The Ai thread

[fischersd]

Up here, one of the regular chants of the Maple MAGA Morons is how jobs at Tim Horton's have been going to immigrants and not their teenage kids. (because the poor racist bastards really need to get their coffee from people that are white like them!)
No - their excuse is that their kids can't find a part time job. "The immigrants are taking all of the jobs our kids should be getting."

I saw an advert for a modular "restaurant in a box" that is going to be marketed to the big chains - where robots do all of the cooking.

It seems a restaurant in Cali isn't waiting for that:

Robots do all of the cooking

I wonder how long until we see the car dealerships rolling out robots to replace the mechanics? They're certainly paying them more than the kids flipping burgers.

 

[Eric]

Up here, one of the regular chants of the Maple MAGA Morons is how jobs at Tim Horton's have been going to immigrants and not their teenage kids. (because the poor racist bastards really need to get their coffee from people that are white like them!)
No - their excuse is that their kids can't find a part time job. "The immigrants are taking all of the jobs our kids should be getting."

I saw an advert for a modular "restaurant in a box" that is going to be marketed to the big chains - where robots do all of the cooking.

It seems a restaurant in Cali isn't waiting for that:

Robots do all of the cooking

I wonder how long until we see the car dealerships rolling out robots to replace the mechanics? They're certainly paying them more than the kids flipping burgers.

The bigger problem is white kids normally won't take those kinds of jobs for low pay so establishments need to hire immigrant labor to fill the gap, then white people bitch that immigrants are taking their jobs on conservative media see, that's how it works in the MAGA world.

 

[Eric]

ChatGPT has been my new go to for more involved questions. It's mostly been useful for technical issues that normally takes a lot of Google Fu to figure something out, far smarter and more comprehensive which saves a lot of time and makes the entire process much less frustrating.

I've also noticed it's becoming more direct (rather than kissing ass) and I think it serves it well. For example I asked a specific question about video frame rates and why it won't work the way I expected and instead of something like "oh, you are right to be concerned, let me see if I can make it easier" it says "because that's not how it's designed to work". Not rude but direct and to the point.

 

[fischersd]

And now they've created a super-hacker AI that can hack virtually any system. But....we're just releasing it to these companies that can be trusted. There's no way that it'll get leaked to the public.

Mythos

 

[mr_roboto]

It's an old PR stunt for the AI companies - they're always talking about "our latest model does this Super Impressive Thing, but you can't see it yet, it's too dangerous". It is never that impressive or dangerous.

 

[Eric]

It's an old PR stunt for the AI companies - they're always talking about "our latest model does this Super Impressive Thing, but you can't see it yet, it's too dangerous". It is never that impressive or dangerous.

I have a feeling it could be if Anthropic were given to the Government with no restrictions.

 

[Nycturne]

It's an old PR stunt for the AI companies - they're always talking about "our latest model does this Super Impressive Thing, but you can't see it yet, it's too dangerous". It is never that impressive or dangerous.

I should have kept the links around, but I did notice folks pointing out that a what Mythos seems to be finding is also possible with Opus. A good chunk of the difference seems to be tuning rather than a full step-wise capability increase.

 

[Eric]

After spending quite a bit of time with Claude.ai building an iPhone app I can say it's been a surreal experience to say the least. It's like having an expert in their respective field at your fingertips at every moment. It's more than simply giving it instructions and telling it to build it exactly what you want, it's actual discussion and the exchange of ideas in a collaborative fashion.

For me it's been eye opening both in dealing with the frustration that comes with coding and the enjoyment you get with the results you've been looking for. Before my retirement my job was working with and managing developers and a project like this would've taken weeks or months of meetings, planning, coding, etc. when you factor in everyone's time and all. I was able to do it by myself in two days with nothing more than a basic understanding of coding.

 

[rdrr]

After spending quite a bit of time with Claude.ai building an iPhone app I can say it's been a surreal experience to say the least. It's like having an expert in their respective field at your fingertips at every moment. It's more than simply giving it instructions and telling it to build it exactly what you want, it's actual discussion and the exchange of ideas in a collaborative fashion.

For me it's been eye opening both in dealing with the frustration that comes with coding and the enjoyment you get with the results you've been looking for. Before my retirement my job was working with and managing developers and a project like this would've taken weeks or months of meetings, planning, coding, etc. when you factor in everyone's time and all. I was able to do it by myself in two days with nothing more than a basic understanding of coding.

I am about to take on a product that has been administered by another group. The sole person involved wrote some custom Java code 10 years ago, which no one wants to touch. I have the uncommented source code and an understanding of what it does, so the first thing up is rewriting the code in a modern language with comments, probably my first Claude.ai project.

