I’ve been watching some of the discussions online and in the media with concern. Quite a few of the statements people make are… not entirely grounded in reality.
This isn’t unusual for discussions on new technology.
It’s a bit like when foreigners talk about Iceland. They frequently paint it as a near-Utopia and us Icelanders as these amazing, well-educated people.
As a general rule, when somebody is busy giving you free PR, you don’t hurry to correct them.
The same relationship exists between the tech industry and the misguided punditry overhyping its innovations.
Our burgeoning AI bubble is shaping up to be a little bit more unhinged than other recent tech industry shenanigans. Quite a few of the people most active at promoting AI are either dedicated grifters intent on draining every last cent out of the bubble they can, or are true believers who just have faith in “the exponential” and that if you wait five minutes, it’ll all be true anyway.
People are going to have to develop a nose for AI bullshit.
Earlier this week, as I was shaking my head over yet another pundit making false claims, I realised that this, specifically, is something I can help with.
While I can’t give my book away for free (do feel free to buy it, though, people seem to like it) one thing I did for the book was to put together a set of information cards. The idea was to have a scannable, straightforward resource to help readers get a quick overview of what’s what in generative AI before they dive into the book.
That got me thinking:
Huh? It wouldn’t be that much work to rework those into a standalone mini-site that people could use to get a better sense of generative AI’s true capabilities and limitations.
A couple of days later and needtoknow.fyi is live.
Generative AI: What You Need To Know is a collection of fifteen cards, covering thirteen core topics in generative AI.
- A high-level overview of how language models and diffusion models work.
- What they’re good and bad for.
- Short and scannable overviews of each major risk.
- An overview of the AI industry’s past history with false promises.
- How these models “reason”.
- And more.
It’s all written with a non-technical audience in mind and each card comes with a list for further reading, pointers to where to find out more in my book, as well as a list of references that back the statements on the card.
The main site gives you a background on the motivation behind the mini-site as well as the research it’s based on, but you can just go directly to the page with all the cards:
I hope people find it helpful in cutting through the nonsense and that it makes it easy for you to spot AI grifters and bullshit artists.