Human-level AI is decades away, says Meta’s AI chief

0
56

[ad_1]

Right now, we can’t tell whether AI will reach human intelligence in 1 year, 5 years, 10 years, or 100 years. We have no more insight into the future of AI than a fortune cookie, but it’s not stopping people from making their predictions. Meta’s Chief AI scientist, Yann LeCun, just said that AGI (Artificial General Intelligence) is decades away, according to a new report.

AGI is basically the final form of AI; it’s AI that’s on par with a human being. This is the kind of AI that you’ve seen in sci-fi movies over the years. Right now, as powerful and scary as AI is, it’s miles behind a human brain. This means that it will be able to think for itself and do more than just digest massive amounts of text. So, AGI is the kind of AI technology that most people think will bring the extinction of humanity.

Meta’s AI chief says that AGI is decades away

In stark contrast to Jensen Huang’s statement that AGI is only five years away, Meta’s chief AI scientist, Yann LeCunn, said that human-level AI isn’t quite so close. Meta just held a large media event to celebrate the 10th birthday of its Fundamental AI Research Team. LeCun spoke at the event and laid down why he thinks that AGI isn’t around the corner.

One point he made makes a ton of sense; he said that we’re likely to get “cat-level” or “dog-level” intelligence before human-level intelligence. That’s just a natural progression, and if he’s right, then who knows if we’ll reach cat/dog-level intelligence in the next five years?

LeCun also touched on text as a source of knowledge. Companies train their AI on mountains of text data. We’re talking about social media posts, articles, books, and other sources. However, he said that “Text is a very poor source of information,”.

That’s another solid point, as human beings learn from several different sources. People learn from touch, smell, taste, sight, sound, experience, love, hate, fear, pain, and the list goes on to infinity. Sure, you can train a bot on text surrounding what it feels like to poke your finger, but it will never learn the feeling of it. “Train a system on the equivalent of 20,000 years of reading material, and they still don’t understand that if A is the same as B, then B is the same as A.”

He poked a hole in what Jensen Huang said

Yann LeCunn did cast a little bit of shade toward Jensen Huang, the CEO of Nvidia. Recently, Huang said that AI will be able to reach AI-level intelligence within five years. Whether or not that’s the case, LeCunn talked about his claim and said that Nvidia is set to profit from this whole AI craze. “There is an AI war, and he’s supplying the weapons.”, he said about Huang. ″[If] you think AGI is in, the more GPUs you have to buy,” which makes sense.

Nvidia is the prime source of processing power for companies building their AI LLMs. Companies like X and OpenAI purchase a dump truck-load of GPUs from the company to train their AIs. This pushed Nvidia’s revenue through the roof. Last year, the company’s Q3 2022 earnings report showed about $680 million in earnings. This year, the same report showed more than $9 billion. That’s a massive jump.

The future is still unclear

At this point, we still don’t know when AI will reach the level of the people training it. Companies are still grappling with the whole “AI hallucinations” issue. The raw processing power behind AI is getting more powerful as time goes on, and LLMs are being trained on more and more information, so it’s likely that it’s going to be smarter next year and the year after.

Also, we can’t rule out massive breakthroughs in AI technology that will leap-frog it years into the future. We’re all still waiting on more information about Q*, the project that OpenAI is working on. This is the breakthrough that the AI researchers allegedly warned would be a danger to humanity. Also, it’s the project that many folks believe led to the company’s CEO, Sam Altman, being fired. That’s all yet to be officially confirmed at this point.

In any case, AI technology is going to continue to improve, and we don’t know how long until it’s at the AGI level.

[ad_2]

Source link