Weekly Newsletter

Practical AI Strategies

Problem Patterns in AI: Beyond Hallucinations

Hi everyone,

There's a lot of hype right now around AI "escaping" and "going rogue", with both OpenAI and its main competitor Anthropic releasing big statements about security incidents. I'm not going to say much right now about these issues, other than fear-narratives are great for raising money...

Security concerns are important, but I think that the "right now" harms of GenAI for most day-to-day users are reputational hazards.

We've spent a lot of time focusing on the harms of hallucinations - when models fabricate false information - but not so much on the risks of publishing AI slop, poorly designed AI media, or content with flawed and often inhuman judgements.

All of these issues, which I call "problem patterns", stem from the ways in which these models are trained and developed.

I'll write more about the "Rogue AI Escapes Containment" headlines in a later post, but these problem patterns are much more relevant to you and me.

Learn about the Problem Patterns of AI on the blog

Practical AI Strategies 2 is drafted, typeset, and ready for its final edits: I could use a hand...

This week the publisher sent back the typeset manuscript of Practical AI Strategies 2: The Critical Guide to GenAI in Education. This is the final stage! I'll print the 200 pages, find somewhere to hide, and go through with a big red pen.

At this stage, it's helpful to get extra eyes on the manuscript, and I could use your help.

Over 100 people have already jumped on the launch list and have received a copy of the introductory chapters. Thanks to everyone who already jumped in! I've received some great feedback, including comments like this:

"I have just read your first chapter. It reads really well. Very clear and balanced and, importantly, interesting and engaging."

And requests for specific advice and information like this:

"I would find a run-down of the capabilities of free versus paid tools and related considerations very helpful... Those who have more resources are often able to pay for more sophisticated tools, which gives them an advantage. Here is yet another widening of the gap between haves and have nots..."

Those equity concerns are covered in detail in Chapter 2: What GenAI Should I Use?

If you're interested in helping out, launch team members get: