In a $30 million mansion perched on a cliff overlooking the Golden Gate Bridge, a bunch of AI researchers, philosophers, and technologists gathered to debate the top of humanity.
The Sunday afternoon symposium, known as “Worthy Successor,” revolved round a provocative idea from entrepreneur Daniel Faggella: The “ethical goal” of superior AI needs to be to create a type of intelligence so highly effective and clever that “you’d gladly want that it (not humanity) decide the long run path of life itself.”
Faggella made the theme clear in his invitation. “This occasion could be very a lot targeted on posthuman transition,” he wrote to me by way of X DMs. “Not on AGI that eternally serves as a software for humanity.”
A celebration crammed with futuristic fantasies, the place attendees focus on the top of humanity as a logistics drawback relatively than a metaphorical one, may very well be described as area of interest. In the event you dwell in San Francisco and work in AI, then this can be a typical Sunday.
About 100 company nursed nonalcoholic cocktails and nibbled on cheese plates close to floor-to-ceiling home windows dealing with the Pacific ocean earlier than gathering to listen to three talks on the way forward for intelligence. One attendee sported a shirt that mentioned “Kurzweil was proper,” seemingly a reference to Ray Kurzweil, the futurist who predicted machines will surpass human intelligence within the coming years. One other wore a shirt that mentioned “does this assist us get to secure AGI?” accompanied by a pondering face emoji.
Faggella informed WIRED that he threw this occasion as a result of “the large labs, the those who know that AGI is more likely to finish humanity, do not speak about it as a result of the incentives do not allow it” and referenced early feedback from tech leaders like Elon Musk, Sam Altman, and Demis Hassabis, who “have been all fairly frank about the opportunity of AGI killing us all.” Now that the incentives are to compete, he says, “they’re all racing full bore to construct it.” (To be honest, Musk still talks about the risks related to superior AI, although this hasn’t stopped him from racing forward).
On LinkedIn, Faggella boasted a star-studded visitor checklist, with AI founders, researchers from all the highest Western AI labs, and “a lot of the necessary philosophical thinkers on AGI.”
The primary speaker, Ginevera Davis, a author based mostly in New York, warned that human values could be unimaginable to translate to AI. Machines could by no means perceive what it’s prefer to be aware, she mentioned, and attempting to hard-code human preferences into future techniques could also be shortsighted. As a substitute, she proposed a lofty-sounding thought known as “cosmic alignment”—constructing AI that may search out deeper, extra common values we haven’t but found. Her slides usually confirmed a seemingly AI-generated picture of a techno-utopia, with a bunch of people gathered on a grass knoll overlooking a futuristic metropolis within the distance.
Critics of machine consciousness will say that giant language fashions are merely stochastic parrots—a metaphor coined by a bunch of researchers, a few of whom labored at Google, who wrote in a famous paper that LLMs don’t truly perceive language and are solely probabilistic machines. However that debate wasn’t a part of the symposium, the place audio system took as a given the concept superintelligence is coming, and quick.