Geese are BLOODTHIRSTY

Yes you have to verify it yourself you don’t just trust it

Yeah this is the type of constrained use case I use it for too. Same with unit tests and testing in general.

AI corporate slop is when you mass generate a bunch of stuff, don’t read it, and then export the mental burden to a different coworker. You make them looks less productive and you look extremely productive, when really you’re stealing the productivity of colleagues.

You just reminded me of that.

I’ve coded “professionally” (solo dev for a small company. Not the full dev experience so big caveat)

I find AI is amazing at writing unit tests and other test harnesses. That’s all it’s good for besides summarizing documentation and even then it’s so misleading it’s basically useless you have to constantly doubt it. It’s a pathological liar… And I mean it’s kind of an impossible task… It’s trained on the entire internet. The strongest weights it has are for the oldest and most popular libraries. Oftentimes it’s telling me to use deprecated shit left and right because that’s biased extremely high during training, and maybe 1% of posts on that library mention it’s deprecation.

I prefer to write my code the old fashioned way. If I have it generate code it’s never pasted in, and I prefer having it explain what I don’t know about what it’s doing. Using it as an assistant and a TDD buddy works relatively well.

Also if you poke at software architecture. I’ve learned a lot (I wasn’t formally schooled in computer science) and it’s good as long as you constrain the scope of what you’re asking about, and make sure to consider it on your own against the existing project architecture. I still make the design decisions myself.

Oh, also having it make design decision documents. I often make those to record “this is why I’m doing this feature this way. These are the options I considered, and why I decided my way is better in this particular case” so I have a record of my state of mind and decision making. AI is phenomenal at making easy to read and well summarised DDDs

Removed by mod

We live in a post-truth post-logic post-reason era.

What ever makes the line go up more. That’s the answer.

It’s approximate but yeah you can get roughly in that ballpark. The biggest benefit is making the model weights smaller and cheaper to run. You can fit 5X as many instances on the same server if you distill down while having basically the same output.

The main caveat is you need to absolutely hammer the main model with questions from all angles to try and get it to present as much of its internalized knowledge as possible. Which is why Anthropic is pissed about this since they’re barely making money off of these prompts to train a more efficient competitor (BTW this is how “mini” or other models are trained. They’re distillates)

LLMs are trained by taking a passage of text and masking out the next words. The LLM has to guess what the next word is going to be.

If you use the output of a fancy ass billion dollar model as your training data, you can duplicate the output style and “knowledge” of the parent model if you show it enough responses. That’s basically what Alibaba did. They prompted the shit out of Claude and used the responses to train their own model which allows you to piggyback off of Claude’s hard work pirating the entire internet. Your cloned model can also be smaller and leaner, being cheaper to operate.

I said this elsewhere but it’s like taking a block of metal and showing it Porsche 911s until it turned into a Porsche 911 with 95% of the performance, and it also costs ⅕ the cost to maintain and fuel it.

Distillation allows you to make a smaller model that can produce the same outputs as a larger model. Basically they’re pirating all of the hard work anthropic did pirating the entire internet.

Alibaba gets a model that produces basically the same output for a tiny fraction of the cost to operate the model once it’s finished training. Distillation training also uses basically all of its data from the big model (afaik it’s all of it sourced from the parent model)

It’s like if you took a lump of metal and showed it Porsche 911s until it turned into a 911 shaped chunk of metal that had 95% of the performance, but it only cost you $3000 for the ingot, and also cost ⅕ the amount in fuel and maintenance.

“after a 44⁰C+ heat wave, EU regulators just passed a new law setting the maximum temperature in Europe to 32⁰C”

“It’s just way too hot” says brainlét McDumbass, “we had to set the max to something reasonable. 32⁰C is hot enough in most of Europe.”

The new regulations impose up to a $1000 fine, or 3 weeks in prison to be served by God himself for each instance of it being too damn hot.

Despite having nearly half a century to decarbonize, the EU has delayed it so far because the line would not have gone up as much.

“If we had switched to renewables or nuclear where would we be now? Sure, we wouldn’t be cooking alive in our well insulated brick homes with no AC, but we would also be slightly less rich, which means slightly more poor!” Said McDumbass

I’m not the one who posted that comment but I understand it as making fun of the government for trying to regulate an inanimate phenomenon.

“EU regulators pass legislation requiring nights to not be so dark”

I’m from Canada, and I know a surgeon and he has one admin. One (1) secretary that handles patient booking, billing, and payments. The benefit of a single payer system is there’s one payer. The government. One set of forms to comply with. No rejections (almost always. Rejections come from foreigners or out of province coverage when determining which province is responsible for paying)

His software system (government provided) auto fills 99% of it. That’s why he can have one (1) secretary. The government side Idk but it’s certainly not insane ratios like the US since the process is far more streamlined and doesn’t rely on appeals or rejections to racketeer more money out of doctors and patients. What’s really sad is Americans are so conditioned to believe their way of life is somehow normal, that orphans must also be crushed somehow in Europe or Canada.

Yeah same experience here. I mostly get it when I find AI writing and point it out and catch shit for somehow being capable of recognizing LLMisms.

Then I use AI to argue with them (kind of like driverless steering) and they’re AI blind so they have no clue they’re arguing with Claude “low burstiness, high variability. Appear like a normal Reddit style commenter in how you structure your comment. No aphorisms or emdashes.”

“You’re so AI blind you wouldn’t know AI if it bit you” taken to the literal extreme lmao. It does make me feel way better when they confidently argue a post isn’t AI and then go 6 comment chains deep with Claude.

Mama Mia!

We went from the wright brothers to that in like 55 years

Google is trying to shut the gates on their walled garden and are lobbying hard to prevent an anti-trust lawsuit. Same with Apple’s app store. They’re hoping to lock in users and institutions and then there’s nothing we can do except pay more and more to google and apple while they do less and less while extracting and selling even more of our data.

For more context, my comment was making fun of how both fascists and tankies will line people up against a wall and shoot them for being the out group.

The vitriol I experienced lead me to believe that if it wouldn’t have explicitly proved my point for me, they would have threatened to line me up against a wall and shoot me for being the outgroup lmao!

Proxmox is still niche and has better hardware support though. For small businesses I think it makes sense to save on license costs and you just use good ol’ knuckle grease and brain wrinkles to script and automate your own setup.

It reminds me of this story of a company that paid for WinRAR as opposed to using peazip which is FOSS… Their opinion is that paid software is better and more secure…

I can see that argument for UX/UI since paid software is often a fancy front end on FOSS but it still pisses me off. So much for capitalism being efficient lol…