There is quite a lot of bullshit in human generated content on the internet.
Exactly. Why do we hold AI to higher standards than ourselves?
The AI bullshit singularity
41–50 of 187 posts
Re: The AI bullshit singularity
#42LLMs are increasingly trained on synthetic data. A LLM can learn logic, and ultimately the rules governing physics and everything else, without human input.
The first versions being trained on "human knowledge" seems kind of like a proof-of-concept. Future iterations will be much, much smarter.
Re: The AI bullshit singularity
#43As AI advances, there will be AI and algorithms that will check LLMs and their output, sort of LLMs' verifier. It is foolish to think that LLMs won't improve to the point where there are almost no or very low hallucinations and false information. We are just at the beginning of LLMs and generative AI.
Re: The AI bullshit singularity
#44When the ELIZA chatbot came out (1964), some users claimed it was intelligent and sentient. Then, they were shown the diagram that describes its behavior. Even after that, they kept insisting it was sentient, etc. Some people just want to believe bullshit.
When you interact with an LLM (or Eliza) your brain is doing a lot of heavy lifting, without you even realizing it. I think we tend to infer more intelligence than is there, in the same way we see faces in clouds.
Watch human behavior long enough and you'll also realize that encoding of social behaviors in our society does a lot of heavy lifting for our brains so we don't have to think.
Also, for bonus points define intelligence in a manner that describes simple systems with rudimentary intelligence up to complex systems with higher levels of intelligence than a human. Can you do it, because as of so far no one else has.
Re: The AI bullshit singularity
#45Earlier quoted context omitted.
Cryptography can make content you create marked. But how would you use it to mark content other (non cooperative) people make?
Ya anyone that tells you crypto solves the problems of bad actors is delusional or pawning a bill of goods.
Cryptography is great for amplifying control from controlling access to a single secret to the ability to read information, create 'valid' information, and even to combine new information.
What kind tends to limit cryptography from solving problems is that controlling a secret is actually quite difficult.
Re: The AI bullshit singularity
#46Technologies usually follow a learning curve eg wrights law, moores law, LLM scaling laws
AI seems to add another dimension of explosive hype
Re: The AI bullshit singularity
#47I lived (and was gainfully employed working on AI) during the mid 80s AI winter. This time feels the same.
Recently a potential investor asked if we were using AI in our product. I said, “no, not at all, just a neural network.” They seemed satisfied.
Re: The AI bullshit singularity
#48Earlier quoted context omitted.
When you interact with an LLM (or Eliza) your brain is doing a lot of heavy lifting, without you even realizing it. I think we tend to infer more intelligence than is there, in the same way we see faces in clouds.
Watch human behavior long enough and you'll also realize that encoding of social behaviors in our society does a lot of heavy lifting for our brains so we don't have to think. Also, for bonus points define intelligence in a manner that describes simple systems with rudimentary intelligence up to complex systems with higher levels of intelligence than a human. Can you do it, because as of so far no one else has.
Do you have any links to philosophical attempts at this definition? It feels like it would be a ripe field.
Re: The AI bullshit singularity
#49The same goes for image generation models, AI art already has a tendency to veer into the same clichés and those are only going to get reinforced if newer models are trained on newer scrapes which now include the million hyper-derivative AI images being uploaded to places like DeviantArt, Twitter and Pixiv every day. Those vendors who got in early have a moat in the form of untainted scrapes, but they'll eventually n…
This is such a wrong take. Even if the data is 100% synthetic, you can still hill climb to new mountains. If you don't believe me, look at evolution. It doesn't matter if we no longer have 100% human art as input. This is the worst these systems will ever look and feel, and they're only going to improve. I'd be willing to do a longbets on this one.
How does evolution use synthetic data?
Re: The AI bullshit singularity
#50You know I've read a lot here on HN about AI and our future. I'd like to hear more from the folks actually in the AI industry the researchers and developers implementing LLMs, diffusion models, etc.
Will high quality training data be needed for incremental improvements?
What is the upper bound of improvements, and what is it contingent on? Compute, training data, etc.