Paris, France – June 14: Meta’s vice president and chief AI scientist Jan Lekun attends Viva …(+)
Getty Images
When we talk about the chronology of artificial intelligence research in our lives, there are some interesting timelines. The right time to move forward.
By combining it with the subsequent wave of technology, like the ones happening now, perhaps multiple fundamental AI revolutions, there’s good insight into what’s going on in the business world. .
Yann Lecun led research in Meta. He is an outstanding voice of technology, and took the stage in Davos this January to talk about some of the dynamics here.
When I interviewed Lecun in today’s AI segment, I thought about this timeline and how to inform the work that MIT people and others are doing now.
Yann Lecun: A Journey to Technology
From a young age he said he always found this type of science attractive. His parents took him to see him in 2001: A Space Odyssey was actually informed by the advice of Marvin Minsky.
“One of the things I always found attractive since I was a child was the emergence of animal and human intelligence,” he says, detailing how it leads to the work of the neural net. did.
He then explained, he signed a role in Meta, making it clear that the work done there was open source and could remain in New York and teach at NYU.
The open source part is also important. And Lecun was highlighted about the nature of work that should have arisen to guide the next AI era in a democratic way.
AI Winter and how it happened
Watch the video from the interview and you can see that they are talking about Marvin Minsky. This individual really informed the world of AI in a basic way, starting with his work at Perceptron in the 1950s and collaborating with others like Seymour Papert (Lecun is a language). He also mentions Piaget and Chomsky, who are talking about academic studies. Essentially, Lecun suggested that the limitations of data and learning models in the 1980s convinced Minsky and others that it was time for AI winter hours or work pauses.
In examining this final slump in long-standing AI research, Lecun pointed out that when he asked Minsky about this and this, the now-deceased scientist had stood up to his previous decision.
“I actually discussed this once with Minsky,” Lekun said. “I was a very young student at the time, otherwise I wouldn’t have appeared.”
The future of super intelligence
I asked Lecun about the artificial general information outlook and he said he didn’t like the term. He likes “AMI” (Advanced Machine Intelligence) and there was a rather bold answer to the mass of voices suggesting that “AI will reach a place that is smarter than us.”
“The idea that somehow intelligence is kind of linear scale is nonsense,” he said. “Your cat is smarter than certain things, you’re smarter than certain things, you’re smarter than certain things, you can beat you with a $30 gadget you can do, chess, you can beat you with chess. …The idea that it somehow is on a linear scale and will be the event when you reach AGI at some point is totally nonsense. It will be progressive.”
Ultimately, we suggested that AI will become “smart” in a human way, but there is no single turning point. What he explained was a kind of concept of ultimate singularity.
“There’s no doubt that at some point in the future there’s a system as smart as humans.
Lecun also weighed the generation AI and how it would be replaced by a different model called the Joint Embedded Prediction Architecture (JEPA). “The genai has a shelf life of three years,” he said.
Two AI revolutions
Back to this idea of open source when looking at how quickly these technologies are evolving, Lecun is committed to open source in the same way that society maintains, and is very diverse. We have suggested how to build a basic model and change the flywheel of innovation. Diverse reporting and democracy. He mentioned Pytorch as a means of ChatGpt and other technology, as the value of “free exchange and acceleration” in LLM and Neural Net Research. Regarding the challenge, he argued that the benefits of AI outweigh the risks.
“Everyone will be smarter,” he said. “AI is not going to kill us all.”
Lecun predicted a different blueprint for objectively driven AI with next-generation tools, safer controller systems, and “common sense.”
And then there’s SSL:
“I think self-monitored learning is probably the most innovative concept that has really completely changed the way machine learning practices over the past decade or so,” he said. “And there’s all these things like systems that are enhanced with associated memories (e.g.).”
He mentioned the open source wearables:
“In the future, all of our interactions with the digital world will be mediated by our AI assistants… smart devices like glasses,” he said. “All of our information diet comes from AI assistants. You can’t come from a handful of companies on the west coast of the US or China. It has to be very diverse. And it’s The only way to be diverse is related to the previous question and is the basic model of open source. The open source foundation model is a vertical application, or all languages, all cultures, all values in the world The system has been fine-tuned to learn. How to acquire a diverse population of AI assistants.
In conclusion, Lecun had some words in history at this time as well.
The last time this (a kind of innovation) happened on a large scale followed the invention of the printing press. And it sparked the emergence of the American Revolution, the French Revolution, and the democracy,” he said. “This had a big impact () right? It’s just the nature of knowledge. So, AI may have the same effect, but the next step (I’m going to go) is the new Renaissance.”
It’s all fascinating to see where we’re headed. We wanted to present these statements to the next generation of people working on Genai and AGI Systems.
