Artificial intelligence comes with its power and pitfalls, something Yann LeCun, who is charging the technology’s boom, knows well. The chief AI scientist at Meta is not the person who falls in line to say everything is all right with AI or join a chorus only to praise the progress the technology has made. He is always cautiously careful.
What he is sure of is the importance of embracing an open-source approach. “We don’t have a monopoly on good ideas,” he told a crowd at Meta’s Build with AI Summit in Bangalore last week. This is soon after having a light breakfast, spent discussing plans for the day.
“Open source as well as research enable an entire ecosystem. It gives a career perspective for young people. I can give you an example of what happened in France some years ago. Paris is now probably the second most vibrant place for AI startups in the world, after a couple of places in the US. I’m sure Bangalore is not far behind and is probably progressing quickly. And the effect has been psychological. That, in turn, has had a huge effect. So, the first effect is that when you have an industry practising ambitious research in AI, it gives hope to young people to contribute to it. Open source tools and the open source community are (there) worldwide. It has the effect of motivating people to go into technology, to perhaps go into graduate studies in AI. They learn how to produce new things, how to invent new things, and then sort of help expand the ecosystem. We can sort of use this phenomenon as a blueprint to accelerate and further the progress of AI in India, which is already incredibly vibrant,” he said.
‘There is a lot of talent in India’
LeCun looks chic like any Parisian should. Now a professor at New York University, he sports a pair of Ray-Ban Meta Smart Glasses, which is all the rage at the moment. Unofficially dubbed the “godfather” in his area of expertise, he can count as friend another “godfather” who recently was awarded a Nobel Prize in physics — Geoffrey Hinton, who has warned repeatedly about AI’s existential threats.
LeCun’s work with colleagues at Bell Labs in the 1980s allowed him to design a neural network that could recognise handwritten numbers at a high level of accuracy. It was a milestone moment. The Association for Computing Machinery awarded LeCun, Hinton and Yoshua Bengio (together known as the ‘Godfathers of AI’) a Turing Award in 2018 for what it called the “conceptual and engineering breakthroughs that have made deep neural networks a critical component of computing”.
The man is keen on the potential young India has to offer. “Going forward, India has an important role to play, not just in technology development and product development, not just for local products or international products, but also for research. There is a lot of talent in India. We see a lot of people from India making major contributions to AI, both technical and scientific contributions. India can do this in a much bigger way than it currently is doing. I was visiting IIT Madras yesterday (last week) and saw a bunch of really interesting projects, particularly in natural language understanding, like a system that, using LLaMA (a family of autoregressive large language models released by Meta AI), can translate hundreds of languages. We have technology now for systems that can do speech-to-text, text-to-speech, or even translation speech-to-speech, including for languages that are not written. It’s important for regions like India, obviously, but also for Indonesia, for example. There are a number of languages in Indonesia. It would enable (many) languages to survive. I think AI is going to help in the process,” he said.
In July, Meta CEO Mark Zuckerberg wrote an open letter to reinforce his company’s stance — open source development of artificial intelligence would allow technologists to learn how powerful AI models are created and use that knowledge to build their own AI programmes. He said a handful of companies can’t keep AI a secret. He seems to be mirroring the mind of LeCun.
“Open source will ensure that more people around the world have access to the benefits and opportunities of AI, that power isn’t concentrated in the hands of a small number of companies, and that the technology can be deployed more evenly and safely across society,” Zuckerberg said.
Of course, such motives can’t be all altruistic since the more other companies use Meta’s services, the more its own products are standardised across the industry, ensuring Zuckerberg bypasses products from companies like Apple and Google. “We must ensure that we always have access to the best technology, and that we’re not locking into a competitor’s closed ecosystem where they can restrict what we build,” he said in the letter.
‘Training will be distributed’
Meta has a powerful AI algorithm called LLaMA and a few months ago it added support for several additional languages — including Hindi, French and Spanish — for Meta AI, the company’s AI-powered smart assistant.
“AI is going to become, basically, a common infrastructure that all of us can use in the future and share. We need AI systems in the future to become a kind of repository of all human knowledge. And, currently, we’re sort of trying to do this, but we can’t really do it (all by ourselves). The reason is that we don’t have the diversity of data in terms of languages, cultural preferences, value systems, centres of interest. We cannot centralise all the data in a single place to train a system that would be the repository of all human knowledge,” LeCun said.
He continued: “Once those systems are pre-trained, even if we had access to the data, it would have to be fine-tuned by people who speak all the languages that the system has. All the cultural backgrounds are necessary, and all the centres of interest. And no single entity, as big as Meta might be, can do this. It has to basically take the entire world. It has to be quite smart. And so, what I see as the future is that the big frontier AI systems will not be produced by big companies or developed countries. I think the main ones will be trained in a distributed fashion all across the world. There will be data centres everywhere in India, North America, Europe, other parts of Asia, and Africa. The training will be distributed.”
All this brings us to the question of whether we should be afraid of artificial intelligence. “I’m familiar with the concept of people who are smarter than me,” LeCun said with a smile.
“AI is going to amplify human intelligence. Everybody is going to be smarter for it. And it’s not just people like us in the tech community or in academia or things like that. It’s going to be everyone. Everyone in India, in rural areas, they can ask questions to their AI assistants in their own language, say about health issues. And it’s a very different future that will become possible.”
To reach the next level of AI — or what LeCun calls advanced machine intelligence or AMI, meaning “friend” in French — there is a need for systems that can help people in their daily lives. This involves developing systems that can understand cause and effect and model the physical world.
