In 2023 artificial intelligences based on neural networks are quite similar to the HAL-9000 computer planned for the 2001 by Kubrick and Clarke in 1968
We 22 years late but we made it !
This page is the right place to discuss and comment on the application Theremino_OpenAI and artificial intelligence in general.
Sam Altman, CEO of OpenAI, stated in an interview that some skills that were considered difficult for computers to learn, such as context awareness and language processing, turned out to be simpler than expected. However, There are also other unexpected aspects that are becoming problematic to solve.
OpenAI has invested a lot of time and energy into “train” GPT3 responses so as to prevent him from saying unsuitable things and this is becoming a problem for two reasons.
- Currently, is not considered “politically correct” affirm that “Women are different from men”. However, Biology, Anatomy, Medicine and psychology say there are differences, like chromosomes, the arrangement and type of internal organs, the average weight, etc. Therefore, it's hard to convince an AI that what they've read billions of times is wrong.
To get the correct answer, then, Rules must be imposed by force, i.e. “If they ask you this, you have to answer this other, even if you find it wrong”. This problem also occurs when it comes to “black and white” and an infinity of other things, including a large number of words, that billions of humans say and write, but that must not be said.
- The deformations imposed by those who “Trains” The neural network creates “Bias” And they could even be worse than the disease they would like to cure.. For a certain group of people, in a certain culture they might be right, But globally and scientifically they could be wrong.
In conclusion: We have an AI that is the sum of everything humans write and say., a sort of “Average Human” that could be very useful both to understand what we are, and to study how to avoid our future mistakes. If we let her go her own way, she could find solutions to our biggest problems..
We are sure that it is right to condition it with our limited vision of reality?