Sept
Sept
Comments
71-
-
✨AI✨
-
It's maaagic. So much so that sometimes we don't know wtf it's doing.
-
Seems like a reasonable thing to bet a whole economy on. /s
I mean, back when it was a huge, poorly understood leap past previous technology it maybe was, but we know now that this is pretty much as good as it can do, just by scaling.
-
-
Is it just me, or are Github Copilot and ChatGPT getting dumber? I'm quite underwhelmed lately.
-
Maybe the more copilot is used, the more code on github is ai garbage, ths more copilot trains on github, the worse it gets.
Probably quite a lot of other things too, but I haven't used it so I don't know if it has got worse.
-
garbage in garbage out - sounds like a reasonable take
-
-
Novelty weared off
-
Yeah I remember the hype floating in the air when someone showed the magic AI writing a scan all files and folders recursively code in python. The result was like of a first year student, which is kind of incredible in itself to be fair, but it's also just the result like of a first year student so...
-
could be, maybe expectations went up too
-
-
It's been reported the latest model has been optimized for reduced power use, and abilities got reduced somewhat in the streamlining.
-
I guess enshittification has been moved off the backlog then..
-
-
There is a reason there is sometimes a notable decrease in quality of the same AI model a while after it's released.
Hosters of the models (like OpenAI or Microsoft) may have switched to a quantized version of their model. Quantization is a common practice to increase power efficiency and make the model easier to run, by essentially rounding the weights of the model to a lower precision. This decreases VRAM and storage usage significantly, at the cost of a bit of quality, where higher quantization results in worse quality.
For example, the base model will likely be in FP16, full floating point precision. They may switch to a Q8 version, which nearly halves the size of the model, with about a 3-7% decrease in quality.
-
Expertly explained. Thank you! It's pretty rad what you can get out of a quantized model on home hardware, but I still can't understand why people are trying to use it for anything resembling productivity.
It sounds like the typical tech industry:
"Look how amazing this is!" (Full power)
"Uh...uh oh, that's unsustainable. Let's quietly drop it." (Way reduced power)
"People are saying it's not as good, we can offer them LLM+ plus for better accuracy!" (3/4 power with subscription)
-
But if that's how you're going to run it, why not also train it in that mode?
-
-
Dont knows for these two but the enterprise version of copilot for office workers is so damn awefull its embarrassing.
It literally suggests to do stuff that it is incapable of doing.
I tried to get some doc as a flowchart using mermaid script, i know for a fact other llms have no problems with that:
After 3 attempts of it trying to code but resulting in an actual error it gives me,
Do you want me to provide this flowchart in visio format instead.
yes, if you can do that, it would be useful.
Generates a powerpoint about the topic.
that is a powerpoint, not a flowchart.
Tries to generate an image of a flowchart with not a single word spelled correctly and arrows going nowhere.
i didn’t ask to generate an image i need a flowchart
Sorry, if you want i can provide a flowchart in viso file format instead.
sure, give it another go
Generates a brand new bad PowerPoint.
I copy pasted the entire conversation to claude and it instantly gave me what i asked.
-
On par for Microsoft software in general. Seems like every week I discover new bugs in outlook.com..
-
-
ChatGPT has its moments but generally it creates more problems than it solves.
-
-
Really curious in what scenarios people would be writing enums with months and weekdays.
Because short of developing yet another library to handle date and time, everything else is likely a disaster waiting to happen...
-
A lot of developers are not passionate, or not curious, or don't know that libraries exist, or all at once (aka stupid). I've seen this everywhere.
-
The developer who was there when I started my last job believed that libraries should be avoided at all costs. He wrote a CSV reader from scratch in python. It didn't work in many edge cases. He didn't like it when I pointed that out. Nor when I showed him that his "better way" in another case was more than 10x slower using a profiler... At least he was using git, but the git history was full of long series of identical commit messages unrelated to code changes, because PyCharm has an option to reuse the previous commit message on a new commit...
He eventually quit and I spent 3 years refactoring his garbage before we finally had a tech team who could take over (I'm a scientist, with self taught coding skills). Pretty sure even after we had a tech team of 7 if was still a better coder than most, purely because I was interested in how coding works, and trying to understand underlying concepts.
-
Or they are passionate and curious about reinventing the wheel.
-
-
Wrapping a blackbox/legacy system would be a good reason.
Declare the old API in your new language, warts'n'all.
-
I did it once to pull out data from a spreadsheet into a database. Specifically, I needed
"${DataType}${Month}"for each month for 3 different datatypesIirc, i used an sql pivot (or unpivot) in that query too
Usually, it's situations like this where you're parsing data from strings, and you need some glue code to interface between the input data, and the date library you're using to actually resolve the datetime
-
-
{god, god, god, season, god, god, ceaser, ceaser, number, number, number, number}
-
And not even the right numbers! Months 9, 10, 11, 12 = "7," "8," "9," "10."
Days of the week = celestial body, celestial body, god, god, god, god, celestial body
-
March was the first month of the Roman year. Not sure why that changed.
-
-
-
its an enum; write the whole fucking thing, please.
-
But maybe the point is for display, and this is what they want to display
-
-
minerals, maxerals...
-
The vibe epoch is the number of milliseconds since Wednesday
-
It's like Excel is in my code editor.
-
Huit
-
Months, but Day. Sept, as already pointed out. The LLM is just following suit.
-
Layer 0
Wednesday, Wednestime
-
Not sure if bored, sarcastic human or highly advanced artificial intelligence.
Nice touch making Months plural and Day singular.
I also like how Wednessecond isn't going to be the end of the list, trailing comma is there.
Cursed.
Wednesmillisecond, Wednesmicrosecond, ...
Wednesnanosecond, Wednespicosecond, Wednesfemtosecond, ...
Can you tell what language it is? Some allow you to have dangling commas (I, personally, hate it but it could be the end of the list, nevertheless).
C#
I understand the argument in favour of dangling commas, but oh gods I hate it. Aesthetically it's just so awful.
It could be Typescript or Rust. But probably a lot of oter languages have enums expressed like this too. Likely not C, because no semicolon, and no prefix on the variants.
Well, there's multiple months in a year, but only one day per day, so that makes total sense somehow.