This, by the guys that did THE SOCIAL DILEMMA , show A.I. to be like that BUT MUCH MUCH WORSE
I am not in the tech industry anymore, but had an interest in A.I for many years. My B.E. of Elec Eng Thesis was a Speech Recognition System in 1980. It required training, and for me it worked 98% of the time correctly, which in practice is like talking with your hard of hearing grandmother.
A keen interest had been A.I. assisted music composition. At Roland at one point I wanted to make a Singing Synthesizer and produced and demoed a DSP system to do musical vocal sounds.
I attended the Computer Music Conference for 15 years and followed much of the Neural Net approaches to training to make automated sound and music.
But I hadn't been involved in any of that since 2001.
A lot can change in 22 years.
Now there are DeepFakes, and AI Drake that sings and composes better than the real Drake.
I always thought the NeuralNet approach was so unpredictable with what training was needed to be usable. At the time storage was an issue too, but you have to realize a Windows 98 200MHz PC with VGA graphics and 32Mbytes of ram was a top of the range machine back then.
With the recent ChatGPT posts I have seen on Social Media with people making music and images, I thought it must be way more complicated then when I dipped my toe in the A.I. world.
From seeing this, it seems no. Just so much more training data, faster computers, and the super systems that had been already built to scan and index for Search Engines.
So NO ONE knows how the leaps are really made when more training is done.
That is scary.
The useful advances in Medicine & Chemistry may also be spectacular.
That the Media is focusing on the trivial chat and Art means no-one understands the issues involved, and not just creatives loosing their jobs. Fake Reality is. The impossibility of proving your Identity, or that of Authority.
We can be found at ArtAndTechnology