Back when blockchain first reared its head, I had a pretty close look at it, to work out if there was a there, there. I came to the conclusion, along with a lot of other folks, that there wasn't a there, there in that case - outside of a few niche cases^[1].

For a while, AI fell in the same bucket - a LOT of money and hype being thrown at something which just wasn't that obviously good, and in some cases, quite destructive to some industries.

But after only a few iterations - remember, ChatGPT went from 3, to 3.5, to 4 in less than 24 months (November 2022) - there does appear to be a there, there. Or at least, the nascent, embryonic start of something.

Is it perfect? Hell no.

Is it moving at an insane pace? Absolutely yes.

Can you do things which would seem like magic to anyone 15 years ago? Totally.

Is it overhyped to the point of stupidity? Well, yes and also maybe no.

Side note: do I think we are heading for AGI anytime soon? No. Or maybe. I don't know.

I do think that AGI - or making some kind of simulation of your brain and sending it out - is more likely than an actual human going to Mars to live. Outside of a "yeah, it was always a one way ticket" suicide ride, anyway. This is worth a read.

Overall, I'm not sure I'm as "all in" as JD is, yet, but I'm possibly not too far behind:

So, being I've not been writing much recently, and that's a muscle I want to exercise more, I'm going to document a bit what I've been doing with AI/ML stuff, what I've found it good for. That kind of thing.

And no, I don't plan on writing these with AI. Thats my job. But I do plan to look at the things I've been using it for, both personally and at work; challenges; where I think it might go; principles I like when I'm designing systems which use it, and, well, anything else I can think of.

Before you ask, I consider AI, ML, LLMs, GenAI etc to all be in the same space. Are they 100% the same? No, it's a wide space. Close enough for this. I'm going to use "AI" as the term, even tho "advanced predictive search engine" could also be used. I guess.

The landscape

Some of the tools I've tried, and I'll be writing about, include:

• Github Copilot
• ChatGPT - 3, 3.5, 4
• Claude (3, 3.5)
• Gemini - Googles chat and LLM model
• Ollama - think docker, for llms, on your M-series Mac.

As well as the various thing you can do with APIs like ChatGPT, AWS Bedrock and a few others. This is not even remotely an exhaustive list, and if you are reading this more than 100ms after I write it, its likely to be out of date. Them's the breaks in fast moving technology.

I've had mixed results, so far, with but the indicators are pointing the right direction. This is not even v1 stuff, and it's moving at a ridiculous pace.

I also don't profess to being an expert. I've been around in software engineering for a long time, and I think I can spot bullshit when I see it. But maybe I can't. I'll be fun to find out.

Lets go then.

1. Bitcoin, for all the hype, and Etherium as a platform might be one, more thru social acceptance than anything. Maybe. Git is, if you look at the underlying structures, a chain of blocks. Is it going to replace land title databases? yeah nah.
2. Things which are not AI: Siri. Alexa.