Every person or entity who writes things on the web, allows Google to index our content in exchange for bringing visitors to our sites.

If Google isn't going to hold up their end of that bargain, then we need find a way to stop giving them our content for free.

It's a very hard problem, I'll admit. But it's a problem that needs to be solved.

I'm pretty sure a whole lot more content is going to have paywalls in front of it, or registration walls. Until someone finds a better model.

Kagi is basically Google without the ads or AI.
Instead, you pay as few dollars per month.
Because you are the CUSTOMER, not the product, the incentives are aligned against enshittification.

There's a free trial that gets you 100 free searches.

I switched after Google and Bing turned the first page of search results into a minefield of ads posing as relevant links. It was a step backward in quality at the time, but they're pretty close to parity now.

"It is getting increasingly clear that young people are being more affected by AI than older workers..."

While it is true that young people are disproportionately affected, it is not at all certain that AI is the root of this problem.

Other than a half-dozen tech-behemoth companies, the entire economy is in decline.

AI is the excuse that executives like to give when they lay people off, but there's always an excuse, and the excuse is rarely a reflection of reality.

If slashdot supported editing, I'd clarify that I mean "turn up the gain" WITH NO MICROPHONE CONNECTED. So you just get noise (literally just noise) from the input circuitry, not signal from anything.

Sorry about that.

If the computer has a microphone input, you can turn up the gain to get white noise from it, and use that noise to introduce as much nondeterminism as you want. Every multiply operation in the inference process could have a little tweak applied to the result, courtesy of cosmic microwave background radiation.

This would not make any difference in the matter of whether the resulting LLM+noise system is conscious. Not would the effect on the LLM's behavior be noticeably different from the existing LLM+PRNG systems that are in use today.

But it would render your "must be nondeterministic" objection entirely moot.

For the near term, you can look at this rule as a statement: "AI actors and scripts suck, so don't even bother trying." And they're probably right. They don't want to get deluged with crap submissions any more than open-source repo maintainers want to get deluged with vibe-coded garbage pull requests.

But it is entirely possible that AI movies will not always suck. There may (probably will) be a day when people start to really enjoy AI-scripted movies with AI-rendered actors. (Iran's Lego-world propaganda music videos are kind of amazing. As is the fact that a repressive regime is producing cutting-edge media. But that's another topic for another time.)

At the point, this rule will just be an artifact of a clique of artisans who want to defend their prestige against a disruptive technology. Like horse-drawn-chariot race officials declaring that motorcar builders are not welcome to enter their races.

Another organizing body will spring up, and it will cater to the desires of producers and consumers who appreciate the new technology.

Have they taken any action against the owners of the GitHub repos hosting Claude's reimplementation of their leaked Claude source code?

I read that as Tensor Casino at first, and assumed for a minute that there was some kind of math/physics joke here, and I just didn't take enough math course to get the joke.

And now I'm thinking that there really is some potential here for Tensor Casino humor.

37x the warming potential per what? Per gram? Per liter? Or per mile traveled?

Because I'm pretty sure the amount of carbon dioxide exhaust released per mile traveled in an equivalent jet-fuel-powered plan is far more than 37x greater in mass than the amount of hydrogen fuel released by a mile traveled in this hydrogen-powered plane.

Mostly because CO2 is an exhaust gas generated by burning jet fuel, whereas hydrogen literally is the fuel in this case. And since it is the fuel, they probably try really really hard not release any hydrogen outside the combustion chamber. And combustion combines hydrogen with oxygen to form (mostly) H2O.

Released raw hydrogen is wasted fuel, which is something that aircraft powerplant designers try to minimize. Whereas CO2 is simply the exhaust that spews continuously from a perfectly healthy conventional engine.

This technology fabricates citations to non-existent medical research. There were news headlines about it just a year ago.

People who expect it to be "more rational" will get what they deserve. But their customers deserve better.

Just yesterday I stumbled on this substack post about a research paper whose authors found that AI scored well on x-ray evaluations even when the AI took the test WITHOUT ACCESS to the x-ray images.

https://drjo.substack.com/p/wh...

The moral of this story is that properly evaluating AI performance in classification tasks requires very very carefully designed tests, because neural nets are very very good at picking up correlations between the desired outputs and utterly unintentional signals in the inputs.

When I worked at MS we referred to stuff written in C/C++ as "native code" and stuff written in C# as "managed code."

But I left around the time that Electron/PWA stuff was catching on, so maybe that's changed.

Yes, wine. A single bottle should suffice.

You still won't be able to run a .exe file but at least you'll be in a good mood.

"....we sure do spend a LOT of time and effort trying to ensure every square theory fits into the perfectly accepted Einstein circle of truth."

We spend a lot of time and effort ensure that every new theory fits with observations.

The fact that our observations align very well with the predictions of General Relativity is another matter entirely.

100% this.

CEOs have known for years that Wall Street prefers to see downsizing framed as "productivity gains due to AI."