Meta studies the human brain to teach “common sense” to its artificial intelligence

The scientific mind of Meta wants to create a large collection of models to help AI develop memory and common sense, but there are still many questions.

Yann LeCun is the vice-president of Meta and has always wanted to answer a question: “Can machines equal man? “. Now LeCun, who heads the AI ​​division, has published a new article showing how AI could become autonomous. A way to get closer to this very difficult answer that has always been sought. In fact, we’ve been investing in machine learning models for 70 years. To make them more efficient, more independent, in short, more human.

New research published by LeCun on Open Review.net proposes an architecture to minimize the number of actions a system must take to learn and perform an assigned task. The starting point is the human brain: in LeCun’s architecture we try to replicate his model to generate a more autonomous artificial intelligence. The most complex module to build is what LaCun calls “World Model”. It looks a lot like a long-term memory. Store, process, learn. This is how AI is heading in the world.

TWITTER |  The structure of artificial intelligence proposed by LeCun

TWITTER | The structure of artificial intelligence proposed by LeCun

Who’s Afraid of Artificial Intelligence?

For the moment it is impossible for a revolt of rebellious humanoids to conquer humanity, and it is thanks to a single very important detail. Artificial intelligence systems are pointless. I am unable to learn directly from real-world observations. Now, according to LeCun’s paper, there might be a way to train machine learning algorithms.

A practical example. A human being, to learn to drive, needs about ten hours and does not need to have an accident to learn that he must avoid it. “A car would have to fall off a cliff many times before it realized it was a bad idea,” LeCun explained, “and then thousands of times before it figured out how to avoid falling off the cliff. »

Trivially, the machines lack common sense at the moment. “It’s a practical problem because we want machines that make sense. We want self-driving cars, we want home robots, we want smart virtual assistants,” LeCun said. Common sense can be built for LeCun. Its goal is to collect a large collection of patterns that can help AI recognize the probable, possible, and impossible, and then act accordingly, imagining new solutions to unknown problems.

Who is Yann LeCun?

Yann LeCun is one of the world’s leading experts in artificial intelligence, machine learning, convolutional neural networks, computer vision and robotics. Meta’s scientific mind, he won the Turing Prize in 2019, the equivalent of the Nobel Prize in the field of information technology. The future of autonomous machines can only be in his hands. It’s a legendary business he’s been running for years. The learning systems on which many modern AI applications are based today are his work. There are already those who recognize him as the father of the next generation of artificial intelligence.

Many experiences, happy or not

In 2013, he founded the Facebook group AI Research (FAIR), Meta’s first foray into experimenting with AI research. Earlier this year, the Meta chatbot BlenderBot3 sparked the debate over AI ethics and biased data. The baddest robot ever, after not even a day of interacting with users on Twitter, he became a racist, anti-Semitic and Holocaust denier conspiracy theorist.

We wish to thank the author of this article for this outstanding web content

Meta studies the human brain to teach “common sense” to its artificial intelligence


Our social media pages here and additional related pages here.https://www.ai-magazine.com/related-pages/