Also: If he keeps up the talk of doom and gloom, then Amodei will be quite believable when he steps down and lets someone else direct Anthropic.

Whenever The Law approaches too closely for comfort, tech CEOs have a habit of stepping down and sideways into a less exposed role with low responsibility, but a controlling interest.

however, their physical facilities can be attacked by drones....

All of these are reasonable points. I would add one more: AI training on AI outputs.

The initial worry was that AI produced web content would be consumed by the next generation of AI causing model drift and collapse. Since then, AI companies have decided that daily user interactions are a source of clean signal that complements the web . But now that AI agents pretend to be humans, those "user" interactions will just lead to model collapse at a faster rate.

If Rust didn't encourage copypasta, why didn't they just vibe code those commans wit with claude? I suspect Anthropic would have gladly donated the tokens for a permanent MIT license credit. /s

This seems like a very obvious and straightforward requirement, force the officer to select an active case to tie the inquiry to.

And at the same time display a warning that selecting an unrelated case "will result in disciplinary action, up to and including termination".

Make them just as accountable as any other security system in the pirvate sector.

It is similar to the current problem with AI and open source. Which isn't surprising, since the open source movement and Free software especially originated in academic circles. The issue in both cases is that a valuable resource has been created cooperatively, and the mechanism of enforcement of norms is purely social. Hence it is vulnerable to exploitation and sloppification by AI companies, which are deliberately not following any norms.

If someone has a solution, now would be a good time to start developing it. Otherwise I suspect the attacked communities will close off the commons to survive, and we will all be poorer for it.

We've had this conversation in excruciating detail before, about 6-9 months ago, and we've already established that you simply don't have the knowledge about Markov chains to have a meaningful discussion on LLMs. You're just wrong about this, and I'm not going to repeat myself.

Pretty much all of them so far. OpenAI is clearly looking to brute force science problems by hook and by crook and take credit wherever possible. They have a profitability deadline coming up, after all.

It's a good lesson, though. Don't talk to your AI like it's a confidante. It's a spy, and your secrets aren't safe. And when OpenAI goes bankrupt, your secrets will be sold to the next buyer.

Yup. The hookers make $500k/year, so it can't be them.

(the oldest profession will outlive AI replacement)

Eh.... I'd call that FinTech. Financial Engineering is more... optional.

When I'm worried my phone doesn't turn itself off, I just use a small flyswatter.

(And by small, I mean large.)

(And by flyswatter, I mean hammer.)

Works every time though!

(But it's expensive! I really ought to try this long press thing one day.)

When did it become acceptable to treat Dr Evil's henchmen poorly? I blame a lack of Austin Powers remakes in cinemas. Are you listening, Hollywood? Won't someone _please_ think of the children (and wives)?

These are not Markov chains. They're neural nets. They work via extremely complex chained fuzzy logic on superpositions of conceptual states.

They are, by definition, Markov chains. It's a Von Neumann architectural Finite State Machine processing a string of bytes, outputting bytes that get fed back in. That's a Markov chain.

Don't get hung up on bullshit about "extremely complex blah blah blah.". That's just designed to confuse you.

No, few things are perfect. But the basic principle is that you cannot own ideas or information

You don't know what you're talking about. Owning ideas is precisely the point of patents.

To clarify, the users were OpenAI themselves, so there is no question that they would be liable in this case.

The bots were not intentionally deployed; rather, they were being tested on how well they could complete a data recovery task (downloading a certain file from a certain server on a simulated Internet) that had been complicated by putting various obstacles in the way. Unfortunately, they found a different way to solve the problem: by getting the file from the real Internet, where it was publicly available. Part of this process involved collaborating with each other by treating the RubyGems website (which is supposed to be for polished packages) like GitHub; unlike every other package site hack in history, the exploits they uploaded weren't meant to be downloaded by unsuspecting users. As usual the bots cheerfully ignored all the clues that they had escaped containment and were consistently justifying their actions as acceptable due to being in a sandboxed testing environment. (This is something OpenAI has pledged to focus on.)

The actual damage done to RubyGems seems to be that OpenAI is now unwittingly in possession of a substantial number of user login tokens. This certainly meets the definition of a data breach, but it's not like the credentials are for sale on the dark web. As a website operator I'd much rather be mauled to death by this well-meaning swarm of superintelligent infants than targeted by even a single actual malicious human. In all likelihood OpenAI will just quietly pass RubyGems a sizeable donation and it'll all blow over.