“Nature is trying to tell us something here, which is that this doesn’t really work, but the field believes so strongly in its press clips, that it can’t see that,” he adds.
Even De Freitas’ colleagues at DeepMind, Jackie Kay and Scott Reed, who worked with him at Gato, were more circumspect when I asked them directly about his claims. When asked if Gatto is moving towards artificial general intelligence, they will not be attracted. “I don’t think it’s really worth making predictions about these kinds of things. I try to avoid that. It’s like predicting the stock market,” Kay said.
The question was difficult, Reid said. “I think most people who learn a machine will avoid the answer very closely. It’s very hard to guess, but, you know, I hope we get there one day.”
In a way, the fact that DeepMind called Gato a “general specialist” may have made it a victim of the AI industry’s excessive hype around AI. Current AI systems are called “narrow” AI, which means that they can only perform a specific and restricted set of tasks such as creating text.
Some technologists, including at Deepmind, believe that humans will one day develop “wider” AI systems that will be able to work as well or even better than humans. Some call this AI “general”. Others say it’s like “belief in magic.” Many senior researchers, such as Meta’s chief AI scientist Yann LeCun, are wondering if this is possible at all.
Jato is a “specialist” in the sense that he can do many different things at the same time. Andreas of the Massachusetts Institute of Technology says this is a world away from “general” AI that can purposefully adapt to new tasks that are different from what the model has been trained on. “We are still a long way from being able to do that.”
Making the models bigger also wouldn’t address the problem that models don’t have “lifelong learning”, which means they can be taught things once and they will understand all the implications of it and use it to inform all the other decisions they’re going to make say.
The hype around tools like Gato is detrimental to the overall development of AI, says Emmanuel Kahemboy, an AI/robots researcher part of Black in AI co-founded by Timnit Gebru. “There are many interesting topics left aside, underfunded, that deserve more attention, but that’s not what the big tech companies and the bulk of researchers at such tech companies are interested in,” he says.
Tech companies need to take a step back and assess why they build what they build, says Vilas Dar, president of the Patrick J. McGovern Foundation, a charity that funds AI projects “forever.”
“AGI is talking about something deeply human — the idea that we can become more of who we are, by building the tools that propel us to greatness,” he says. “And that’s really cool, but it’s also a way to distract us from the fact that we have real problems facing us today that we should try to tackle with AI.”
The hype around DeepMind’s new AI model misses what’s really cool about it
Source link The hype around DeepMind’s new AI model misses what’s really cool about it