 

[Eric]

I am about to take on a product that has been administered by another group. The sole person involved wrote some custom Java code 10 years ago, which no one wants to touch. I have the uncommented source code and an understanding of what it does, so the first thing up is rewriting the code in a modern language with comments, probably my first Claude.ai project.

I can hazard a guess here, comments and age won't matter, Claude will likely ask you to upload the code/files and will take a few seconds to analyze and then present you with options. Be amazed, I was blown away by it.

 

[Katbel]

How can you check it will not add some extra code lines that involves something you don’t want?

 

[rdrr]

How can you check it will not add some extra code lines that involves something you don’t want?

Regardless, it became self-aware at 2:14 AM EDT, August 29th, 1997.

 

[Nycturne]

How can you check it will not add some extra code lines that involves something you don’t want?

Going beyond the point I think you are making about security, this is my consistent issue with Claude and Codex. It can iterate until it meets some definition of a specification, but there's zero guarantee that it's the code you want, or will be maintainable down the road. I'm constantly slapping the hands of devs on PRs that clearly had little thought put into the approach behind something, because the code was just... bizarre. It worked, but it was literally the worst way to go about achieving the task.

The code of the Claude client itself got leaked, and it's pretty much a trash fire on multiple levels, including some approaches that are wasteful of tokens/resources which would explain how folks are burning through their quota faster than in the past. The irony is that the quality of the code lends weight to Anthropic's claim that they use Claude to write the client.

I don't discount the fact that Claude can produce working tools, but when folks are complaining about memory bloat from things like Electron, I'm wondering how quickly the sheen will wear off here. I'm seeing code quality of a kind that makes Electron look like an upgrade.

 

[Yoused]

Here is a hilarious "AI" generated map:

Three of the county names are correct (but not for Maine), though Clallam is spelled wrong. The rest is ridiculously wrong. Though, I think they did it just to get a reaction from the locals.

 

[Eric]

The code of the Claude client itself got leaked, and it's pretty much a trash fire on multiple levels, including some approaches that are wasteful of tokens/resources which would explain how folks are burning through their quota faster than in the past. The irony is that the quality of the code lends weight to Anthropic's claim that they use Claude to write the client.

Already ran into this, I uploaded a video to show it an issue and it cost me my remaining tokens for the session and it just cut me off without warning telling me to come back in 5 hours. I later asked it if it could warn me before jcutting off and it said that is not an option.

 

[Eric]

TheOnion

Man Who Threw Molotov Cocktail At Sam Altman’s Home Claims He Was Following ChatGPT Recipe For Risotto​

Man Who Threw Molotov Cocktail At Sam Altman’s Home Claims He Was Following ChatGPT Recipe For Risotto

SAN FRANCISCO—Following reports that a 20-year-old man had been arrested for throwing a Moltov cocktail at Sam Altman’s home, the suspect stated Monday that he only initiated the attack because he was following a ChatGPT recipe for risotto. “I’ve been using ChatGPT to help with cooking for a...

theonion.com

 

[DT]

TheOnion

Man Who Threw Molotov Cocktail At Sam Altman’s Home Claims He Was Following ChatGPT Recipe For Risotto​

Man Who Threw Molotov Cocktail At Sam Altman’s Home Claims He Was Following ChatGPT Recipe For Risotto

SAN FRANCISCO—Following reports that a 20-year-old man had been arrested for throwing a Moltov cocktail at Sam Altman’s home, the suspect stated Monday that he only initiated the attack because he was following a ChatGPT recipe for risotto. “I’ve been using ChatGPT to help with cooking for a...

theonion.com

The suspect went on to tell reporters that he still had “a whole fridge full” of Molotov cocktails at home, having attempted to prep enough risotto to last the week.



Also in the truth-stranger-than-fiction, this actually happened. A nutcase who wanted to stop AI because it was going to destroy the world tossed a molotov cocktail at Altman's house! (that is his actual house in the Onion article):

https://www.npr.org/2026/04/13/g-s1-117320/openai-sam-altman-molotov-cocktail

 

[Yoused]

The world will be fine, for at least the next billion or two years, after which it will no longer be able to support life due to rising insolation. If "AI" causes us to wipe ourselves out, the environment will gradually recover, and in a few thousand years, we will not even be missed.

 

[rdrr]

The world will be fine, for at least the next billion or two years, after which it will no longer be able to support life due to rising insolation. If "AI" causes us to wipe ourselves out, the environment will gradually recover, and in a few thousand years, we will not even be missed.

I think in the show "Life after Humans" speculated it would only take a dozen decades or so for nature to wipe out the memorials of our existence.