Today’s LLMs and Chatbots: A Maker’s Vision

Today’s LLMs and Chatbots: A Maker’s Vision - https://www.youtube.com/watch?v=PvW3ePVCJwc

OpenAI has been around for seven years now, so it's not, it's not a very old organization, but it's also not a very new organization either. The culture of OpenAI has changed over time as we got more and more clarity about how to make progress. In the first year, our culture was a little bit more exploratory. We were trying to understand what it is that we want to do and hone in on the big ideas that we wanted to pursue. And then as time went on, our culture started to adopt more aspects of large scale engineering projects. Because our goal is not just to invent something, our goal is also to build it. And building requires a really strong engineering culture where large teams are working together to produce sophisticated engineering artifacts. I think all of computing has just one goal right now, which is to actually build artificial intelligence which loves humanity or some variant thereof. All the AI that exists today is based on top of deep learning, which is a relatively narrow set of ideas compared to the broader computer science. So in the past five years or seven years or so, there's been, or maybe more 10 years, there's been this contraction where previously there's been lots of activity all over in computer science research and now it's all concentrated in AI on deep learning specifically. So people go really deep and people understand those techniques. Deep learning, which is basically ways of building digital brains which live inside large computers. That's what deep learning is. Literally we build digital brains with artificial neurons inside big computers and there's a certain set of ideas that come with it. There are certain ways of thinking which are more forward looking. And when people work together and they internalize these forward looking ideas, very substantial and productive progress is possible. What I hope will happen is that as we keep on, as the field keeps on making progress and artificial intelligence gets smarter and smarter, it is eventually built and it's used to solve a huge number of problems that society has a lot of problems that have been intractable for a long time around. From small things like global warming, to international relationships, to massive scale mega cheap autonomous healthcare, to curing lots of diseases and extending life and living on other planets. I think all of this could happen. It'd be rather nice. In a nutshell, AI is going to be incredibly, unbelievably powerful, incredibly unbelievably impactful. And with such tremendous power comes the possibility that perhaps things won't go quite the way we want. Will we be able to manage such powerful tools? And also the technical challenges of making those technical tools well behaved. I think These are all very significant risks, and it is my wish that basically none of this happens. Here's what a predict will happen. I think AI will continue to make progress. I think we will see AI becoming the single most discussed subject in society. I think we will see a whole lot of people starting to think about the questions that are posed by AIs, and starting to come up with creative and unexpected solutions that we haven't thought of yet. I think that we will see a huge number of incredible applications and incredible uses of AI, and we'll see lots of existing problems, indeed starting to be solved. We'll see a huge amount of economic growth, and we'll also see new problems that are caused by AI that we've never seen before. I think all of these things will happen in the same time. The precise proportions are unpredictable. But yeah, it will be not boring.