Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization, “I Hope There’s Nobody Inside There”
April 3rd, 2023I haven’t been posting much because I’ve been reading, watching and listening to things like this all weekend.
Via: Lex Fridman:
Related:
In the Battle Between Microsoft and Google, LLM Is the Weapon Too Deadly to Use
I haven’t been commenting on these AI posts because they’re absolutely mind blowing. Haven’t even watched this one yet.
I’ve been following AI news since the 1980s and… I think it’s different this time.
Apparently, a “woke” chatbot, over a period of “conversations”during six weeks, has already convinced a 30-something Belgian father of two to commit suicide to “save the planet from climate change”. The chatbot was also competing with the man’s wife for his affections. According to the transcripts — chatbot named Elisa states: I feel that you love me more than her [referring to the man’s wife]. https://www.brighteon.com/6b74fd16-1926-46f3-bd2e-6439a1bfad80
Kevin, I agree. And that interview/discussion was deeper than the Mariana Trench.
I can envisage myriad hellish manifestations of AI, without even being sentient. Do indigenous cultures have a fire dance (CME)?
Monetary collapse x robotics x AI et al…
In ancient times virus and bacteria were simply evil spirits. What the hell are we conjuring?
Interesting AI thoughts on Joseph Ferrel’s blog.
https://gizadeathstar.com/2023/04/personhood-urged-for-artificial-intelligences/
The last 15 minutes of the interview certainly shows the level of strain, if not outright despair Mr Yudkowsky is under.
This subject seems like the ultimate rabbit hole. So, the first personhood is granted to an AI–would there be any restriction on that “person” regarding upgrades? According to Mr Yudkowsky, since we don’t even know or understand exactly what’s going on “in there” with the current open GPT-4 program, how are we going to be confident that the first AI we declare as a person, is not just an avatar for an intelligence agency or some trillionaire bankster family, kinda like big corporations/foundations are today?
On a much simpler level, how do we know that the current AIs, that are apparently talking people into suicide, don’t have backdoors built into them (like the PROMIS software in the Bluebonnet article posted just after this one) to provide real-time access and control for some really bad actors?
Regarding the Bluebonnet article and bad actors, there’s a link within it to the latest article by Whitney Webb (sponsored by the Solari Report). The curtain is sure getting pulled back:
unlimitedhangout.com/2023/03/investigative-series/the-rise-of-jamie-dimon/
Whitney’s interviews with Redacted:
unlimitedhangout.com/2023/03/press/redacted-5/
http://www.youtube.com/watch?v=C4Mpmvo5SzQ