fwipsy

I find e-readers to be a pretty good compromise to reduce eyestrain. Older Kindles are pretty cheap here in the US at least and can side-load free ebooks (e.g. from project gutenberg.) (I guess Amazon may be removing that functionality in newer firmware so you may need to apply a jailbreak to prevent updates/prevent it from being removed.)

I guess that many PCs don't generate the dumps because they shut down after a shorter period, and they're just analyzing kernel dumps from a subset of PCs.

It's like Wordle but for Blue Prince rooms. You're trying to guess a specific room, the game will tell you what you got right or wrong about the last room.

Even so, not sure it qualifies for flagging

I would call it a Switch switch SwitchBot Bot.

Ah, I was missing the obvious. Thanks for pointing that out. I thought that the scandal was that these instructions applied to Meta's AI; if they were reaching users' own AIs that's a whole different story.

Web games don't make money because they're expected to work reliably, people play them on impulse. If it breaks 1% of the time, that's hardly a problem at all. If a kernel breaks 1% of the time, nobody would use it ever.

A raytracer demo doesn't need to be engineered to the same standard as the Apollo onboard computer software. A $20 paint app doesn't need to work as reliably as banking software that handles millions of dollars of transfers every day. All software exists on a scale from trivial to critical. It's like saying a toy drone needs to be designed the same way as a F-16.

They never claimed it was a bug that they were included. It's a bug that they were revealed to users.

Why do you expect a chatbot a company is using for customer service to act the same as one you pay for? Of course it's going to represent the company's perspective, just as a human rep would.

> a way to scan the model weights for backdoors

The AI safety folks have been working on a way to verify model behavior more generally. The term I've heard is "mechanistic interpretability" and it's actually really hard, though there's been some progress.

Yep, he mentions the video you linked as well. (Mentioning because I watched the video you linked first.)

But we still require cars to have seatbelts, turn signals, ABS, etc, etc. AI companies have a responsibility to build AI which is safe for normal purposes; users are responsible for not misusing them.

Some applications (e.g. a kernel) need to be carefully engineered with every possible edge case considered. Some apps (e.g. a web game) don't. This is a way for careful engineers who need to build dependable software to integrate AI without compromising quality.

I'm glad I read this, but I wish the HN title included the subtitle too--it would be nice to be able to know know what the article was about before clicking it.