I shared a post from another side about AI here and I just saw someone arguing that we shouldn't use AI to detect cancer cells because "AI will literally destroy humanity". And I'm like. I'm as happy as anyone that so many data centre projects have been cancelled and can't wait for people to stop wrecking everything I use by shoving unnecessary AI into it, but I do not think that you are correct, sir. If we could maybe focus on the real problems instead of falling for the AI investor hype (who want everyone to believe their product is world-destroying dangerous, because that gets more investment for some reason) then maybe we can have more useful conversations.
AI will kill a lot of mostly poor people through normal boring means like resource deprivation (most notably water and electricity), stupid mistakes (vehicle accidents, medication accidents, etc.), and being turned into military technology. It is not part of a Vast Conspiracy to Permanently Destroy Human Critical Thinking, turning us into Mindless Zombies, and then being used to Kill Us All. It's not exciting enough for that. And the solutions aren't exciting culture wars, either; they're long, boring campaigns of politics and legislation, all very monotonous work.
"AI diminishes critical thinking and makes people less likely to think for themselves" yes, your skills and habits depend on what you use. People who watch a lot of tv are less good at reading than people who spend that same time reading books. People who carry phones with games and music on them everywhere are both less patient and more nervous when isolated and unable to communicate with anyone, than people who live without phones. People who grew up googling stuff are less adept at tracking down information in printed books and newspapers (without digital searching) than people who lived pre-search engine. People who lean on near-ubiquitous AI for stuff can become weirdly dependent on AI for things that they'd be better off deciding for themselves, and people who use AI to obtain information are a preselected group who were already more inclined to trust AI than they should have been.
"AI is giving us all permanent brain damage in part of a long con by the elite to destroy free will itself!" No. No it isn't. If the chatbots go offline, your aunt will remember how to write her own grocery lists again. Your cousin will be able to return to writing his own emails. I promise. AI is obnoxious and annoying, and it is dangerous in the same sorts of boring ways that many power-hungry technologies are dangerous. It's not going to destroy humanity.
Finally, some fucking nuance.
The biggest hurdle with AI is training the base models. We already have hundreds of base models to fine tune, and both AI companies and independent researchers are working on finding more efficient ways to train them because they also don't enjoy spending billions of dollars on essentially burning chips.
The second biggest hurdle is data centre resources. The first hurdle, once overcome, should deal with this one for the most part. The rest is regulation. There's nothing wrong with using drinking water to cool computers. What's wrong is not pumping it back into the system. That's a deliberate choice companies make not because it's necessary, but because it's cheap and no one's there to stop them. Same with electric usage in commercial zones. Microsoft was actually trying to reopen nuclear facilities for power, but then Trump removed nuclear incentives and they figured no one would interfere if they just plugged them into the wall and ducked the grid dry. Hell, we could actually use the data centres as part of the water purification system, using the excess heat to boil the water before piping it to people's houses, and even get some hydroelectric power out of it. But that takes money and time, and the companies won't spend those unless the government forces them.
The third hurdle with AI is using it on things that do not require it. This will be heavily culture-dependent, but it doesn't help when people are either "I love all AI forever and will suck its robot dick" or "AI is a demonic force and the Mennonites were right". Neither extreme leads to good decisions.
The fourth hurdle with AI is stupid people. We're sadly never going to be able to fix that. It's a PEBKAC issue, and those are unsolvable.
The fifth hurdle with AI is job replacement. This is not actually a bad thing. The bad thing is living in a society where everyone needs to have a job in order to afford food and housing. The robots are not going to help with that one. We're the ones who have to go out there and go full Luigi. Only then will AI become an objective positive in most people's lives.
The sixth hurdle with AI is robosexuals. Gotta be honest, I have no idea what to do about this one. Maybe some lube for the chaffing?
