hello

Foundation models have passed a tipping point and they are cropping up everywhere, in a huge variety of use cases. We can transcribe images, translate text, do object detection, and generate anything that our hearts desire. The simplest approach to solving a problem with machine learning now almost invariably begins with “send it to Claude, OpenAI, or both and see what happens” before digging into specific model implementation details. Many natural language processing (NLP) tasks have been “solved” with this, at least in a very first pass of the data. Do you want to extract topics from a text? Do you want to summarize it? Do you want to judge whether it’s offensive or not? Send it to an LLM and log the output. Maybe later you use the output to train something specialized, but that first pass is very effective for standing up a proof of concept. ...

I have done a number of talks and appeared in a number of videos over the years. It has been some time since many of these videos and talks have been made (mostly over 5 years from this date), but there are some interesting nuggets amongst them. Most of these are on applications of machine learning in finance - you may find them interesting. Talks Thrifting Alpha Thrifting alpha is a talk about using ensemble learning to combine low alpha (weakly predictive) signals for algorithmic trading into stronger composite signals for trading a portfolio of stocks. ...

Artificial intelligence (AI) products are now ubiquitous, but I think that few of them are able to hit on the right interface. Chat became the dominant functioning path too early, and now it is stapled onto everything, regardless of whether it is suitable for the medium of the information it is trying to channel. When it comes to knowledge management, I personally do not like heavy usage of generative AI tools. For me the writing and the iterating is what makes the process worth doing, and from a variety of works, we know that the act of writing contributes heavily to cementing new ideas in your mind. ...