David Deutch, one of the founders of quantum calculations, as the media write, is sure: the general artificial intelligence (Agi) in the future will be not just a program, but a person. In the Strange Loop podcast, he noted that the current AI, like Chatgpt, remain “obedient optimizers”, who can only compare templates, but not explain and not create new theories.
In his opinion, the key feature of real intelligence is the ability to explain and predict. Today’s linguistic models do not know how to, so consider them a step towards Agi is erroneous.
Doych’s most daring thought: when AGI appears, it will have the status of a person. Such a mind will “own” the computer on which it works, and demand rights – at least labor. An attempt to turn him into “property” will become a serious moral mistake.
At the same time, Doich does not support the ideas about the “uprising of cars”. He believes that threats will be more related to human mistakes and unwillingness to recognize new forms of reason. If humanity lasts a million years, it will be able to master the galaxy – along with artificial intelligence, and not contrary to it.
He also skeptically refers to the Turing test: the machine can simulate human speech without thinking.
