- Current artificial intelligence systems like ChatGPT don’t have human-level intelligence and aren’t as smart as dogs, said Yann Lucan, head of AI at Meta.
- Talking about the limitations of generative AI such as ChatGPT, LeCun said AI is not very intelligent because it is trained only on language.
- Meta’s LeCun said that in the future there will be machines that are more intelligent than humans, but that shouldn’t be seen as a threat.
Meta Yann LeCun’s chief AI scientist spoke at the Viva Tech conference in Paris, saying that artificial intelligence does not currently have human-level intelligence, but it could someday. I said yes.
Chesnot | Getty Images News | Getty Images
Current artificial intelligence systems like ChatGPT don’t have human-level intelligence and are barely smarter than dogs, says Meta’s head of AI, amid growing debate over the dangers of burgeoning technology. Stated.
Developed by OpenAI, ChatGPT is based on so-called large-scale language models. This means that AI systems are trained on large amounts of linguistic data to prompt users with questions and requests, while chatbots respond in a language we understand.
The rapid development of AI has raised concerns among leading technologists that the technology could pose a danger to society if left unchecked. Tesla CEO Elon Musk said earlier this year that AI is “one of the greatest risks to the future of civilization.”
At the Viva Tech conference on Wednesday, Jacques Atari, a French economic and social theorist who writes about technology, said AI is good or bad depending on how it is used.
“If we use AI to develop more fossil fuels, it will be a disaster. [to] “If you develop worse weapons, it will be worse,” Atari said. “On the contrary, AI could be great for health, for education, and for culture.” rice field.
At the same panel, Yann Lucan, chief AI scientist at Facebook parent company Meta, was asked about the current limits of AI. Focusing on generative AI trained on large language models, he said artificial intelligence is not very intelligent because it is taught only by language.
“These systems are still very limited and have no understanding of the underlying realities of the real world because they are trained purely on text, on large amounts of text.” said Lucan.
“Most of human knowledge has nothing to do with language, so part of human experience is never brought to AI.”
LeCun added that AI systems can now pass the bar exam, the exam required to become a lawyer in the United States. But while AI can’t load a dishwasher, a 10-year-old can “learn it in 10 minutes,” he said.
“It shows that we are missing something really big to reach not just human-level intelligence, but even dog intelligence,” Lucan concluded.
Meta’s head of AI said the company is working on training AI with video, not just language, but that this is a more difficult task.
As another example of current AI limitations, a 5-month-old baby will see a floating object and try not to think about it too much, he said. However, your 9-month-old baby will be surprised to see this item and realize that the object should not float.
“I have no idea how to replicate this ability in today’s machines,” LeCun said. “Until we can do this, we won’t have human level intelligence, and we won’t have dog or cat level intelligence.” said. [intelligence]. ”
“It is well known that humanity will face many dangers in the next 30 to 40 years,” Atari said in a pessimistic tone about the future.
He said climate change and war were his top concerns, and he feared robots would “turn against us.”
During the conversation, Meta’s LeCun said that in the future there will be machines that are more intelligent than humans, but that they shouldn’t be seen as dangerous.
“We should not see this as a threat. It will be like a staff member that supports the ,” LeCun said. He said.
The scientist added that these AI systems should be created to be “controllable and fundamentally subordinate to humans.” He also rejected the idea that robots would conquer the world.
“Fear spread by science fiction [is]If robots are smarter than us, they’ll want to conquer the world…there’s no correlation between being smart and wanting to conquer,” LeCun said.
While weighing the dangers and opportunities of AI, Atari concluded that guardrails must be put in place for the development of the technology. But he wasn’t sure who would do it.
“Who draws the border?” he asked.
AI regulation is a big topic at Viva Tech. The European Union is pushing for its own AI bill, but a top French government minister told CNBC this week that the country wants global regulation of the technology.
