post-it

They'll become a reality on the same day as solar roadways.

Not what it says.

We fit a trend to the quarterly quantities of AI compute sold, as measured in H100-equivalents. We find that computing capacity has been growing by 3.3x per year, equivalent to a doubling time of 7 months.

This is just a graph of cash changing hands. I transferred a loonie from one hand to the other one trillion times this morning and have the largest data centre in the world.

And this is where the value is for me: I can simply make things higher quality than I could do them alone.

Yeah, that's the thing for me. LLMs have made my work easier and faster, and they've made my side projects easier and faster. I think there are very sensible and valid critiques but so far the tool works for me.

Well there's no need to suppose. While I think if it hasn't been exploited in 30 years, there probably isn't an attack surface, you can always demonstrate and report an exploit.

I disagree. Looking forward to reading your amicus brief.

You're not seriously out here posting raw numbers without considering the base rate are you

Right, but I hadn't really heard it before until I started using AI daily.

Because Trump likes being antagonistic. And the notion that plundering America is just a distraction from the files is itself a convenient distraction from the plundering of America.

Why has not a single person been prosecuted since the files were released?

Because the perpetrators run the government.

I'm not sure anything is a distraction from the Epstein files. I don't think the administration cares about the Epstein files. What would be the fallout if they were all released? We already know that a lot of wealthy people were raping children. It's not like the US is going to prosecute.

I caught myself saying "push back" the other day. I've never said it before, it's a Claude-ism.

If you're a job seeker, most of the jobs are fake for pretend growth optics.

Maybe, but it continues to be one of the best places to find work.

That's the neat part though, isn't it? It's a product that's so good that there's no everyday alternatives to it. I was researching cat litter options recently and cat owners do a lot of thinking and talking about litter, because there's a variety of different materials, none of which are solidly better than all others in all situations. But aluminium foil is so good that we don't even think about it, because it's by far the best product for every application that we use it for.

Closest they can get to piloting a mech.

I bought as few textbooks as I could, but the few that I did buy are sitting in my parents' basement bookshelves somewhere.

The fundamental problem is that reality has a surprising level of detail (as described in an article by the same name). The founder is the same kind of guy that believes the government should just get rid of inefficiencies.

Patents incentivize investment in R&D.

In theory or in real life?

Why did you do it this way?

One thing that I find helps is just avoiding the word "why" as well. Restructuring to say "how come" or "I'm wondering..." or "am I understanding right that..." helps avoid putting people on their guard.

It even works on AIs, interestingly enough.

I thought the same thing. I do think the first section was written or at least edited by hand. The rest could've been an email, so to speak.

At minimum you have to filter out 90%+ of people that simply don't have the mental faculties to evaluate what is and isn't a valid argument, before you even get started.

I don't think this is true. There are times when I do think it's true, and when I start feeling that way I know it's time to step back because I can no longer engage constructively.

Text is a hard medium to have a back-and-forth in. The features that make it useful for explaining also make it easy to feel ignored and insulted.

I think a lot of people also go online and write things when they feel argumentative, so comment sections self-select for people who want to argue.

Whenever I feel intellectually superior to someone, I try to remind myself that I can barely change the oil filter in my car, and there's a lot of people out there who can't write a line of Python but who save tens of thousands of dollars doing their own maintenance.

I don't know why Opus would try to create an entire library when I told it specifically to do something simple that would take 2-3 lines of Python.

Because it reasons in one direction. First it encounters some kind of issue with 2-3 lines of Python that might make it not work, and then it goes onto plan B, which is making a library, but it doesn't circle back and compare the effort of making the library to working around whatever might make the 2-3 lines not work. Except sometimes it does, because it's inscrutable.

This is an excellent comment from your HN link:

The core issue there isn't that the LLM isn't building internal models to represent its world, it's that its world is limited to tokens. Anything not represented in tokens, or token relationships, can't be modeled by the LLM, by definition.

I've noticed that LLMs struggle with tic tac toe, even when I get them to draw it out. And it makes sense! Tic tac toe is an extremely visual game, and difficult to reason about in a single dimension.

A more appropriate mirror test for LLMs is to get them to state facts about their training data. Percentage of arts vs science for example.

LLMs are not capable of this kind of reflection.

It seems like we forget that LLMs are next token prediction systems.

It's pretty clear to me that above a certain size threshold, LLMs are more than a sum of their parts. The sheer amount of training data seems to embed a higher level of reasoning.

One of the larger DCs is 7.65 GW

"is" is doing a lot of work in that sentence. The data centre doesn't exist.

This already isn't the case for the popular models. The knowledge baked into the weights tells the model how to talk and reason, but for world knowledge they do a web search right off the bat most of the time.

It could be a taxi driver if you like. Or an anarchist passing by on xir way to a protest.

Live Photos are what iPhones call it and I love them. I made an app to turn them into GIFs, I often make wigglegram GIFs out of them. "Giffer" on the App Store[0] if any readers are interested.

[0] https://apps.apple.com/ca/app/giffer/id6767937960

The streetlight effect:

A policeman sees a drunk man searching for something under a streetlight and asks what the drunk has lost. He says he lost his keys and they both look under the streetlight together. After a few minutes the policeman asks if he is sure he lost them here, and the drunk replies, no, and that he lost them in the park. The policeman asks why he is searching here, and the drunk replies, "this is where the light is"

All of your suggestions are better but they're hard, so someone casually evaluating an AI isn't going to do them.

Now that you know, do you feel culpable?

Of all the things to regulate why bother with this one? It's not like IPv6 is better for the environment or useful to the consumer.