It’s also arguably the only thing propping up the American economy right now. They don’t have anything to bail with, if the AI bubbles goes on fire. Their economy may well come tumbling down in short order.
On top of that, you don’t have to stay bleeding-edge. Computers have mostly reached a steady-state these days. A decade old computer now is still a fairly usable machine. It won’t be fast, but it also won’t be a slug.
Not the screen, but it would be nice if I could replace the keyboard on my laptop, since it doesn’t work properly any more.
But since it’s built into the lower body, it costs about as much to replace as the device is worth, since I’d be replacing a big piece of the casing, so I just make do with also carrying a Bluetooth keyboard.
I was thinking more along the lines of the big evaporative cooling towers you see with a lot of reactors, since they rely on vaporisation to cool off, most of that water can’t be reused, short of getting picked up in the water cycle again.
Unfortunately not. Computing hardware doesn’t get that hot.
The chips would be fried if you had them at the temperatures where they were generating enough steam to run a turbine.
They cool the hot side of the loop with running water, like reactors do, rather than using air.
So the loop is closed, but they’re using running water to cool it off faster, rather than needing a bigger radiator on the hot side.
Weird argument they’re making, since distillation doesn’t give you the AI. It just gives you the style of the AI.
If you distilled Google’s newest Gemini into a GPT-2 model, all it would do is just sound like it. Your GPT-2 model isn’t suddenly going to be as capable as the model you distilled.
They usually do. They just do the same thing a lot of nuclear reactors do and also evaporatively cool one end of the heat exchange loop.
I would be curious how that will work with anti-trust legislation. At least currently, there’s an argument that Sony does not prevent people from buying games from other sources, or other publishers from publishing physical games for their consoles.
That’s no longer the case if they switch to a purely digital model, since they become the sole source of all the games on PlayStation, and have unilateral authority over them. The only place you can get a game on a digital model is Sony.
Such a unit exists and it is also called tokens, that can measure the capability of a model and the size of a running operation in a model.
I think you might have it mixed up with parameters, rather than tokens. Parameters are how big the model is, and are an indirect measure of how capable it is. Bigger models tend to be more capable.
But what they use for calculating your bill is something different today.
The tokenizer varies a little, but I don’t think it’s changed measurably from tokens. You pay an amount for a million tokens worth of processing. The tokeniser difference just alters how text is converted to tokens, but the tokens themselves don’t change all that much.
If anything, I’d honestly put the issue more with reasoning chains in models, where they basically babble to themselves inside of a <think> tag, that most interfaces hide/collapse. It makes them work better, but vastly increases the amount of tokens per operation.
They have been getting longer and more sophisticated with newer models. So you might have a model now that basically repeats the output multiple times whilst refining and drafting the non-reasoning output.
If you’re making it generate a lot, that’ll balloon the usage, and thus price.
Social media has also done its hardest to try and push people away from using it. Between the culture being awful, and there being an increasing number of roadblocks to using it, that ironically ruins discoverability for anyone who might want to use social media.
For example, if you want to use Reddit, and see a link, there’s a lot of posts that you can’t see without having an account and logging in. That’s a big ask for something that you’re not even sure that you want to sign up for, which would only be worse since you couldn’t sidestep that using the old reddit interface.
Meanwhile, Twitter not only makes it so that you can’t see much of anything without being logged in, but they’re trying some new scheme where if you have an account, you need to download the app and give them your biometrics to confirm that you’re human before you can use your account.
If you’ve scarcely used either site, why would you start now? Everything wants you to jump through more and more hoops to verify that you’re actually a human, and if you don’t have an account, the content that you can see doesn’t seem to much of anything interesting. When not logged in, some subreddit and posts are completely inaccessible, and on Twitter, you can only see the tweet, but not the replies, or recent user posts.
Both of those were the main draws for each site. Why would any new user want to use them now? The only thing that they have is their reputation, and that will slowly go away with time.
Once upon a time, for example, Twitter was once the haven for beginner programmers, because they had a nice, free easy-to-use API that anyone could use to make bots and learn how to use APIs in general. Reddit was not far behind that. But that’s mostly gone now. Reddit no longer approves API keys for the most part, and is working to shut down the public APIs that it has left, and Twitter has locked theirs behind a paywall.