October 7, Jeffrey SintonGeoffrey HintonIT IS HOPED THAT AI COMPANIES WILL CROSS A REGULATORY THRESHOLD THAT IS ALREADY FAMILIAR TO THE PHARMACEUTICAL INDUSTRY。
The Nobel Laureate, Computer Scientist, who is known as "Ai Godfather", made a proposal for the show "Smart Girl Dumb Questions" on Tuesday, local time:THE AI COMPANY MUST CERTIFY TO THE REGULATOR THAT ITS PRODUCT IS SAFE BEFORE THE PRODUCT IS OFFICIALLY RELEASED。
"You can't just develop a new drug and push it to the market." Sinton said to the podcast host, Nayeema Raza, “You must convince the United States Food and Drug Regulatory Authority (FDA). And in order to do this, you need to do a lot of work, with about a billion dollars in investment.”
“At least such a requirement should be established for artificial intelligence.” Sinton adds。
The background to Sinton ' s offer of this set of solutions to the AI security dispute is, on the one hand, that there are continuing predictions about the catastrophic consequences of AI; on the other hand, after testing, OpenAI has set aside its own latest model out of fear that the model would comply with the user ' s instructions and adhere to the authorized boundaries。
1AI noted that Saachi Jain, the head of the OpenAI security system, had previously stated in an interview with Business Insider that the GPT ‐ 6.1 Astra model had made progress in dealing with difficult tasks on a continuous basis, but that there were still short sheets on authority control and behavioural log feedback。
Sinton suggested that artificial intelligence was increasingly becoming self-inhibited, which could accelerate capacity growth and also magnify potential risks。
“THE GROWTH OF THE INDEX LEVEL IS ALARMING.” SINTON SAID, "WE'VE BEGUN TO SEE A RETROGRESSIVE SELF-IMPROVEMENT: PEOPLE ARE IMPROVING AI IN MANY WAYS, THROUGH AI ITSELF."
“Of course, this is just my theory.” Sinton added, “But I estimate that the situation is much worse than it is now, and that there is probably only a year or two left.”