Mori responded to this observation by citing an example. “Some researchers asked ChatGPT to answer the questions in World Values Survey, and supposedly the answers it gave tracked perfectly with Western values. Another interesting thing about it was that when they told ChatGPT to ‘answer like a Japanese person’, it gave a Japanese-like response.” This only further attested to the difficulty and importance of alignment.

＊The trolley problem: a thought experiment that asks what one would do in a situation where a trolley has lost its brakes and is speeding down the track, and if it continues it will kill five people, but if one flips the switch and sends it down another path, one person will be killed.

Mori probed even further into the topic of AI’s most cutting-edge fields, particularly about AI agents that can act autonomously. “The issue is whether that kind of personal AI agent really has your best interests in mind. For all you know it could be an advocate for the interests of some company. It would be hard for a human being to tell which AIs can be trusted,” he pointed out.

Lastly, Mori asked Kelly how he has managed to remain a “protopian” (one who believes the world is slowly marching toward incremental improvement) despite being in San Francisco, a place where extreme views coexist, like those of AI Boomers who believe that AI will bring about a utopia, and AI Doomers who predict a dystopia instead.
 
Kelly said there were two catalysts involved. “The first was my experience traveling throughout Asia in my younger days. Since then, I’ve seen how different countries and regions like Taiwan, India, Pakistan, and Afghanistan have grown from poverty to wealth in 50 years, and it led me to believe from the bottom of my heart in people’s ability to make the impossible possible. The other would be the changes I’ve seen in the digital world. The arrival of the Internet allowed people to connect with each other, and this made me realize how it could unleash the ‘forces of good.’”
 
Kelly then addressed the audience directly. “In the early days of the Internet, the experts said that ‘no one’s going to buy food or cars online.’ Now, many people are saying the same kinds of things about AI. But I don’t think so. You have to expand your thinking a bit, use your imagination to envision the best-case scenario, and then move in that direction little by little. It’s important to have the ability to believe that nothing is impossible.”
 
“Mr. Kelly, you really are putting knowledge into action. Instead of being pessimistic about the future, I hope that today, we can start moving forward to prepare for the best future, armed with our imaginations,” Mori said, concluding the talk.