[ad_1]
STANFORD, Calif. – At first glance, it may be difficult to understand why an expert in cognitive development of children was one of the keynote speakers at a conference on artificial intelligence.
It turns out that children – and those who understand how they learn – can have a lot to learn from artificial intelligence experts on how to improve their systems.
Artificial intelligence systems have been able to absorb a lot of data and use it to make predictions based on all this information, said Alison Gopnik on Monday during a roundtable at a sponsored event by the Stanford Man-centered Artificial Intelligence Institute. But they are not very good at generalizing from small amounts of data, said Gopnik, a professor who studies cognitive development at the University of California at Berkeley. They are also not good at collecting data to make these generalizations or to become familiar with the world from clues given by other intelligent entities around them, she added.
But babies and young children excel in all these things, she said.
"So, these three things – model building, exploration, and social learning – are clues to how kids can learn a lot, and these are things that are just the beginning. terms of what the AI can do, "said Gopnik, the author of" The scientist in the cradle: what early learning tells us about the mind. "
Read it:Artificial intelligence may soon be omnipresent: here is how it could overcome 8 different industries
Researchers in artificial intelligence are already using what psychology experts such as Gopnik have discovered about how children learn and apply it to their field. Gopnik herself works with researchers at her university to develop AI systems intended to be curious, like children. They are designed to collect data on their own, she said.
One of the key ideas that has allowed AI research to progress so rapidly over the last five to ten years is the awareness of the need to design a curriculum, a teaching program of their systems, said Demis Hbadabis, co-founder of the Alphabet-owned AI laboratory. DeepMind, who was sitting on the same panel as Gopnik. They could not expect their systems to master the tasks immediately, but had to let them develop gradually, mastering the process steps, Hbadabis said.
"You can not just go from zero to one," he said. "You must actually create simpler versions of the task and develop the way we teach children."
One of the promises of modeling AI after the brain development of children is that the technology could be much more effective. Instead of relying on huge data sets and great computing power to make sense of the world, such childish artificial intelligence systems could potentially rely on far less data and power, "said Chris Manning. Professor of Linguistics and Computer Science at Artificial Intelligence Stanford Laboratory.
"You can get a more effective learning of several orders of magnitude," said Manning.
Do you have a tip? Contact this reporter by email at [email protected], send him a message on Twitter. @troywolv, or send him a secure message via Signal at 415.515.5594. You can also contact Business Insider securely via SecureDrop.
[ad_2]
Source link
