Notices by trisweb (trisweb@m.trisweb.com), page 2

Corollary: the mission of product development (note: not a specific job title) is to understand your customer’s job better than they do, so that you might create a tool to guide them to do it better.

If you’re making a tool based on a poor understanding of their job, aren’t you simply targeting poor execution of that job? Who would want that? Who would want to take part in building that?

@trochee oddly enough this thought came from a discussion about Figma’s new AI features.

I don’t think they’re playing 3D chess with the point of these features, they’re just following the hype train and trying to be reasonable about how it might help people, with a lot of uncertainty in how that happens.

My conclusion here is, if they understood the jobs their customers actually do (in Figma’s case the discussion was around getting out of high fidelity and actually modeling solutions and conceptual directions) then they could design tools, AI or otherwise, that are genuinely valuable. Like, I don’t know, something that helps designers work through a fantastic approach to their process in a humanistic and user-centric way, rather than just helps plagiarize high-fidelity UIs for useless conceptual ideas that have no connection to reality.

I don’t believe that their customers are demanding “replace humans with AI” or expect that to be realistic (yet).

You can’t make up just how badly this shit works.

@mike @andypiper @tchambers @shitpostalotl yep this is mostly correct via my experience on a single-user server. Following matters the most, but others’ boosts do cause those specific users and messages to be pulled in, so a nice side effect when looking at a user with no follow relationship is I can see the posts that my server got by whatever means, which means there was some activity I can go on.

Overall, fine-grained controls on federation, and especially controlling moderate proactive fetching of posts and content, is IMO one of the best opportunities to improve Mastodon.

I'm scared how misunderstood LLMs and GenAI truly are; they are so far from a genuine intelligence it's not even funny. They are statistical models for language, trained on a lot of words and sentences, representing the statistical likelihood of words following other words, sentences following other sentences, just with a lot of data and samples to pull from. That's it. There's no intelligence, there's no path to intelligence; it's simply statistics chaining words and symbols together.

@futurebird we need a knowledge bank, kind of like a seed bank but for knowledge untainted by AI.

@clacke @engarneering yeah that makes sense, I figured I have no recourse it’s just upsetting to lose control over something you’ve written in general.

Follow up: it is not within my rights to remove my content, apparently. Upon submitting it conveniently becomes Creative Commons licensed. Huh.

@j12i that’s their recommended approach, and I’ve done it already. Or you can delete your account which makes everything just reference an anonymous contributor. I believe that’s how they say they adhere to GDPR and California privacy act.

Byeeeee stackoverflow

lol, going to need to try a different method

This is what you get when you try to delete your content from #StackOverflow: they treat you like a criminal defacing their site, rather than as a human being with privacy rights.

Will be saving this conversation and sharing publicly. Your move, Stack Overflow…

@carnage4life Apple simply wasn’t the one to undertake it.

Competing with Tesla will be easy for auto makers with actual experience manufacturing automobiles. Tesla isn’t even good at that.

@thomasfuchs just a heads up, mostly the issue with not reading replies is because all replies aren’t pulled into my local server, not that they’re displayed wrong.

@marcprecipice @drewda I still miss it. I think I almost interned for you once? A missed opportunity for sure.

@virtulis @aral that would be fine. Edit: it may not be fine, as it still makes it inaccessible to folks hosted by a provider.

@loke @aral @shoq Par for the course for him. He's a pseudointellectual blowhard who's entirely divorced from the real world.

@thomasfuchs @mudasobwa @JonnyT @mac I think it's important to recognize that in-person does work as well for many people and situations. That doesn't mean that remote work doesn't, it's just a nuanced decision that should be left in the hands of the people working on the team.

We shouldn't blanket assume that working from home doesn't work for any given role or team, just like we shouldn't say that working in person doesn't work for any given role or team.

I think the major danger of #LLM #AI is not their fundamental capabilities; they're truly useful and incredible in how they can aid in language comprehension and generation. That's their job!

The danger is that we insist on giving them a human interface of conversation, the semblance of being a human intelligence. Because they can do it so well, they're enchanting---and deceptive.

We can use them for what they're good at; but we should stop fooling ourselves into thinking they are intelligent.

men will create a fully generic entirely customizable meta-software system that enables customers to enter their own custom data model, which then is built into a bespoke application just for them including special relationship types that turn on unique UIs and abilities, instead of talking to a customer one time