I’ve been meaning to write this for a couple of weeks, so let’s get to it - things are moving to fast to reflect too long; which is it’s own risk.
In March, I wrote about how I was using AI in coding , which was Codeium (now Windsurf) in VS Code for completions, and ChatGPT and Claude online for architecture questions and code gen that was more than half a function.
There’s still plenty of controversy about LLMs for coding, and not without reason. But I thought I’d run through what I’ve tried, and where I’ve landed for using AI. Also what the pitfalls are, where it’s useful and how it’s changed my practice.
The training data for large language models generally is problematic. There’s no doubt that they have been trained on copyright material. With code it’s slightly less murky since there is a high availability of good quality open source data with attached licenses to train models on. No doubt this include code written by people who don’t approve of it being used by AI, but I think the popular reading of most open source licenses is that using it for training is fine.
If you’re interested in how generative AI works, check out Ishan Anand ’s Youtube series “Spreadsheets are all you need ”. He steps through the basics using an Excel spreadsheet that encompasses most of GPT-2. Just doing that is an impressive (and hilarious) feat, but he also has a knack for teaching, so you’ll come away with a good understanding of AI and how some of it’s limitations manifest.
Ishan is selling a course, which I guess these are the first three lessons of, and I got a lot out of them. It’s also possible to download the spreadsheet he uses in the course to play with.
Now Ollama has made it simple enough for anyone who can use a terminal to run large language models locally, naturally I’ve gone overboard downloading too many to play with. I’m increasingly feeling they definitely have a place in the devops/coding arsenal of tools, but which model is best?
If you go on HuggingFace to look at a new model you’re interested, they often have great comparisons like this.
This post looks at the context for some of my thinking about AI for supporting software development, and where I’ve landed on it for the time being.
I briefly wrote about ChatGPT’s coding ability at the end of 2022. The wide availability of this tool marked the beginning of what I think can fairly be described as a revolution. The controversies that have crystalised since have not dampened my amazement of this step forward in what compute can do, especially around natural language processing.
This week, the internet has been all about ChatGPT , the rather remarkable natural language AI with a very large model. If you’re a twitter user, you were probably amazed, but now eventually tired of seeing examples of it’s output. I’ll add to that with an example of a SwifUI CoreData based todo app it wrote for me from a single sentence prompt below. Rather than look at other people’s examples you should definitely go and play with it yourself - it is very impressive. Along with the image based AI’s it’s made 2022 into a historical year for AI.
Since I have minimal design skills, I went back to Fiverr (the digital gig economy platform) to get some icons done for CodeTrimmer - explaining that I wanted something like a “pair of scissors floating over some computer code”. At the same time I’ve been playing with DiffusionBee - a free Apple silicon version of the Stable Diffusion artifical intellligence that generates images from text prompts. The image above was created on an M1 Macbook using DiffusionBee.