87. Evan Hubinger - The Inner Alignment Problem
Towards Data Science
06/09/21
•69m
About
Comments
Featured In
How can you know that a super-intelligent AI is trying to do what you asked it to do?
The answer, it turns out, is: not easily. And unfortunately, an increasing number of AI safety researchers are warning that this is a problem we’re going to have to solve sooner rather than later, if we want to avoid bad outcomes — which may include a species-level catastrophe.
The type of failure mode whereby AIs optimize for things other than those we ask them to is known as an inner alignment failure in the context of AI safety. It’s distinct from outer alignment failure, which is what happens when you ask your AI to do something that turns out to be dangerous, and it was only recognized by AI safety researchers as its own category of risk in 2019. And the researcher who led that effort is my guest for this episode of the podcast, Evan Hubinger.
Evan is an AI safety veteran who’s done research at leading AI labs like OpenAI, and whose experience also includes stints at Google, Ripple and Yelp. He currently works at the Machine Intelligence Research Institute (MIRI) as a Research Fellow, and joined me to talk about his views on AI safety, the alignment problem, and whether humanity is likely to survive the advent of superintelligent AI.
Previous Episode
86. Andy Jones - AI Safety and the Scaling Hypothesis
June 2, 2021
•85m
When OpenAI announced the release of their GPT-3 API last year, the tech world was shocked. Here was a language model, trained only to perform a simple autocomplete task, which turned out to be capable of language translation, coding, essay writing, question answering and many other tasks that previously would each have required purpose-built systems.
What accounted for GPT-3’s ability to solve these problems? How did it beat state-of-the-art AIs that were purpose-built to solve tasks it was never explicitly trained for? Was it a brilliant new algorithm? Something deeper than deep learning?
Well... no. As algorithms go, GPT-3 was relatively simple, and was built using a by-then fairly standard transformer architecture. Instead of a fancy algorithm, the real difference between GPT-3 and everything that came before was size: GPT-3 is a simple-but-massive, 175B-parameter model, about 10X bigger than the next largest AI system.
GPT-3 is only the latest in a long line of results that now show that scaling up simple AI techniques can give rise to new behavior, and far greater capabilities. Together, these results have motivated a push toward AI scaling: the pursuit of ever larger AIs, trained with more compute on bigger datasets. But scaling is expensive: by some estimates, GPT-3 cost as much as $5M to train. As a result, only well-resources companies like Google, OpenAI and Microsoft have been able to experiment with scaled models.
That’s a problem for independent AI safety researchers, who want to better understand how advanced AI systems work, and what their most dangerous behaviors might be, but who can’t afford a $5M compute budget. That’s why a recent paper by Andy Jones, an independent researcher specialized in AI scaling, is so promising: Andy’s paper shows that, at least in some contexts, the capabilities of large AI systems can be predicted from those of smaller ones. If the result generalizes, it could give independent researchers the ability to run cheap experiments on small systems, which nonetheless generalize to expensive, scaled AIs like GPT-3. Andy was kind enough to join me for this episode of the podcast.
Next Episode
Few would disagree that AI is set to become one of the most important economic and social forces in human history.
But along with its transformative potential has come concern about a strange new risk that AI might pose to human beings. As AI systems become exponentially more capable of achieving their goals, some worry that even a slight misalignment between those goals and our own could be disastrous. These concerns are shared by many of the most knowledgeable and experienced AI specialists, at leading labs like OpenAI, DeepMind, CHAI Berkeley, Oxford and elsewhere.
But they’re not universal: I recently had Melanie Mitchell — computer science professor and author who famously debated Stuart Russell on the topic of AI risk — on the podcast to discuss her objections to the AI catastrophe argument. And on this episode, we’ll continue our exploration of the case for AI catastrophic risk skepticism with an interview with Oren Etzioni, CEO of the Allen Institute for AI, a world-leading AI research lab that’s developed many well-known projects, including the popular AllenNLP library, and Semantic Scholar.
Oren has a unique perspective on AI risk, and the conversation was lots of fun!
If you like this episode you’ll love
Promoted