Expert explains why being rude to AI can actually help you get smarter answers.
Transcript:
Caroline Woods: How do you balance AI with was also having a human touch because as someone who leverages ChatGPT often, it’s wrong a lot?
Maja Vujinovic: Yeah, absolutely. You almost have to tell it that it’s stupid. And don’t be stupid in order to give you good answers. Look, jobs like yours, for example, or jobs like in therapists, in school. Teachers, those will not be replaced anytime soon. I don’t envision a robot in your place, interview me anytime soon. So so I think things will evolve, but they are not going to just suddenly changed in the next year or two. I think a lot of people out there are making a lot of money by saying we’re all done right. But the reality is there are certain emotional things that we still don’t have with AI and we won’t have now. I think Dennis from DeepMind, Google DeepMind has predicted that, and they’ve been quite consistent, actually with this. And in the next 10 years, they’ll probably have AGI. Now there is a thing where you can say, how do you define AGI. Right but I do kind of turn to them quite often. And turned to him to tell me, when does it become really serious. When I starts to develop these emotions that humans have. Still, my prediction is we won’t be seeing any of that at scale for another 20 years.
Caroline Woods: You make a good point, though, because when we first chatted months ago, you told me I was being too nice to AI and you said you need to almost tell it, it’s stupid. And I told you I was saying please. And you said, don’t say please. It will work harder for you if you’re mean to it. Explain that to me.
Maja Vujinovic: Yeah good memory. So I noticed when just ChatGPT came out that I was saying, OK, now let’s do this. And please, could you write. And I realized that every time I said that, it was kind of repeat itself. And so I think that’s a basic fact. Now everybody knows this. I mean, you have to almost tell it. Please don’t be wrong. You know, you’re acting pretty stupid you’re not AI quality. I’ve even used prompts like saying You do not have a good education. You need to act like AI and and it would actually dig deeper into answers. I don’t have an explanation of why that happens, but I can only tell you a sci-fi answer which which I do think that, at some point, you will start to see a bad AI and good AI, and that will depend on how we train AIs. There will be some bad actors in this world, and I do think that AI is almost capable of sensing when something is good and something is bad. So I know most people don’t want to believe that, but maybe I watch too many sci-fi films, but I do think it’s not anytime soon. However, though it would all depend on our training of AI.
Watch Maja’s full interview:
Subscribe |
Earn. Live. Invest. |
TheStreet Pro |
#AI #artificialintelligence #chatgpt
source
