AlteredEgo

Thanks, and yeah lol, the guy is definitely coherent. He's using a lot of technical jargon and also has that weird "cult leader" / CEO way of speaking to draw people in. Talking weird seems like a prerequisite to make it to the top lol.

I think what he's saying with "We don't need to oversell this" is "hyping of AI" is inflating the economy and creating a bubble that has adverse effects on everybody. The problem is overhyped growth and inflation, not AI in itself. And AI is growing too fast because of grifters. A type of "pre-enshittification".

Which seems a reasonable take. I think the anthropic guy said something similar with a moratorium or AI development, to slow down.

"We have to find ways to make these models raise the standards of living for every American"

He's presumably a horrible person aiding in the genocide of Gaza and empowering the worst of humanity, so this most likely just PR lies. He'll probably lobby against every single policy that raises the standards of living. He touting the same "extreme left just as bad as extreme right" (neo)liberal bullshit. The same "adults in the room centrism" nonsense. Probably offering to use Palantir to combat people like Mamdani (who are effectively moderate). And "half the people criticizing Israel don't think it has a right to exist" is vile too.

But the headline is just disinfo. It's probably just grifters slinging dirt back. Or "flack" in Chomsky's propaganda model. The article reads like a disgusting polemic hackjob. Really nasty journalism by Joe Wilkins, meant to distract from any discussion of the actual issues of what he's saying, and what many CEOs are probably thinking.

That would be the one in a billion case where it could clash, unless it's "transformative artwork" and except from copyright, like you turn it into some kind of satire or social critique.

The larger problem with this discussion is that you'll just deny facts. AI models don't just "blend images", that is disinformation. The models are too small and the input way too large. They are learning like an artist would, and if you would ask an artist to recreate the mona lisa or some iconic GTA5 scene he studied from memory, you'd get not dissimilar results.

What will happen is simply that AI models with be monopolized, free use will be impossible and only corporations and capitalists will have access to using AI models to make some kind of money. Imagine an army of robots that you can just tell to do some job. But you'll not be allowed to build your own or operate one without paying license fees. The laws you look forward to will seal the permanent economic enslavement of humanity. And that will be that.

I've recently imagined if you could build some kind of "micro monorail". Like some simple 4x4 inch metal rail that is used for a very small "pod" similar to the podbike / a velomobile. Like a 1 seater or 2 face to face seater which only weights like 200 kg instead of large heavy trains. You could just use earth screws and poles and weld them to the microrail, have it close to the ground or rise up 6 meters high to cross over existing roads or obstacles. You could also have a welding robot on that same rail to endlessly extend the rail and do all this installation automatically. And everything could be recycled.

I'm not sure you can follow this somewhat mad idea, but this could connect rural and urban areas without the need to build costly roads, and at the same time transport internet, energy, goods, trash, maybe even water. Fully automated and with ultralight vehicles this could be the cheapest, least invasive and lowest energy transport ever.

Anyway, I'm a fan of rail, but it does have a problem: You need railroad switches and that adds a lot of complexity compared to a car. Every stop, turnaround or crossing needs it's own stopping rail and switches which adds a lot of moving parts. So for some parts this could be a great solution, but for others it's overkill. Basically in a city, there is a limit to how close rail can get you to your house.

So in a city you're stuck with waiting for larger train cars and longer walking times. Small individual railcars won't work, and so they can't replace roads or cars. I believe that it's like a fundamental mathematical density limit to rail. And if you go shopping, you don't want to carry heavy bags or crates for like 10 min to your house. You can, but it kinda sucks.

And I don't believe that within a city at 50 kmh energy consumption between a self driving "podbike" with rubber inflatible wheels on roads and a tram or metro is significant. It probably only is about 10% at max, or maybe even less.

So in an urban area self driving "micro cars" would be the most universal scaling solution that would be most comfortable to people as well because they can drive you and your baggage directly in front of your house and can pick you up.

Obviously I haven't run the numbers, and I don't understand why there aren't any self driving narrow 1 or 2 seater cars build out of bicycle parts. Isn't this the obvious solution? There are so many advantages. Just build a million of them for one metropolis and completely switch over, ban all cars and test how it works. And make it publicly owned, basically free to ride robo taxies. It'll be cheaper for the city. Just the amount of real estate you'd gain from no big cars would be worth so much money.

Yeah I'd love that, that would be the future.

I looked for studies, but completely autonomous driving has too little data yet. But from A matched case-control analysis of autonomous vs human-driven vehicle accidents | Nature Communications, severe injury and fatalities are lower. I suspect "per driven mile" it's even less and it depends on model.

Not sure where you get your numbers shout DEATH AND DESTRUCTION!!! about self driving cars. I can only assume it's out of your ass.

It doesn't even matter. 65 fatalities over a few years is nothing. What matters is how the technology improves in the next years and how the numbers look once there is significant adoption. That it's as good as it is now basically means you're already wrong.

Would be nice if you could tip the actual cook. Like the food preparation chain is visible on the receipt with boxes for sending a small tip. But only after the meal turned out to be amazing.

But overall it's cool that delivery robots will probably mean an end to tipping culture.

Or, these shitty ai companies could pay creators a small amount to use their content for training. What is the problem with that?

It's logistically impossible, and economically non viable. The AI companies are a bubble that will soon burst. And then consolidate and monopolize.

But you want to ban all open source / weight AI models that ever possibly trained or learned from public information. You want to monopolize and give the AI companies ultimate power and control over AI, by creating the legal framework to make models never free.

Copyright covers reproduction, not learning. Books, articles, comments are all public information.

Oh wow plastic cartons for eggs are still a thing? What a total waste of oil

considerable death and destruction

Alternative facts?

It's about artistic freedom. I know how hard it is to program and make games and can imagine a better way using advanced (future) AI tools.

The problem with game development is not just the amount of work, and the capital required, it is the intersection of complicated technology and artistic outcome. Almost all disciplines in computer sciences are part of video games. It isn't just hard and more time intensive than a single person can do, there is a higher risk of complete failure. You need high level of "technical intelligence" to even do it, which precludes a large number of potential artists.

We have very few computer games that I consider "art". This not made to be pretty or profitable or entertainment, but something that shows you something new about the world. They do exist and they are labors of love or accidents, but they are rare. This is very different from other media like books or even movies. And movies are expensive and complicated affairs, but at least they have a linear nature.

The fundamental issue with computer games as a medium is that it is near impossible to invent a truly new game, with new gameplay and a worthy story that is not linear, and then turn it into a viable product (not even talking about commercial success, but something that is approachable to a wide audience. Like art ought to be. See Façade (video game) - Wikipedia for example.

AI has the potential to change all of this. The tools will become drastically more powerful and hardware and models cheaper to use.

Now, in a 100 years you'd surely have the ability to create virtual worlds by descriptive prompt, like ordinary people in Star Trek can do with the holodeck. How do we get from here to there? What technology is needed? Imagine you prompt "make a GTA V clone but underground with dwarves" and you get a complete game generated where you can now play or create or tweak a narrative. That's what the holodeck can do in star trek, and what I always wished for ever since I was a little kid.

Only then will we have the artistic freedom to create video games as art.

Or think of dialogue that plays with you as you deviate, already examples exist with TTS. Or a game or dungeon master playing with you, throwing challenges in your way. Or a crafting component in the game, where you don't have to painstakingly paint textures and nudge vertices to create some new style of armor or weapon or a house. You can just describe it in detail.

There are tons of problems with AI and society and the disgusting capitalists who will try to monopolize it. Making it easier or cheaper to create games isn't one of them.