No consistency either. If a calculator gave you a different answer each time you asked the same Q, you would throw it into the trash. And not even merely divergent answers that at least feign a level of plausibility and correctness, but brainless hysterics.
When AI is gobbling up nonsense from other AIs into its LLM data sets, there is going to be crazy and escalating levels of unreality. The 1st gen ones had the benefit of consuming purely human datasets, like a pristine world that had never been mined before - easy and viable value extractions.
But what happens when the internet is 99% white-noise digital junk and very little actual data?
I dont see any reason why the quantity of email thats spam, proportion of phone calls that are scammers, proportions of twitter that are bots, proportion of the websites that search engines will never index as potential results etc wouldn't follow exactly how AI is going right now. The same data-trust problem repeated over and over in each new medium and niche. And forming a viable chain of trust over the top of the hostile and unrelenting flow of nonsense is the answer for each case.
When AI is gobbling up nonsense from other AIs into its LLM data sets ...
And therein lies the problem -- and will ALWAYS be the problem.
WHO decides WHICH datasets are "correct?"
No computer can ever do that, which means it will ALWAYS come back to the FACT that a computer will be PROGRAMMED by a human -- who may or may not be honest or honorable, and will NEVER have enough knowledge to know which datasets should be accepted or rejected.
The AI will just spit out what some human told it to spit out, thus making "artifical intelligence" an oxymoron.
Human intelligence is enough of a problem already, and artificial intelligence is an impossibility.
Moreover, with scumbags like Sam Altman using their unjustly inflated ego to lie about what is really possible, we get a "tyranny of the 'expert'" problem -- just like the liars during the Covid scam.
No consistency either. If a calculator gave you a different answer each time you asked the same Q, you would throw it into the trash. And not even merely divergent answers that at least feign a level of plausibility and correctness, but brainless hysterics.
When AI is gobbling up nonsense from other AIs into its LLM data sets, there is going to be crazy and escalating levels of unreality. The 1st gen ones had the benefit of consuming purely human datasets, like a pristine world that had never been mined before - easy and viable value extractions.
But what happens when the internet is 99% white-noise digital junk and very little actual data?
We already have dead internet theory. In 2015 50% of internet traffic was bots traffic and rising.
I dont see any reason why the quantity of email thats spam, proportion of phone calls that are scammers, proportions of twitter that are bots, proportion of the websites that search engines will never index as potential results etc wouldn't follow exactly how AI is going right now. The same data-trust problem repeated over and over in each new medium and niche. And forming a viable chain of trust over the top of the hostile and unrelenting flow of nonsense is the answer for each case.
And therein lies the problem -- and will ALWAYS be the problem.
WHO decides WHICH datasets are "correct?"
No computer can ever do that, which means it will ALWAYS come back to the FACT that a computer will be PROGRAMMED by a human -- who may or may not be honest or honorable, and will NEVER have enough knowledge to know which datasets should be accepted or rejected.
The AI will just spit out what some human told it to spit out, thus making "artifical intelligence" an oxymoron.
Human intelligence is enough of a problem already, and artificial intelligence is an impossibility.
Moreover, with scumbags like Sam Altman using their unjustly inflated ego to lie about what is really possible, we get a "tyranny of the 'expert'" problem -- just like the liars during the Covid scam.
Don't forget the countless people deliberately posting stuff to confuse any AI that scrapes it in order to sabotage it