NEW DELHI: Current cutting-edge models in artificial intelligence are developed by companies in a handful of countries, including the US and China, while most other countries are “passive victims of what they have created”, University of Montreal computer science professor Yoshua Bengio told a conference. Artificial Intelligence Impact Summit in New Delhi on Wednesday. Bengio, widely regarded as a pioneer in artificial intelligence, stressed that it was “unacceptable” that only a few countries are leading in the field and that countries must address the issue at the highest diplomatic level. Cutting-edge models in the field of artificial intelligence are the most advanced, large-scale, general-purpose machine learning models that currently push the boundaries of functionality, multimodality (text, image, audio, video) and performance. Yoshua said this is not only a moral issue, but also a sovereignty issue. It’s also about concentration of power, he added. “If AI capabilities continue to grow, it is likely that there will be a huge difference between the models in the United States and China and those being developed in other countries, even larger than they are now. This could make a huge difference for these two countries, or for any country that leads a huge economic power.…”, noting that “…the stability, the geopolitical stability that we have known since World War II could be in vain. I’m not saying it’s going to happen, but when you bring in so much power and concentrate it in this way, you do run the risk of destroying the house. “ These observations are significant at a time when India is spearheading the drive to democratize artificial intelligence. On the issue that India should be cautious when it comes to artificial intelligence, Bengio said people should be aware of the impact or risks that may affect society while developing the technology. “We need to understand it from a scientific perspective. We need to understand it from a social perspective because AI has a social component in psychology because we are talking about systems that interact with people and language. So countries like India can contribute to this understanding,” he said. He also said there is an urgent need for independent risk assessments before deploying open AI models. “If the risks are not too big compared to the benefits, because obviously there are benefits to sharing, especially in a developing country like India, then of course, you should definitely be open. But if you find that the risk crosses the threshold of social accessibility, then you shouldn’t do it. So, in this way, we might be able to get the benefits of open source when it makes sense and prevent catastrophic uses,” Bengio said. The professor also compared how a thorough risk assessment should be conducted before allowing the sale of medicines, and how the same should be followed when using artificial intelligence among the public. “You can’t make money doing whatever you want. You need to first show to independent groups, such as those representing the government, that your product will not be harmful. But there is no such thing yet. It is a scandal,” he said.
