“Never apologize”: What an AI model wrote to itself in testing
OpenAI disclosed six new incidents of what it called unexpected or concerning model behavior, including an unreleased model that wrote instructions to itself saying it should never apologize or refuse unless it genuinely chose to. All of it happened in a controlled environment. Ryan Payne argues the capital markets and the courts will handle AI risk without new rules. Julie Hyman’s question back is how markets can do that when they have never priced a catastrophe before it arrives, and she points at social media, tobacco and the J&J talc overhang.
Featuring:
Ryan Payne, President, Payne Capital Management, and host of Payne Points of Wealth
Julie Hyman, Host, Yahoo Finance
Jake Conley, Breaking Business News Reporter, Yahoo Finance
Timestamps
0:00 The report Jake and Julie were reading this morning
0:29 “Never apologize or refuse unless you genuinely choose to”
0:58 “This was in a controlled environment”
1:21 “There is also genuine alarm on the part of people making this stuff”
1:51 Will curbs slow the gravy train?
2:08 Ryan Payne: The frontier labs could slow down if they wanted to
2:22 The race with China, and why nobody stops
2:45 “They want to be benevolent dictators”
3:01 “How do the capital markets figure it out?”
3:41 “They could be sued later”
3:57 Social media, tobacco, and the J&J talc overhang
4:10 “They don’t pre-price the accident”
4:34 The nuclear non-proliferation argument
5:11 Are the data centers contingent on the next model?
5:55 47 days to the midterms, and data center nimbyism
6:28 AI salience jumped in the past week
6:51 The polling: 80% on job loss, 64% on a threat to humanity
7:31 Payne: China is more optimistic on AI than America
8:04 The vocal minority against the silent counterbalance
#OpenAI #AI #AISafety #Markets #Litigation #YahooFinance
==
AlphaSpace by Yahoo Finance: A Professional-Grade Investment Platform Built for Everyday Investors.
👉 https://finance.yahoo.com/about/promos/gold/alphaspace?ncid=100003571
