Yes, the title of this Dan’s New Leaf post is indeed another play onwards. Just like the last word of the last sentence . . .

. . . OK enough of that illustration. Illustration, you ask?

Yes. Illustration. I mean, it is merely wordplay – all part of my normal cyberpoetry – but it’s also meant to be an illustration.

I’m demonstrating what ChatGPT does, in a cyberpoetic way. Predicting the next word, in response to a set of words it has an analyzed, the prompt, LLMs are actually a mathematical form of wordplay.

It’s just that training neural networks to handle the wordplay provides such powerful benefits that it has not only intrigued us; it has scared us. Enough to start talking about new laws.

One of my favorite cyberpoetry rhymes – enlighten and frighten, is really at play when it comes to ChatGPT and its wordplay. We are so enlightened, we’re now frightened.

Speaking of enlightening wordplay, in cyberpoetic metaphor – our boilerplates are really an extension of the wordplay we enjoyed as Mad magazine juveniles. and speaking of boilerplates . . .

We are finally finished with our AI policy boilerplate. I know, it seems to have taken forever. But given you can use ChatGPT to generate policies now, we felt like we needed to take our boilerplates to the next level. We hope you can see that in this policy template and we encourage you to provide feedback.

But before you run out and adopt an AI Policy over the holiday – which I wouldn’t recommend anyway – please know that we already want to update it. We’ve already discovered something going on that has given us a new lens on the concept of governing artificial intelligence.

Our AI policy boilerplate takes a risk-based approach. This matches, in intent and approach, to the new EU AI law. But as we started investigating the regulations proposed for that law, we noticed the EU has defined risk categories differently than we have. Therefore, we’re thinking we should at least include the EU risk categories as an option to ours. And we’re also looking at a possible hybrid.

One other thing we want to learn over the holidays is whether or not the EU AI regulations call for unintended consequences. Or are they going to avoid micromanaging tactics. Will they repeat prescriptive bad ideas, like the GDPR speed bumps we all hit now, every time we go to a website? We didn’t always have to take the risk of clicking on a pop-up window telling the website they can, in essence, use cookies.

That wasn’t intelligent.

The I in AI stands for intelligence. Our policies need to be better than the wordplay generated by a large language model.

And to mitigate AI risk, we have one amazingly powerful, inexpensive, and distinguishing tool in our utility belt. (Another play on words, Batman.)

While those enlightened to the risks of artificial intelligence can be in the dark as to how it works, we must protect ourselves from the frightening risk of AI, with the bright light of intelligence, and that one feature of intelligence that separates the organic from the artificial – awareness.