We've reached a point where machines can do pretty much everything better than humans and animals too, even without AI. Because people add solutions to problems

We've reached a point where machines can do pretty much everything better than humans and animals too, even without AI. Because people add solutions to problems and if they're better than previous ones, then it's downloaded to all the machines which perfectly run the software. Self-driving card etc. No animal can compete with the accuracy, reaction speed, etc. Racing Drones will be able to fly better than 99% of human operators. Cars will be able to drive better than 99% of human operators. And consistently. Google already gives direction better. Trading bots trade better. They are preferred by people and corporations, respectively. I wouldn't be surprised if computers can brute-force better scientific theories just by being fed motions of stars or other data. Like deducing Kepler's laws in 2 minutes and then going much much further, into laws that use 19283-dimensional vectors and are 99.9% of the time predictive. Monte Carlo Tree Search has done very well with chess, with AlphaZero beating all human-designed programs like Rybka, and playing much more elegant chess too. Where we need AI models is training on human-produced data (art, text on the internet, etc.) And there, it can be trained to maximize various metrics, including human satisfaction, or actual investment into an enterprise, and soon -- laughing at jokes, or agreeing to go out with someone romantically, etc. Just pick the stuff that works the best, and show it to humans as a superstimulus. In addition, the models are developing "understanding" of the latent space, the way a student in a classroom would develop this from listening to the professors explain stuff (rather than do experiments themselves). The latent space, thanks to humans, models the real world quite well, from a human perspective, and the computers can become polyglots and polymaths... ALL of them. Just download some compressed model weights onto a small computer, and it becomes more knowledgeable about topics than any human. And just like other software, the weights can be improved. But the next level is Swarms. Swarms of bots in online accounts at online platforms and communities will be able to optimize "reputation" points that they accumulate from existing humans in the communities (social capital). By doing this 24/7, and mimicking timing of humans etc. they can eventually (in aggregate) command far more reputation than any set of humans. They can also coordinate over long periods of time, as sleeper accounts, and eventually undertake any reputational attacks, including and up to having a person's own friends turn on them, ruining a famous person's reputation, having them convicted of a crime, or even foment a war etc. Whether or not humans are behind these directives at that point will be irrelevant. It's a bit like giving everyone atomic bombs or the ability to create novel viruses, except online. Someone's bound to misuse it. And because the swarms are so destructive and able to blend in, the internet will become a Dark Forest and people will try to retreat into real-life communities, where they will try to eliminate bots by having people show up periodically at events, but never completely succeed (because these same people will run bots on their own accounts). Already for years, corporations have been deploying algorithms to optimize people's addiction to TikTok, Instagram etc. with predictable effects of depression in teens, anger in adults arguing politics, etc. And that was before AI. I wrote about this here in LA Weekly: https://www.laweekly.com/restoring-healthy-communities/ https://www.laweekly.com/restoring-healthy-communities/