Sublime

Kahle: Yes. One thing I learned from Marvin Minsky (one of the founders of AI) was, "Pick a big enough project, something that's really hard, something that over the years you can work on." I've found that that has been a great guiding piece of wisdom. If you just set out to go and make a lot of money, then the problem is, what happens wh

... See more

Some tasks are easier to improve via system design. While LLMs appear to follow remarkable scaling laws that predictably yield better results with more compute, in many applications, scaling offers lower returns-vs-cost than building a compound system. For example, suppose that the current best LLM can solve coding contest problems 30% of the time

... See more

Incrementally improving an explanation of a technical idea might take a single author weeks to do, but could go on to save a day for thousands, tens of thousands, or hundreds of thousands of students, if it becomes the best option available.

This is a challenging task even for today's state-of-the-art LLMs such as GPT-4, largely due to their inability to generate accurate input arguments and their tendency to hallucinate the wrong usage of an API call

Whether or not we come to understand “intelligence” any time soon, we do have a major challenge on our hands in bringing together computers and humans in ways that enhance human life. While this challenge is viewed by some as subservient to the creation of “artificial intelligence,” it can also be viewed more prosaically — but with no less reverenc... See more

In most areas of computing, programmers write code or train an AI model to achieve specific objectives, such as driving a car without crashing or generating humanlike text. Stanley creates systems that instead evolve, seeking novelty for its own sake. These systems sometimes discover stepping stones toward solutions that couldn’t be achieved via a ... See more