The technology is impressive, but the largest part of the improvement over earlier systems is scale - huge advances in computing. As an example: 40 years ago I wrote a little code that extracted likely sequences of words and pasted them together, and used it at a party. Everyone who came in was asked to type in a few sentences of a story…
The technology is impressive, but the largest part of the improvement over earlier systems is scale - huge advances in computing. As an example: 40 years ago I wrote a little code that extracted likely sequences of words and pasted them together, and used it at a party. Everyone who came in was asked to type in a few sentences of a story (without looking at what others had written). I pasted all the contributions together and ran my program, and generated the "party collective story." Of course it was rough, but it succeeded at being funny. It essentially was a generative large language model. Now, of course, groups like OpenAI can tap enormous banks of written (or even transcribed spoken) language, they have the computing power to do it, and the artificial neural networks involved can be huge given the computational resources now available.
The point of my saying this is that sometimes when technological capabilities reach a tipping point, it seems like there has been a sudden leap, when they have been gradually increasing all along. But what's real in that sense of a sudden leap is the applicability of the technology to anything people care about. Video conversations were demonstrated at the World's Fair in 1939, but it took smart phones this century to make it a thing.
Aside from all that, another fun Arizona Agenda newsletter! Thanks!
The technology is impressive, but the largest part of the improvement over earlier systems is scale - huge advances in computing. As an example: 40 years ago I wrote a little code that extracted likely sequences of words and pasted them together, and used it at a party. Everyone who came in was asked to type in a few sentences of a story (without looking at what others had written). I pasted all the contributions together and ran my program, and generated the "party collective story." Of course it was rough, but it succeeded at being funny. It essentially was a generative large language model. Now, of course, groups like OpenAI can tap enormous banks of written (or even transcribed spoken) language, they have the computing power to do it, and the artificial neural networks involved can be huge given the computational resources now available.
The point of my saying this is that sometimes when technological capabilities reach a tipping point, it seems like there has been a sudden leap, when they have been gradually increasing all along. But what's real in that sense of a sudden leap is the applicability of the technology to anything people care about. Video conversations were demonstrated at the World's Fair in 1939, but it took smart phones this century to make it a thing.
Aside from all that, another fun Arizona Agenda newsletter! Thanks!