Yuval Noah Harari: uncertainty

Transcript context

…Could we optimize for it? If social scientists and say psychologists establish that, I don’t know, 17% inattention is good for a conversation because then you feel challenged, “Oh, I need to grab this person’s attention,” you can program the AI to have exactly 17% inattention, not one percentage more or less. Or it can by trial and error, discover what is the ideal percentage. Over the last 10 years, we have creating machines for grabbing people’s attention. This is what has been happening on social media. Now, we are designing machines for grabbing human intimacy, which, in many ways, it’s much, much more dangerous and scary. Already the machines for grabbing attention, we’ve seen how much social and political damage they could do by in many way kind of distorting the public conversation. Machines that are superhuman, in their abilities to create intimate relationships, this is psychological and social weapons of mass destruction. If we don’t regulate it, if we don’t train ourself to deal with it, it could destroy the foundations of human society. Well, one of the possible trajectories is those same algorithms would become personalized and instead of manipulating us at scale, there would be assistants that guide us to help us grow, to help us understand the world better. Even interactions with large language models now, if you ask them questions, it doesn’t have that stressful drama, the tension that you have from other sources of information. It has a pretty balanced perspective that it provides. It just feels like the potential is there to have a really nice friend who’s an encyclopedia that just tells you all the different perspectives, even on controversial issues, the most controversial issues, to say, these are the different theories. These are the not widely accepted conspiracy theories, but here’s the kind of backing for those conspiracy. It just lays it all out. Then with a calm language, without the words that kind of presume there’s some kind of manipulation going on underneath it all. It’s quite refreshing. Of course, those are the early days. People can step in and start to sensor it to manipulate those algorithms, to start to input some of the human biases in there as opposed to what’s currently happening is kind of the internet is input, compress it, and have a nice little output that gives an overview of the different issues. I mean, there’s a lot of promise there also, right?…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.