In a vacuum, GPTs are a significant advancement in Machine Learning and AI. It’s the hype, overhype and blatant greed that’s made this whole thing toxic.
They were a significant advancement, yes. But they were a dead end and that's why we dropped them seventy years ago
That's not to say they're useless, but the hype does negate literally ever use case that shouldnt get you abd your family dragged around the country on meat hooks-admittedly they're quite good for some of those.
That's like saying "All Modern Games are a dead end. We've had Gouraud Shading since 1971."
Just because all 3D games use a "dead end" shader doesn't mean they can't be a multi billion industry. Games look great using primitive shading algorithms that we know are a dead end to achieving photo realistism.
If someone was claiming they were working on making a game that contained all games, we would all laugh at them because we know gaming well enough to understand there absurd. But when we talk about computer intelligence like that, some people are like, "well, maybe" because they don't understand computer intelligence.
They're also starting to charge more for the BIG ASS models.
The amount of times these things have to regenerate code to get it to work properly during the SOTA "agentic" operations, you can burn through your token allotment and get throttled super quickly.
Google did something similar. They made search worse on purpose so people would do more searches and see more ads, boosting revenue.
But yeah, diminishing returns means investor money drying up means squeezing customers and hoping all that putting AI into everything translates to lots of paying customers.
25 replies
We are 1 model away from AGI
Yeah, fuck all the people who saif this tech was a dead end? What did they ever do fir comouter science?!
In a vacuum, GPTs are a significant advancement in Machine Learning and AI. It’s the hype, overhype and blatant greed that’s made this whole thing toxic.
They were a significant advancement, yes. But they were a dead end and that's why we dropped them seventy years ago
That's not to say they're useless, but the hype does negate literally ever use case that shouldnt get you abd your family dragged around the country on meat hooks-admittedly they're quite good for some of those.
I‘d love to read more about that thing that happened 70 years ago.
They are basically just giant markov chains. Knock your heart out
That's like saying "All Modern Games are a dead end. We've had Gouraud Shading since 1971."
Just because all 3D games use a "dead end" shader doesn't mean they can't be a multi billion industry. Games look great using primitive shading algorithms that we know are a dead end to achieving photo realistism.
I'm also not saying LLMs and Markov Chains don't have a use. Just it's a dead end towards general artifical intelligence
If someone was claiming they were working on making a game that contained all games, we would all laugh at them because we know gaming well enough to understand there absurd. But when we talk about computer intelligence like that, some people are like, "well, maybe" because they don't understand computer intelligence.
Kindly, I believe the -ist is replaced by -ism, rather than having both. (Realism)
More like 75. Turing was involved but not like the only person involved.
They had one trick: scaling. And that trick is getting diminishing returns.
They're also starting to charge more for the BIG ASS models.
The amount of times these things have to regenerate code to get it to work properly during the SOTA "agentic" operations, you can burn through your token allotment and get throttled super quickly.
Google did something similar. They made search worse on purpose so people would do more searches and see more ads, boosting revenue.
But yeah, diminishing returns means investor money drying up means squeezing customers and hoping all that putting AI into everything translates to lots of paying customers.
and they would have gotten away with it, if it wasn't for you meddling kids!
You could write "corpos", "tech bros", "capitalists" or just "everybody" at Lois' place.
90% of tech bros quit right before they're about to create agi
depends on if they have their own bunkers.
https://www.vice.com/en/article/openai-co-founder-wanted-a-doomsday-bunker-in-case-agi-caused-the-rapture/
One more parameter could make the difference!
One more trillion please
Waiter, waiter! More GPUs please!
and a water table to cool it. oh and add a bunch of generators - fuel 'em with methane, why not.
With enough nvidia gpus they'll all begin to learn how to read right?