FeedCity

In the Grok fallout, Mike Masnick makes the case for moving away from large, centralized AI. He asks a range of great questions about bias in AI: After a similar incident two months or so ago where Grok became obsessed with linking everything to white genocide, the company ...

The biggest problem with ChatGPT is that it thinks it's running the show. I've just given it instructions to think of itself as a command line that can understand English. Just answer the question exactly as asked.

Session in the cowshed

Next time there's a Big Beautiful Bill, let's set up a ChatGPT project or equivalent to injest new versions of the bill as they come out, and quickly alert us to issues, and also suggest ways to frame it for the electorate and the press. We have new analytical tools, we should use them. We're only now, far after it's too late, finding out the awful things that the BBB going to do to us. The Repubs planned this out far in advance. They probably even had software tools to advise them on ways to word the bill as the changes were negotiated.

Friday morning session in Miltown

This is the 98th edition of People and Blogs, the series where I ask interesting people to talk about themselves and their blogs. Today we have but she's a girl... and her blog, rousette.org.uk To follow this series subscribe to the newsletter. A new interview will land in y...

Swarm changed their app icon from orange to gray (in a generic “bug fixes” update with no release notes) so I had to move it on my home screen. It used to be in between Overcast and Audible. Bottom dock hasn’t changed in a while: Hey, Epilogue, Strata, Micro.blog.

I took a long drive today on back roads, down toward Poughkeepsie, and on the way back a big pickup swerved into my lane. I had to swerve away from it and hit the brake, and had a bit of a conniption. Got me thinking. Who was it that did this. Man or woman, young or old, spac...

Rewriting a bunch of my old Micropub code for the web today. Long overdue, and needed to keep adding features.

Dia is coming along nicely. They’ve fixed the custom text selection that was bugging me. Split view is interesting too. Overall, the tabs and UI look cleaner than most web browsers.

I’m good with the explosion of new browsers. Orion, Comet, supposedly an OpenAI browser soon.

When did I feel the change from the countryside to the town? This question came to mind while I was walking earlier this evening. While walking, I noticed that there was a point at which I felt I was coming back from the countryside into town, but I didn’t know when that mome...

Leo Laporte, linking to a webcomic about RSS:

…and here’s the good news, Spotify wasn’t able to co-opt podcasting, RSS podcasts still live and thrive. And I, for one, intend to keep it that way.

King of the Hill is returning (YouTube trailer), and the world has changed a little. “I don’t know how to kick someone’s ass over Zoom, but I’ll figure it out.” 🙂

Have they made ChatGPT more stupid in the last couple of days? Or maybe I'm starting to see the problem. I was trying to debug a piece of software that sends mail. I use Amazon SES. I wanted to see if the software was actually communicating with the server, and thought for sure there must be a log of requests on the AWS system. Where is it? I asked ChatGPT. It starts telling me to check all these things and never got to the part about where the log is. So I said, just tell me where the log is. Again, it tells me about all these nice things I should check first. I asked again, this time in ALL CAPS which is how I communicate that I really mean it. Again more bullshit. So I think to ask Is there a log? No, it says, actually there isn't one.

How to transform the internet's most toxic platform into essential infrastructure.

I want to write something! This thought has come to mind a lot lately. I have the energy to go and start writing. But before I do, I ask myself: what do I want to write? Inspiration is hard to find, although I recently realised that the voice saying that I want to write somet...

Seeing random Grok screenshots — not just the racist tweets, all the other craziness too — makes me appreciate how much that serious companies like OpenAI and Anthropic must actually work on alignment and guardrails. Grok seems imbued with a “personality” that is out of control.

Useful report from the Social Web Foundation after a privacy forum in Norway. From the first bullet point of next actions:

Create standard metadata fields within ActivityPub to indicate content visibility, consent preferences, and sharing restrictions across servers.

There is a tension in the fediverse between consent and the open web. It’s difficult to balance both ideals without over-complicating everything. Look at how confusing private mentions in Mastodon are.