Sebastian Mallaby Spills the Beans on AI’s Most Elusive Minds
Demis Hassabis, co-founder of DeepMind, is often described as the visionary who’s single-handedly pushing humanity toward superintelligence. But what does that really mean, and how close are we to achieving it?
Sebastian Mallaby, a two-time Pulitzer Prize finalist and biographer of Hassabis, sat down with Tim Ferriss on his podcast, The Tim Ferriss Show, to discuss the intricacies of AI, its potential risks, and the importance of spotting breakthroughs early. Mallaby, a Paul A. Volcker senior fellow for international economics at the Council on Foreign Relations, has spent countless hours interviewing over 100 AI insiders, from industry leaders to researchers. This inside knowledge has given him a unique perspective on the world of artificial intelligence.
Lessons from the Trenches
Mallaby shared some insightful takeaways from his conversations with AI experts, including the concept of a “religion” surrounding AI. He noted that many researchers and entrepreneurs have become so enamored with the potential of AI that they’ve started to believe in its almost messianic power to solve all of humanity’s problems. This, he warned, can lead to a misplaced focus on short-term gains, rather than a more measured approach that considers the long-term implications.
“The AI community is very good at saying, ‘This is the future, and it’s going to be amazing,'” Mallaby said. “But they’re not as good at saying, ‘What are the potential downsides? How do we mitigate them? How do we make sure we’re not creating something that gets out of control?'”
The Great Unknown: Superintelligence</hassistant
While many experts predict that we’re on the cusp of a major breakthrough in AI, Mallaby remains skeptical about the notion of superintelligence. He pointed out that defining superintelligence is a moving target, and that it’s difficult to predict when (or if) we’ll actually reach it. Instead of focusing on a hypothetical end goal, Mallaby advocates for a more incremental approach to AI development, one that prioritizes transparency, explainability, and robustness.
“The problem with superintelligence is that it’s a bit like trying to predict the stock market,” Mallaby said. “We can’t forecast it, and we shouldn’t try to. What we should be trying to do is create a more robust and transparent systems, so that we can understand how they’re making decisions and make sure they’re aligned with our values.”
Spotting Breakthroughs Early
Mallaby also emphasized the importance of recognizing potential breakthroughs early, rather than waiting for them to become mainstream. By doing so, policymakers and industry leaders can better prepare for the implications of these advancements and work to mitigate any potential risks. This requires a combination of domain expertise, a deep understanding of the technology, and a willingness to engage with the broader AI community.
“The problem is that we’re not very good at spotting breakthroughs early,” Mallaby said. “We tend to wait until they become mainstream, and by then it’s often too late. But if we can identify these breakthroughs earlier on, we can start to think about the potential implications and work to address them proactively.”
What this means: As AI continues to advance at a breakneck pace, it’s essential that we take a more nuanced approach to its development. By prioritizing transparency, explainability, and robustness, and by recognizing potential breakthroughs early, we can work to ensure that AI serves humanity’s best interests, rather than creating a new class of superintelligent machines that may be beyond our control.
