if the majority of people were incapable of meeting the standard of AI, then AI wouldn’t have enough data to train on to be better than the majority. it’s a lossy database of things that already exist, it cannot create novelty or surpass its training set.
AI doesn’t produce data suitable for training AI. It’s a huge problem when AI generated slop makes its way into the training set because it generally degrades the quality of the model. Like a photocopy of a photocopy.
So where is all the data its trained on to surpass most people come from? Do you think they’re curating what they feed it based on IQ scores or something? Verifying accuracy, competency, etc? Or are you aware they just turn on the reddit/stackoverflow/github/etc. scrapers and start pumping them full of unfiltered 100% pure grade A internet bullshit?
Yes, I’m aware of that. But you don’t seem to be aware of the fact that most cultural output (and yes, that includes Reddit shitposts) is produced by a small minority of people. Most people never contribute anything. So however shit the AI slop may be (which I’m not in any way denying), it’s still better than what the majority of people can manage. Just look at the percentage of people that are functionally illiterate.
Im sorry dude but if your argument is reddit and stackoverflow are the basis for being “better than what the majority of people can manage” then I just have to respectfully disagree.
To be fair, I do think the average accepted stackoverflow answer displays far more competence than the average human.
One of the few things I use LLM’s for is giving me overviews of best-practice in things I’m not familiar with (before reading the posts I find to get more in-depth understanding)
if the majority of people were incapable of meeting the standard of AI, then AI wouldn’t have enough data to train on to be better than the majority. it’s a lossy database of things that already exist, it cannot create novelty or surpass its training set.
Most people can’t do that either.
AI doesn’t produce data suitable for training AI. It’s a huge problem when AI generated slop makes its way into the training set because it generally degrades the quality of the model. Like a photocopy of a photocopy.
So where is all the data its trained on to surpass most people come from? Do you think they’re curating what they feed it based on IQ scores or something? Verifying accuracy, competency, etc? Or are you aware they just turn on the reddit/stackoverflow/github/etc. scrapers and start pumping them full of unfiltered 100% pure grade A internet bullshit?
I would assume they rank the stuff by vote-count.
Yes, I’m aware of that. But you don’t seem to be aware of the fact that most cultural output (and yes, that includes Reddit shitposts) is produced by a small minority of people. Most people never contribute anything. So however shit the AI slop may be (which I’m not in any way denying), it’s still better than what the majority of people can manage. Just look at the percentage of people that are functionally illiterate.
Im sorry dude but if your argument is reddit and stackoverflow are the basis for being “better than what the majority of people can manage” then I just have to respectfully disagree.
To be fair, I do think the average accepted stackoverflow answer displays far more competence than the average human.
One of the few things I use LLM’s for is giving me overviews of best-practice in things I’m not familiar with (before reading the posts I find to get more in-depth understanding)
Then you obviously live in a bubble.