AI: thoughts, experiences, predictions

[Monkey]

I'm always using Full Self Driving FSD did everything to maneuver around them

I was going to say "is that even legal"...then i remembered who's been in charge of your country for the last few months.

[kyle]

I was going to say "is that even legal"...then i remembered who's been in charge of your country for the last few months.

I've been using it for almost 4 years, literally when I do drive something manually I question why are they making this legal, when there are many deaths each year to accidents from manually driving, Supervised FSD, is much safer, I almost got in a wreck in an intersection, when I did manually drive. It can see in all directions at a time, where I can't. Supervised FSD also forces you to watch the road, or else they take it away from you. People who strike out consider getting another Tesla, because they have a hard time living without it. I've only ever had 2 strikes, both were actually software issues, and not from me not paying attention. One was because I set my navigation too soon, I was testing something and I navigated to the spot I was at, which gave me a strike, the other was also something similar, I wanted to drive by a location, and I overrode to keep going without removing that location It takes 5 strikes to loose it, and strikes get removed after 7 days

[Word]

Don't mean to troll but the first thought that came to mind was the geese probably died because DOGE fired someone tasked with feeding them or guiding their movements but thank God the same guy who is responsible for DOGE builds cars that can blind out the trail of destruction his political meddling leaves behind

[sinewav]

I'm personally using Perplexity from time to time.

Just wanted to say thanks for mentioning this tool, I've found it pretty helpful. Of course, I'm often asking questions I don't know the answer to so I can't say for sure how good it works, but the citations help.

[Monkey]

Don't mean to troll but

Turns out these geese were illegal immigrants. ICE gave them ample opportunity to turn around and fly back to the country they came from but they ignored the order. They were subsequently shot down as a most urgent matter of national security.

[Word]

There were some more geese that are now doing time in an El Salvador prison as a chain gang

[sinewav]

Making a prediction: If we think of AI development in terms of an arms race between the United States and China, then China has already won. I haven't written down my reasoning yet as I'm pretty busy right now but I'll come back and post about it later.

[sinewav]

Continuation of above...

Previously I said that the winner of an AI arms race would be the country with the best laws, not the best researchers. I still believe this to be true.

Let's assume for a minute that, generally, a country's laws reflect the attitude and culture of it's citizens. Right now the people most distrustful of AI come Western states such as the US and Germany, while the most enthusiastic people can be found in India and China. AI in the United States in particular very much feels forced upon us by Tech Overlords. This doesn't seem to be the case in South and East Asia. A key component to success in any technology is diffusion of that tech, and today it looks like China has better positioning in that regard.

China's frontier models also seem the have an advantage of supporting more languages, and specifically the languages of their neighbors. Again, if diffusion is a marker for success then providing models usable by more people out-of-the-box is a huge win.

The US has waffled it's AI policy for a few years, and the latest American AI Action Plan is just as much a disaster as the previous one. It states repeatedly that the US must "beat" China and spread US values to the rest of the world. The problem there is first, naming China puts the US on their back foot. Adversarial motivations will not lead to success in AI. Second, in the opening of the action plan it specifically says Frontier models should not include references to "Inclusion" and "Climate Change". This makes Frontier models useless for education. Who would want that? The US has terrible values and the government wants to code those into our AI exports. Meanwhile China, at least publicly, has adopted a tone of global cooperation. And there is good reason to believe they are being honest in this case.

Currently nine out of ten open source/weight models are Chinese. On top of that, China is publishing at a faster rate than the US and some of the most interesting and useful research is coming from China (new architectures and optimizations). So while I said at the beginning the best researchers wont decide a winner, it certainly doesn't hurt to be a science powerhouse.

Does open source/weight matter in this race? I think it does. While we don't know the real nature of how AI will shape the future, I feel like we are seeing some parallels to the early web. A lot of the first web interactivity was bound to Java Applets and then tools like Shockwave/Flash and even Silverlight and RealPlayer. But today it's all JavaScript. All those proprietary tools pushed the web further, but ultimately the tools that won were ones everyone could modify and customize. It could be that while Google and OpenAI and Meta are pursuing AGI and "SuperIntelligence" (and billions of dollars) it's possible people will simply choose to train on top of open source/weight models for their specific needs.

It's still early in the game, but framed as an arms race the US may only appear to have an advantage over China.

[Word]

I'd argue China is facing a demographic crisis for the next 30-40 years, so it'll make it a priority to use AI to fight some of the symptoms of that. Also, Xi is in his 70s and had the political system tailored to his needs. Once he's dead or has a successor that can't keep the Chinese people in check (again, the demographic crisis will make things difficult, with or without AI), there will be some room for political upheaval. Same with Russia once Putin is out of the picture. If the US gets their crap together in time, there's a good chance for them to win the AI race. Our AI will be shaped according to our most urgent political needs so the result can be very different from China's.

Also, if Trump managed to alienate Europe/NATO/Africa/the UN/former allies for good, then there'll generally be less cooperation from those countries with the US in a field that could strengthen independency from the US. I don't think highly of the EU commission but if our European researchers collaborate, there's some potential there as well. We have our own right-wing politicians and anti-science guys with considerable political power though.

I hope that climate change/the waste of resources and human lives for pointless wars/RFK jr and the next pandemic leave all of us, as a species, something worth winning.

In the worst case the winner is neither the US nor China but someone like Peter Thiel. Yesterday I watched Jon Stewart's podcast with some historian and Tony Gilroy talking about the causes of revolutions and I'm still thinking about that a lot right now. The tech part starts at like 59min, but I linked you the timestamp where it gets political.

[sinewav]

I watched Jon Stewart's podcast...

Thanks. I find Jon Stewart kind of annoying but I listened to the bits you linked. I just downloaded Andor S1 & S2 so maybe I'll watch the podcast after I binge watch that show. Also, if you are interested in revolutions see academic writer Jack Goldstone.

I think it's unlikely that some rando billionaire will be the one controlling the world through AI, billionaires aren't clever people, just lucky. And just to be clear I don't think it's possible to "win" an arms race, people generally use the term to measure advantage in something doesn't have a defined goal. AI perfectly fits that description since the term itself is hard to define.

[Word]

Also, if you are interested in revolutions see academic writer Jack Goldstone.

Thank you! I will.

I think it's unlikely that some rando billionaire will be the one controlling the world through AI, billionaires aren't clever people, just lucky.

Well, maybe not a single guy, but we already live in a time where there is a tech oligarchy that has lots of control over formerly (more or less) independent media, our data and politics. I didn't want to define "winning the AI race" as "having control over the world", which is probably more common sense, but to me, "winning the AI race" just means "attaining and using AI that can destabilize/destroy/undermine your enemies/opposition and swing public opinion mostly in your favour, all this in a more successful way than your opponents do this", and we are already seeing that being used. I don't think anyone will really be in total control of anything for a long time, especially not democratic parliaments because so many people are fed up with democracy, but there will be people who make at least short-term profit from the chaos they sowed.

[sinewav]

I'm personally using Perplexity from time to time.

Just wanted to say thanks for mentioning this tool, I've found it pretty helpful.

Even though I assume every AI company is unethical, it's better when they aren't openly bad. Cloudflare recently shamed Perplexity for it's web scraping practices, but I learned today that Perplexity is partnering with Trump's propaganda site. This is a step too far for me, so I'm looking for an alternative and I'll post about that outcome later.

[Word]

"ChatGPT "Absolutely Wrecked" at Chess by Atari 2600 Console From 1977"

[Lucifer]

"ChatGPT "Absolutely Wrecked" at Chess by Atari 2600 Console From 1977"

I saw that. I'm still not convinced it isn't satire.