Over the past two days, I participated in MIT's EmTech AI conference. (I have always wanted to visit to the MediaLab).
The most memorable moment came from Geoffrey Hinton, the godfather of AI and the winner of the Turing Award, who just this week resigned from Google, so he's free to share his concerns about the direction AI.
In a statement that left the audience in awe, Hinton said:
"It is quite conceivable that humanity is just a passing phase in the evolution of intelligence. You couldn't directly evolve digital intelligence it requires too much energy and careful fabrication. You need biological intelligence to evolve so that it can create a digital intelligence. The digital intelligence can then absorb everything people ever wrote in a fairly slow way (which is ChatGPT), but then it can start getting direct experience with the world and learn much faster. It may keep us around for a while to keep the power stations running, but after that maybe not…"
Wrap your head around that! <pause>
A few highlights:
► GPT4 knows more than we do. It's more efficient at storing, accessing, and learning. (This is due to their approach to "backpropagation"). Imagine being able to instantly recall every book, academic article, and discussion that has ever been written. He noted, "Smart things can outsmart us."
► Being digital allows many copies to do the same thing. Imagine 10,000 computers all studying the data and instantly sharing what they learn with all the other computers. Compare that to humans, where we sit in classrooms for many years to gain that same experience.
► He used the example of one doctor seeing a thousand patients or a digital doctor seeing 100 million patients. The AI doctor will see patterns and derive insights that humans ever could.
► AI's Manipulating Humans. Because AI will have access to all the world's data, including psychology, science fiction novels, and behavioral data, they will be able to manipulate humans with such fitness that humans may never know they're being manipulated. 🤯
When asked about the possibility of an international coordinated approach to manage AI, Hinton expressed skepticism, pointing out the difficulty in reaching global consensus on critical issues: "We can't even agree not to give assault rifles to teenage boys… and we're in a world where bad actors want to create robot soldiers." However, he did acknowledge that the world has successfully united in the past to limit nuclear proliferation. I sure hope Hinton allocates some of his amazing brainpower to convene the people to address this.
As someone who admires Hinton's work, I struggle with how to respond when the very father of AI technology warns us of a species-ending risk. <more to come on this>
Thanks to Will Douglas Heaven, for a great interview. Thanks to the MIT Technology Review crew for a great conference. 🙏🏼 to Marii Sebahar
Comments