A conversation with Mira Murati, CTO of OpenAI (full interview)

AI everywhere: transforming our world and empowering humanity

Mira Murati Th’12, chief technology officer of OpenAI. This exclusive discussion will explore the pioneering advances and ethical considerations of AI. Murati will share his ideas, experiences and groundbreaking developments to better understand the role of AI in shaping our collective future.

Moderator: Dartmouth Trustee Jeffrey Blackburn '91

Mira Murati shared some eye-opening insights, including her blunt statement:

/"I think so, and maybe some creative jobs will disappear, but maybe they shouldn't have been there in the first place if the content that comes out isn't very high quality, but I truly believe that using it as an educational tool, creativity will develop our intelligence, creativity and imagination./"
/"And also, I think people don't realize how much of these tools are already used, and it's not studied at all. And so we should study what's happening right now with the nature of work, the nature of education, and that's going to help us predict how to prepare for these increased capabilities in terms of jobs in particular, I'm not an economist, but I certainly expect a lot of. jobs change, some jobs are lost, some jobs.
/"- So yes, these systems are already at human level in specific tasks, and of course in a lot of tasks they are not. If you look at the improvement trajectory, systems like GPT-3, we are maybe let's say, toddler level intelligence And then systems like GPT-4 are more like high school student intelligence And then, in the next couple of years, we'll look at intelligence at. doctoral level for specific tasks./"
/"So we found that this formula works really well, data, compute and deep learning, and you can put different types of data, you can increase the amount of compute, and then the performance of these systems." AI is getting better and better. And these are what we call scaling laws. They are not real laws. They are essentially a statistical prediction of how well the model will perform. improve as you put more data and more calculations into it. AI advances today./"
/"Well, yeah, exactly. And so there's all this debate right now, are you doing more security or are you doing more capability research? And I think that's a little misguided because well Of course, you have to think about security in the deployment of a product and the guardrails around it. But in terms of research and development, they actually go hand in hand. And from our point of view, our approach. is very scientific. So let's try to predict what capabilities these models will be, what capabilities they will have before we actually complete the training. And then, along the way, let's prepare the guardrails for how we do them. let's manage. That hasn't really been the case in the industry so far, and then there are these emerging capabilities that we call, because they're emerging. We can see some sort of statistical performance, but we don't know if statistical performance means that. the model is better in translation, biochemistry, coding or whatever. And developing this new science of capacity forecasting helps us prepare for what lies ahead. And that means…/"
/"I really think we need to understand how we use these tools and AI to advance education. Because I think one of the most powerful applications of AI will be in education, in advancing our creativity and our knowledge and we have the opportunity to essentially build a very high quality, very accessible and ideally free education for anyone in the world, in every language or cultural nuance that you can imagine. You can really have a personalized understanding and education for anyone in the world And of course at institutions like Dartmouth, the classrooms are smaller and you have a lot of attention, but you can still imagine. have one-on-one tutoring, even here, let alone in the rest of the world – As a supplement – Yes. This happens very late, perhaps in college, and that's one thing. so fundamental, the way you learn, otherwise you can waste a lot of time and the courses, the program, the problems posed. , everything can be personalized based on how you actually learn as an individual./"