I had a visiting professor in my class this week. He gave a lecture on AI and how to best determine possible use cases and then how to prioritize them. Great discussion as he provided a structure: a way to classify different types of use cases and to identify the type of AI required to solve the problem. When dealing with a fast moving and quickly changing environment, a little bit of structure can go a long way!

OK, I had to laugh… I was watching a Youtube video about reading ledger lines in music and an ad appeared for an AI course (why it appeared is a separate issue!) The premise of the ad was that all users over 40 use AI incorrectly and therefore needed to take the course to figure it all out. Spoiler alert: I am over 40. My experience may be unusual (Along with The Wired Schoolhouse I'm also a professor) but the good AI users I know are ALL OVER 40… it's mostly younger users who are doing the one prompt and then cut and paste thing. Honestly… who cares about age; let's just use tools as best we can so we move the planet forward.

I would call the creator of the ad to complain but it's likely past their bed time.

We're mostly familiar with large language models (LLMs) like chatGPT, Copilot, etc. But increasingly I'm seeing small language models - these are smaller and lighter AIs. They have specific knowledge in a subject area but not a broad knowledge of the world at large. The trade off is general knowledge and deeper reasoning for speed and lower resource needs. Key issue is that they're specialized and more easily "tuned" for specific needs, i.e. project management. AND you can run them on a laptop! Worth thinking about - use the small model for lap top based specific tools such as meeting data, summarizing specialized documents or decision support while continuing to use large models for more complex or abstract problems.

Interesting article in the paper today - Quebec City is using Google AI technology to improve traffic flow. Using data from Google Maps and a specific purpose AI, traffic light timing is adjusted to optimize traffic flow resulting in lower drive times and thereby reducing carbon dioxide emissions. 19 cities around the world are using this technology now. The article doesn't address the issue of real time adjustments but why couldn't that happen? I suspect this is how many AI implementations will happen - rather quietly and under the covers but making real differences in the real world. Hopefully positive differences of course but that's not quite clear yet…

ChatGPT 5 is out and while the reviews are mixed (at best) it's clear that progress, for all the big AIs, is ongoing. After all, they're not burning the money and there's a lot of money… So? So, we're going to pivot a bit. Why create content that an AI can find for you? Sure we could likely sell the product for awhile but I think we need to add more value… not just about using an AI as a tool but more as a partner… a fast, pretty smart, but still makes mistakes partner.