Both open source models and closed source models are playing essential roles in the development of artificial intelligence (AI) technology, said Yang Song, a technical staff member of OpenAI and incoming assistant professor with the California Institute of Technology.
"I think in the future there will continue to be open source models and closed source models. They will become more and more powerful over time and maybe compete with each other a little bit. I view both of them essential to the field," Song said in a recent interview with China Global Television Network in Beijing.
"Open source models are very important to democratize AI research. It really allows people to have wide access to this technology. In the meanwhile, I also understand that not all models can be open-sourced. There might be security safety issues that concern people. There might also be IP considerations," he said.
Song said there is a gap of computational power between the AI academia and industry, and that the academia is geared toward early ideas on innovation.
"Starting from the very beginning, the academia just doesn't have much compute, and [while] industry has more compute. But mostly that means we focus on different things. Our research topics can still be done in academia without access to lots of, lots of compute. I remain optimistic that academia, even though without access to tons of compute, can still make very reasonable progress in the field of AI," Song said.
Both open source, closed source models essential to AI development: expert
Attempts to curb China's scientific and technological advancement are futile, a fact that has already been proven, said Kishore Mahbubani, former permanent representative of Singapore to the United Nations, in an interview aired Friday.
In an exclusive interview with China Central Television (CCTV) in Beijing, Mahbubani said he had stated this position in one of his articles published in the United States.
"Actually, I published an article, you know the two, I guess two leading journals in the United States on international relations. One is Foreign Affairs and the other is Foreign Policy. And last year I co-authored an article with two other co-authors, saying that all the efforts to stop China's scientific and technological development will fail. And it has failed always. You know, for example, the Soviet Union tried to prevent the spread of nuclear technology to China, China develops its own. The United States didn't want to share its technology on international space station with China. China develops its own space station. So clearly, efforts to stop China in the area of scientific innovation and technological development have failed. And so it'd be wiser for the West, including United States, to work with China other than to try and stop China seek development," he said.
Regarding China's progress on robots, Mahbubani said China is leading the world in the sector and hopes the country will share its expertise with the rest of the world.
"If there's one country that is preparing for the future well, it is China, because one in six human beings in the world is Chinese. But one in three robots in the world is Chinese, and one in two baby robots being born every day is Chinese. So China is producing far more robots than any other country is. So clearly it's preparing for the world of the future when we will have, for example, labor shortages, as you know, as you develop an aging society. So China is wisely investing in robots. But I hope that China will also share its learning and expertise with other countries. Also because the robots like that can also be helpful even to developing countries cause you can enhance the productivity of their populations, of their factories and so on so forth. So the world should be happy that China is leading the world in manufacturing, producing robots," he said.
Attempts to stop China's sci-tech development doomed to fail: former Singaporean diplomat
