Most of us probably understand that what we are calling "AI" is a total departure from what the term has meant for decades. Yet many people have accepted an underperforming algorithm to be marketed as "intelligent" instead of what is really meant by the term- a thinking machine, something that can cogitate.LLMs and similar tech aren't terribly complicated. Use a weighted grid and essentially generate a semi random path based on input. That's why you can slightly change a prompt, or recrop a photo (pic related) to get wildly different outputs. Science may not know exactly how intelligence works, but I think intuitively, everyone knows that this is definitely not it.I get that this is a contentious topic, but I'm genuinely interested to hear what you all think. If you're pro AI, do you believe it actually is thinking, similar to how a human does? Are you more just hopeful it is the right path and, with lots of money, the Silicon Valley types will figure it out? And anti-AI guys, are you truly anti AI or do you just think it's a fancy text generator? Would you still be anti if what you were being sold was closer than a positronic brain rather than a janky chatbot?
>>110216662>Google search AI"Flash" models doled out for free for every google search are not the same as token-burning frontier models, presumably. A lot less processing power goes into those than most models because Google has to use a lot of them for little to no profit.
>>110216662i think your post wrongly assumes everyone's divided into "pro" and "anti" camps, where everyone in the "pro" camp is a total evangelist who sees nothing but ai and the "anti" camp sees absolutely no useful aspect of the tech and thinks it should all disappear.i'm "pro-ai" but i am also fully aware it's just a fancy text generator. the tech is laughably simple, a second year comp sci major could write an llm in a few days and the ai companies are doing everything they can to have everyone believe there's more going on than they're actually is. but that doesn't mean it can't be useful. if you have a question and you're fairly confident it's popular enough to have appeared in training data a few times, it's much easier to ask an llm than to browse some rando websites looking for the answer and meanwhile clicking out of ads, gdpr popups, newsletter popups, paywalls etc. and if you code you can have an agent do mundane tasks for you - if something is high code but low complexity and you're able to write a precise prompt where there's not much wiggle room in terms of answer, it's unlikely for it to screw up.unfortunately both of the examples i covered are also brought to the absurd by the ai evangelists who think verifying information is for suckers and coding is dead. i'm guessing those are the people you describe as "pro-ai" since they're quite prevalent nowadays and definitely louder than the mild user crowd like myself.so in conclusion i am pro-ai but i am also 100% aware it's a fancy text generator (that's really a fact not an opinion desu).
