On December 4th, AI Sweden hosted an exclusive event with Nobel Prize laureate professor Geoffrey Hinton.
“I've been amazed by how fast things have happened in the last few years. I thought I wouldn't live to see it,“ he said in a conversation with AI Sweden’s Senior Scientific Advisor Daniel Gillblad.
An audience of high-level decision-makers and leading AI scientists from AI Sweden’s partners were given the special chance to hear professor Hinton reflect both on his work on neural networks in the 1980s as well as more recent development in the field – and what they mean to society at large.
When asked if there are missing components for reaching human-level AI, Hinton answered:
“Nothing that is crucial. We may hit a limit where we run out of training data, but then what we can do is get them to do reasoning and generate their own training data, much like AlphaGo plays against itself and generates training data.”
One question from the audience touched upon this rapid development and the ever-changing landscape organisations that want to use AI in their businesses have to orient themselves in. “Hang on, you're complaining that we keep doing new things,” professor Hinton joked before addressing the question.
“It used to be that you could make some nice product and then just rake in the profits for the next 20 years. And it isn't gonna be like that. It's gonna be that you have to be agile and just assume you're gonna get your profits over [just] a few years.”
The discussion further addressed professor Hinton's concerns about AI safety. He expressed frustration with journalists who conflate different types of risks. He also elaborated on his belief that regulatory pressure must be applied to the largest AI labs, compelling them to invest more in this area.
Professor Hinton shared that while some challenges need to be addressed with huge resources and international cooperation, there are others he believes to be more easily solved. Such as how to ensure privacy when training models on sensitive data.
The panel that followed delved into the impact of AI on society, noting the important contributions research brings while underlining the need for involving both industry and the public sector.
I think there’s a lot of value in supporting researchers, not least PhD students, and research environments. Not only to progress on the research, but also to be able to communicate the research. And both industry and the public sector play very important roles in order to ensure that the research has relevance.
Amy Loutfi
Professor in computer science & co-director Wallenberg AI, Autonomous Systems & Software Program
That's the challenge, really; to make people start working on issues that they feel are relevant, whether it's some curiosity-driven thing or whether they're solving real life problems. Enabling people to explore these things has spillover effects, and the spillover effects are what we're actually trying to get to. That's what's creating innovation and all these new fantastic things that we are seeing in our world today.
Hanifeh Khayyeri
VP of Computer Science, RISE
The bottleneck right now is not to create better AI. The bottleneck is to solve real problems in the world to unlock the benefits of all of this. I think we should focus more on that in society. I wish I wasn't one of the few AI entrepreneurs. I wish there were thousands of us, of people who are at the bleeding edge of taking the systems and the capabilities to solve real problems out there.
Anton Osika
Co-founder Lovable
Photographer: Sandra Humer