Human brains learn new information better than AI robots, say experts
Technology
AI needs to process data hundreds or thousands of times to take it in
(Web Desk) - Human brains still learn new information better than artificial intelligence robots, experts say.
Tech bosses are hyping up a future of self-driving cars, instant cancer checks and computer-generated music.
But scientists say our good old grey matter has an extra step when learning that makes us smarter — for now.
They compared the human learning process to that of AI models and found we consolidate new pieces of knowledge, skills or memories before fine-tuning nerve connections to reduce errors when we recall them.
But Terminator-style machines learn by process of elimination, blocking errors until it has the answer, called “backpropagation”.
It means AI needs to process data hundreds or thousands of times to take it in.
But clever clogs humans, such as Albert Einstein, can connect ideas instantly.
The Oxford University study’s author, Dr Yuhang Song, said: “Learning in the brain is superior in many critical aspects.
“For example, compared to the brain, backpropagation requires more exposures to learn and suffers from catastrophic interference of new and old information.
“When trained on new tasks, performance on previous tasks is largely destroyed.
“Our biological systems often need to rapidly adapt to changing environments and have an advantage in this respect.”
Hundreds of experts signed a letter warning of humanity’s “risk of extinction” by AI.
