It’s not like that. Tokens are an inherent computational property of how a model calculates the probabilities and such to generate text.

Having said that, what a token means in terms of computation varies wildly between models and is not directly comparable. So attributing a money value to tokens in general, independently of the model, is weird by nature.

And even within a model, the number of tokens needed to generate a response is very variable too, depending of the model itself and the parameters with which it has been configured (thinking mode, temperature, etc.).

So yeah, companies can pretty much set any price they want and there’s not much anyone can do about it.

Claude is trained on stolen data (the whole Internet), so I can’t have any sympathy for Anthropic when someone steals from them.

Given that the LLMs could follow the short lists of words well but not the longer lists, and that they were processing images, not text, I think it’s more likely that their context just filled up and they forgot the original instructions (or they were assigned a lower weight in the computation).

But Windows is doing the enshittification thing since the nineties, generally every other release:

• Windows 95: Good
• Windows 98: Bad
• Windows 98 SE: Good
• Windows ME: Bad
• Windows XP: Good
• Windows Vista: Bad
• Windows 7: Good
• Windows 8: Bad
• Windows 8.1: Bad
• Windows 10: Good
• Windows 11: Bad

8.1 is the outlier, but otherwise this thing works like a clock. You could argue that 11 is enshittifying more than the rest, but that’s probably because LLMs are enshittifying everything, I think.

For many years I have thought that elementary schools should be free of any modern electronic tech. In middle school it should be introduced but not connected to internet (word processors, interactive encyclopedias or coding like we used to learn it in the '80s or '90s, BASIC, Logo, that kind of stuff).

Finally, introduce internet in high school, in a controlled way, focusing primarily on knowledge and research sites. No LLMs included.

Of course you can’t control what happens in the students’ homes, but that should not affect what happens at school, specially if you replace homework with schoolwork, which I also think it’s better in the long run.

I would love to see a “Careless People”-style book for Google, Amazon and Palantir. Though the Palantir one would probably be called “Evil People” directly.

Plain wireguard. Or maybe Pangolin?

https://github.com/fosrl/pangolin

Lemmy isn’t social media? Given that it is mostly a federated reddit clone, I would argue it is. Probably even more social than reddit considering that reddit has now more bots than people, unlike Lemmy (at least for now), and I think that a platform full of bots talking among themselves is not really “social”.

I think they’re a dead-end mostly because of the exponential cost vs. performance. The decreasing returns are obvious, and the companies are trying to adapt by raising token prices, but that will not be enough with the current user numbers (or even double or triple, if we believe the analysts). I think that, at least with these large LLM companies, we’re actually beyond the point of economic equilibrium with this technology, at current energy and water prices.

And yes, training is more expensive than usage. That’s probably the reason why Anthropic suggested a pause in LLM development (training), supposedly because of the fear that AI could become Skynet, but really because they are getting an IPO soon and if people see their current balance numbers, the IPO would fail and the bubble would probably pop. Which really proves my point a little: the economics of these companies “improving their LLMs” (training) don’t make sense at current energy prices.

That’s why I said less negative impact.

And we can’t change human nature, but we can force social media companies to change.

I think the author is mostly right about the current state of AI, but his future predictions (or worries) are based on a false premise: that the massive LLMs will keep improving in the future.

As far as I have seen the improvements have clearly slowed down, while the energy consumption is rising linearly (or worse). It’s like the energy (money) vs. performance graph is logarithmic, and the companies are expending double the energy to get a 10% improvement. Something like that is not sustainable, and the money seems to indicate so.

I really think that LLMs are a dead-end for AI. A really useful dead-end, once the bubble pops and with time, we get a useful working model for them, probably based mostly on local LLMs, maybe using specialized training data.

A couple of days ago I finished reading “Careless People” by Sarah Wynn-Williams. This book really shows how awful the people at the top of Facebook are (it probably applies to other similar companies as well). And it’s not that they want to be evil, it’s more like they are playing a game of fame, pride and power and don’t care about the consequences. It made me recall the quote about “the banality of evil”.

This. Pre-2010 social media, which basically showed a feed with the posts of the people you followed, had a lot less negative impact on people. The engagement maximization strategy multiplies “negative” news, because outrage apparently engages more than good news. And current algorithmic feeds like TikTok destroy attention span and make people dumber.

And while we’re at it, ban (or at least heavily regulate) dating apps, the only business where the companies sell the opposite product of what they supposedly offer. If people find long-term partners through the apps they would stop using them, so the companies are incentivized to help you not find love and real relationships.

The funny thing is, being a developer, I was talking to someone the other day about how in my opinion, middle managers were way easier to replace with AI than developers (less domain knowledge, less creative work, etc.). And basically, at least according to the article, that’s the people that Cloudflare fired.

Not that anyone should be replaced by AI, but it’s ironic that many of the people who are shoving AI down our throats are the easiest to replace in an organization.

That’s for other sites that sell Steam keys, which obviously use Steam infrastructure hence the limitation. If you sell your game on Steam and other real platforms, like GOG, Epic or Uplay (or whatever it’s called these days), you can set the price you want.

Of course Steam offers a lot more functionality, and many devs choose to tie their games to Steam because of that, not because Steam forces them to do it.

At least the AMD system management requires physical access (the AMD PSP does not have a network stack). Intel ME / AMT does have a network stack, and it hides its packets inside the host traffic. That’s the reason of the black holes on many Intel CPUs when listening on ports 16992-16995 (the host does not see incoming traffic to those ports because the AMT intercepts it).

Stalin suing Hitler. I’m rooting for both of them to lose.

Not anymore. After public outcry, Google established 2 alternatives:

1. Through developer mode, you will be able to “forego security” (according to Google) and do a lengthy and obnoxious procedure to unlock your phone for installing any app.
2. They’re defining a way for official (sanctioned by Google) 3rd party app stores to work, and you will be able to install apps verified by them, instead of Google.

Stallman was right: https://www.gnu.org/philosophy/right-to-read.en.html

This is from 1997, mainly about books but it mentions debuggers and “free kernels” (linux and… hurd?). Considering that they also want to force OSs to require their users’ age and some Linux distros won’t do it… illegal Linux distros may be in our future.

That doesn’t solve the problem: it only disables the duck.ai assistant, but doesn’t disable the search results that are normal websites generated with AI (the real issue here). To be fair, that’s not DDG problem specifically but every search engine problem.

The web is now full of these fake websites, which is a real problem because on the search results they look legit and only when visiting them you realize it’s AI crap.

And the funny (or sad) thing is that current AIs are being trained on these fake hallucinating sites, so even they are suffering from false information, which in turn is given to humans and used to make more fake websites and… the result is up to your imagination.