ChatGPT 4o, handwriting, fake science and a little optimism

confusing and stage-y demo with no testing, research, or safety documentation

The big news on 4o (which isn’t the long awaited GPT5) demonstrated looking and emotional speaking capabilities. I call out the concerns and troubles with that highly theatrical and curated demo here.

I will add one more concern for my readers here exclusively: your ability to stop it from saving your chats is going away. There is no word on what it will do with images of your face and sound files of your voice. Recall that, as we reported here, Sam Altman has been running around the globe with Orb trying to get people to scan their retinas. Also recall that even infomercials require “user results may vary” for their commercials. Wait until launch on Monday. I predict hilarity will ensue. And maybe a little identity theft thrown in.

handwriting is necessary for deep learning and transcendent thinking

At a moment when Open AI is promising not even type but completely writing free tech that does writing for you, research is showing the significant impact handwriting has on learning and development. As an ardent longhand writer I hope we don’t move backwards in our already declining student performance.

fake generated science causes 19 academic journal shutdowns

The Wall Street Journal reports that academic journals are being flooded with fake science. Is it time to return to handwritten manuscripts?

a deeper look inside the tech

We feel an obligation to be a no hype zone for you but we remain technoptimists regarding the potential of cross-functional data management. The dangers of recklessness in an ethics-free zone are significant but we believe it will get better. In the meantime we do our part to increase understanding. Take a look at this easy to understand breakdown of Kolmogorov-Arnold networks. What’s the upshot? Previous tech used an approximation theorem but it is very unweildy and xan only perform in certain conditions. These networks (KANs) use less parameters and are far more efficient. This tech might have dangerous front men creating confusing hype that will surely turn consumer sentiment against them, eventually. In the meantime curious people who care are thinking about what cross-functional data management means from humanity-centered perspectives. What would be terrible is if these infomercial style demos dumb down the tech so much that people turn away from it all together. The prize lies in allowing more competition and more people in to play in the arena.

This Friday at 11AM we are going live with Charles Mauro on Linked In. Follow my personal feed to hear a live feed of us talking about human factors engineering, IP, and the human cost of the technologies on the horizon.