Terence Tao: AI should not be a monopoly held by one or two companies

Tao says AI is theoretically capable of doing many kinds of work and could be used for harmful purposes such as deepfakes, but current systems are still limited and he is not highly concerned about them taking over. He expects more capable systems in 10 or 20 years, alongside more experience in how to deal with them.

He also says it is not good for something as important as AI to be a monopoly controlled by one or two companies. He notes that open-source models exist, though they lag behind big commercial systems, and says some regulation will be needed because AI can damage trust in photos, videos, and elections.

Reading notes

#

• Tao considers AI a very powerful technology, with the potential to affect almost everything we do, but says current systems are still limited and good mainly at guessing answers.
• He sees a risk of AI being used for deepfakes, propaganda, and influence in elections, although he says the scenario of superintelligence dominating the planet is still far off.
• Tao states that it is not good for AI to be in a monopoly of one or two companies, because the technology is too important to depend only on a few agents.
• He says that there are open models, still behind commercial ones, and that the cost of training large systems should fall over time.
• Tao advocates some regulation of AI and warns that the possibility of falsifying videos and photos can make people stop trusting even genuine records.