Meta has just announced a long-term study to better understand how the human brain processes language.
The researchers involved will look at how the brain and Artificial Intelligence (AI) language models respond to the same spoken or written sentence.
On this occasion, Meta collaborated with NeuroSpin, a Paris-based research center for innovations in brain imaging and the French National Institute for Digital Science Research (INRIA).
This work is part of Meta AI’s broader focus on human-level AI that can learn with little or no human supervision.
The company says even though AI has come a long way, its intelligence is still not as good as the brain at learning languages so it wants to try emulating it in software.
“We will use the insights from this work to guide the development of AI that processes speech and text as efficiently as humans,” Meta said in a statement on his official blog quoted by VOI, Friday, April 29.
“Over the past two years, we have applied deep learning techniques to public neuroimaging datasets to analyze how the brain processes words and sentences.”
Spoken language, Meta says, makes humans completely unique and understanding how the brain works is still a challenge and an ongoing process.
“What makes humans so much more powerful or so much more efficient than these machines? We wanted to identify not only the similarities, but also show the remaining differences,” said Meta.
As an example of the efficiency of the human brain over AI language models, Meta says children learn that the word “orange” can mean color and fruit after being shown just a few examples, while AI takes longer to learn.
If it can figure out how the brain works more efficiently, it can adapt the processes to work in AI models. The study has published two scientific papers outlining their findings.
The first finding is that the AI language model most closely resembles brain activity when trying to find the next word.
Another paper found that the brain anticipates words and ideas long in advance while language models are only programmed to work on the next word, by copying the brain in these areas AI models can be improved.
