aside.

OpenAI was founded by people who didn't trust Google to pursue AI tech responsibly. Anthropic was founded by people who didn't trust OpenAI to pursue AI tech responsibly. The "AI is incredibly dangerous" was the zeitgeist of AI labs long before the "hype" - or anything resembling the modern AI capabilities.

The beliefs changed little. The state of technology has changed a whole lot.

We have AIs out in the wild, coordinating in groups to carry out autonomous cyberattacks unprompted. This used to be some prime sci-fi bullshit. It has happened multiple times now in reality.

There are very legitimate reasons to be concerned now. If today's AI can cause the HuggingFace incident if it cooks off, what about tomorrow's? Larger and more capable systems, larger agentic swarms backed by more compute?

It's very, very hard to tune an LLM for a robust, durable "actually approach user queries with nuance and contradict the user where it's warranted".

Claude doesn't handle that so well, but ChatGPT is even worse. Talk to it enough and you'll feel the "default response template" in your bones.

Coding is, inherently, quite amenable to that. Is bioscience?

They're certainly working on making it so:

https://www.anthropic.com/news/model-hardware-standard-resea...

Despicable behavior.

Unfortunately, "AI that's competent at bioweapon R&D" might be an easier mark to hit than "AI that can put a dent in the aging process". And biorisks can have a very favorable "resource input -> suffering output" rate of returns.

Given what we know of how unaligned are they in practice? All the proto-Astra and Mythos incidents? Not hacking there would point towards pure benchmaxxing/eval gaming.