Clojurians Log v2

You mean like how google once found out the gold was in the search queries? (From https://www.amazon.com/Age-Surveillance-Capitalism-Future-Frontier/dp/1610395697)

Yes, and based on what normal people ask from LLM they could completely understand what that person ideology is & ...

I guess whatever we could see in the Snowden movie is just cute compared to what some organisations can do now 😕

My wife & I both joke that our search history (and now my Copilot history) is so weird and random, that surveillance folks would probably go permanently insane trying to understand it. We're both incredibly curious and look up every random thing that pops up in conversation...!

(those are Copilot-generated titles for conversations we've had, not always the actual questions I asked!)

Haha, I have a similarly weird ChatGPT history 🙂 Part translated from Polish: • Initials on a 1674 flask • Ping issue between Macs • Handwritten SVG transcription • Will the number 141 return • Japanese Compiler Optimizations • Song about Important Joe • Pope into a frog • Piping lines to prog • Dedalus’s Crime Paragraph • Clannad Scottish Gaelic Songs

Important Joe? 🤣

I heard the phrase on the radio and told ChatGPT to „Write a song about Important Joe” on a whim. The result was ... underwhelming. 🙂

Haha, wouldn’t worry too much about that. One thing I really hate about this yes-no-AI debate that I end up in is that it ends up being polarising. I would love to have a more nuanced public debate about its merits but also its costs that doesn’t exclude people based on their AI-stance.

Haha, this is not @ you btw. It just reminded me. Being curious is great. Being skeptical is also great!

I’m waiting for #ai-assisted-skepticism

yeah im in both and really not sure where i land on a lot of this stuff and find parts of both sides compelling. And there's a real internal struggle that has come with all this

I'd say it's perfectly fine to sit on both sides of that boat. Sometimes, you manage to find yourself a really effective way of using these fancy new toys - and that makes you happy. Other times, you receive a sloppy AI-written business ticket with really bad ideas which introduce unwanted complexity - the one that should not have been created at first place. And that makes you sad and miserable. There are good days, there are bad days.

heh I realise these comments are partly tongue in cheek but if anything the AI-assisted coders should be the most qualified to be skeptics themselves, having daily firsthand experience with the thing? Skepticism ≠ rejection

On both sides but rooting for the luddites

The technology isn't the problem to me it's who controls it and how it's used. Especially when it comes to deskilling humans, making their conditions worse. I have to use LLMs at work to keep the usage metric ticking up so I'm invested in it in some way

And the ecological and social impact of all the data centers they're building 😞

(I live in Ohio which is fairly high in data centers already and various towns are campaigning heavily against having new data centers built near them, because they're often already socio-economically disadvantaged and they don't want adverse impacts on their water and power bills)

I think those are good points. I would also like to add the danger of even stronger filter bubbles that reinforce what people were already thinking (“You are absolutely right”), plus that the models are trained by only a few companies, some of which participated in the first wave of filter bubbles. I just really appreciate the everyones included by virtue of “low barrier to entry technology” that the web used to be. That made the web a diverse place, and I like that.

(Not saying that the web is perfect. Just saying that the decentralized nature of the web made it a place that could be better than what came before)

Yeah, the fake confidence and sycophantic reassurance is definitely a problem for people without critical thinking skills 😞 I've seen Copilot be completely and utterly wrong about public, factual stuff, and then argue with me for several minutes before it produces the correct answer -- and then it produces an explanation of why it was still "right"! Recent examples: arguing at great length that Robert Downey Jr. was not going to be Doctor Doom in Avengers Doomsday (and even arguing that the film had not been announced by Marvel!), and trying to convince me that a movie quote was a completely different character in a different movie. In another argument, it actually convinced me of the wrong answer -- until I dug deeper via a web search.

Oh, and I was trying to replace the battery in a car key fob a few weeks ago, and Copilot was helping me -- and asked me to take a photo of the key fob opening (so it could verify where to apply pressure to open it up) and then arguing with me that the photo I uploaded was actually a broken USB adapter and not the key fob. Grr!

I have so much trust issues with AI generated answers. Weirdly enough, when I read someones blogpost about a dumb bug I had something: proof of work. Someone authored a blogpost about the thing that I am dealing with. Now with AI, I don’t know what to trust anymore. Is this a hallucination? Is the source a hallucination?

Its funny how that works. Also with something like a PR review or something. Knowing the author and their style made me a more effective reviewer. Now, all bets are off, it could be them, it could be their agent. They could have tightly controlled their agent, or they could have oneshotted this. AI also erodes that trust, I guess.

(I’m trying to be balanced here, let me know if you think I’m not)

Review is definitely harder. On my team everyone is heads down and creating a lot of code. Mostly with a lot of human judgement but one dev is throwing slop at the wall which pushes the work onto the reviewer / PM etc. The business folk I am struggling with. I hope to never hear “Claude/ChatGPT said this: …..” again. Too many low effort requests that can turn into a wild goose chase.