pull down to refresh
20 sats \ 8 replies \ @optimism 16 Aug \ parent \ on: Sam Says ChatGPT Is on Track to Out-Talk Humanity AI
I think you're missing the scale.
First, models aren't singular or continuous. If OpenAI has half a billion active users it means they need millions of model servers.
Also huggingface says they host over a million open source models, I myself have a bunch of experimental ones on my laptop that I didn't upload to HF, yet they've produced slop.
Wow!
You can take me as dumb about not knowing the scale of servers or all of what going behind the AI scene.
However I also think Sam wants the world to remind about the lack of in person communication which has been on the decrease in modern world.
reply
Not taking you as "dumb" at all!
However I also think Sam wants the world to remind about the lack of in person communication which has been on the decrease in modern world.
I think Sam is with every word coming out of his mouth pitching his company (which arguably should be what a good CEO does) but it also means we should probably not take anything he says at face value. everything is a pitch and meant to make you buy his product.
reply
Oh yeah, well diagnosed! That's definitely everything he means whenever he opens his mouth.
I'm like, I'm not against AI, but tbh I haven't used any of them, and being a literature student I'll never use it. It's just my innate linguistic creativity that often leads me to such observations.
reply
I'm not against AI either, but I am against the current AGI narrative, because it's imho deceptive wishful thinking based on fake-it-till-you-make-it and other bad practices popularized by Silicon Valley VC bros with no moral compass.
For "literature", sometimes search results (including AI-enriched results) don't give me what I'm looking for so I pose a question to an LLM. This happened yesterday, so I asked my big
qwen3 model:"Compile a list of literature that will enhance understanding of the difference between game theory in n-person and 2-person games."
I got some recommendations which I am still in the process of exploring (I got distracted by SNL so this is still pending) but in no way am I going to use some model that is only able to respond to me because it's been shaped to respond to me in a particular way, and allow it to give me the actual knowledge that I am seeking. Instead, it will help me grow insights of my own by sharing me source data, not modulated data.
reply
Yeah correct, the narrative is just filled with FOMO or money making bait.
If you ask AI to explain the works from Old or Medieval literature, please inform us. I don't believe AI can ever make it. There are some works like Milton's, Shakespeare's, Spenser's etc. that go so deep into the classics and mythologies, I don't see any chance for AI there.
reply
I think it can explain it if you explicitly train it on:
- The works in question and the context of the works
- A skill of explaining works
And leave out a lot of the "generic" stuff that are built into popular LLMs today because they are faking AGI.
That "training" is really more like a software layer or database tuning, not magic. An LLM isn't an entity. It doesn't live. It isn't aware and especially not self-aware. That's all programming to make it appear like it is, but it's a cheap trick, just a little more expensive of a trick than the HAL9000 IRC bot had.
reply
Yeah, someone would have to devote years to train AI on these vast works where most words/phrases/sentences can have more than one explanation (most of the tomes many). Anyways AI can help in tasks, but when it comes to explaining human creativity in Arts, it will forever be ugly.
reply