Apparently “I let ChatGPT control my life for X days/hours” has become a genre of Youtube videos. 🤨
I decided I didn’t like the original title for this week’s newsletter entry, so I changed it 🙂
“Prompts are unsafe, and that means language models are not fit for purpose”
In addition to being potentially vulnerable to black-hat keyword manipulation and plundering the commons, these systems are a big security hazard as designed
“Prompts—and with them language models—are not fit for purpose”
Watch for bait-and-switch. Intentionally or not, some leader-level positions with purportedly large teams turn out to be individual contributor positions
“Thoughts for In-The-Job-Market Product Leaders”
😅 I’ve fallen for this a couple of times 😬
“The Hawthorne Effect or Observer Bias in User Research”
Rather than mitigating it, IME most rely on this effect in order to get the predetermined results they want.
They should not call themselves “I” and they should not refer to themselves and humans as “we.”
“Ban LLMs Using First-Person Pronouns — Crooked Timber”
This is a start. Doesn’t go far enough, but it’s a start.
“The Senate’s hearing on AI regulation was dangerously friendly”
A number of experts and industry figures say the hearing suggests we may be headed into an era of industry capture in AI.
An important thing to note is that whenever an AI vendor says that “we don’t understand how these things do what they do” that isn’t because the systems are inscrutable and mysterious, it’s because they aren’t documenting their shit.
“Why Timnit Gebru Isn’t Waiting for Big Tech to Fix AI’s Problems“
Is there another micro.blog client that has a usable sharing widget? Because if I can’t find a decent replacement I may have to figure something else out for my notes blog.
Well, this sucks. Micro.blog’s new iOS client is a complete downgrade on the iPad. Scaled up phone UI and the sharing widget now only shares the url of the page, not title and selected quote like it used to.
That thing when romance author Twitter shows a deeper and more technical understanding of language models, how they work, and what they do than most of tech.
So is there a greater bullshit phrase than “latent spaces” as used by AI fans?
Worrying in public seems to be the theme of the day for me. Just published “Writing when tech has broken the web’s social contract”
“Drag Your GAN: Interactive Point-based Manipulation on the Generative Image Manifold”
This is interesting: 1. seems to work with smaller training data sets, 2. faster than diffusion models, 3. an actual user interface.
Every paying outlet—even many that only ‘pay’ in attention—is going to be hit hard.
“The Fanfic Sex Trope That Caught a Plundering AI Red-Handed”
If, like me, you get the login wall, apparently the reason why the AI writing tool Sudowrite gets the results it gets is because it trains on AO3
“Debt Collectors Want To Use AI Chatbots To Hustle People For Money”
So, this sort of stuff is exactly the reason why the “high-risk” category exists and is more heavily regulated in the proposed EU AI Act.
Just been collecting some of what people have said about my two books and I’m honestly touched.
“Praise for my ebooks ‘The Intelligence Illusion’ and ‘Out of the Software Crisis’”