TechTakes

Training a model on its own slop supposedly makes it suck more, though. If Microsoft wanted to milk their programmers for quality training data they should probably be banning copilot, not mandating it.

At this point it's an even bet that they are doing this because copilot has groomed the executives into thinking it can't do wrong.

People who work at Microsoft.

Source: Me, I used to, was driven moderately insane by their highly advanced and pervasive outbreak of corpospeak.

They are impressed by LLMs because they reproduce their inane dialect.

A company that forces you to write a "Connect" every half-year where you reflect on your performance and Impact™ : (click here for the definition of Impact™ in Microsoft® Sharepoint™)

chatgpt

By creating a language only they are able to speak and interpret, the managerial class is protecting its existence and self reproduction, while keeping people of other classes out or only let them in after passing through a proper reeducation camp, e.g. MBA program.

Business grads who think it makes them sound smart. I have to deal with way too many of them. It's infuriating, because behind it all I know just how dull most of them truly are.

I have no doubt that a chatbot would be just as effective at doing Liuson's job, if not moreso. Not because chatbots are good, but because Liuson is so bad at her job.

Wait until you experience AI "assistants" who extract info from your e-mails and texts to hallucinate wildly wrong appointments and summaries which then are cluttered throughout your calender and notes apps.

At this point any respectable workplace should just throw the whole Microsoft nonsense out and hire a sysadmin who knows what an immutable linux distro is.

LLMs are the next wave of popups here in the second quarter of the 21st century. I've become skilled at removing all the requests to let AI help me in whatever I'm actively doing. I about lost it recently when Excel threw one at me at work. NO, I DON'T WANT YOUR HELP!

Having a better guided search in a help feature I don't mind. But stop pushing it in everything, just have a way to get to it (and have it WORK when I use it!)

Unlike with coding, there are no simple “tests” to try out whether an AI’s answer is correct or not.

So for most actual practical software development, writing tests is in fact an entire job in and of itself and its a tricky one because covering even a fraction of the use cases and complexity the software will actually face when deployed is really hard. So simply letting the LLMs brute force trial-and-error their code through a bunch of tests won't actually get you good working code.

AlphaEvolve kind of did this, but it was testing very specific, well defined, well constrained algorithms that could have very specific evaluation written for them and it was using an evolutionary algorithm to guide the trial and error process. They don't say exactly in their paper, but that probably meant generating code hundreds or thousands or even tens of thousands of times to generate relatively short sections of code.

I've noticed a trend where people assume other fields have problems LLMs can handle, but the actually competent experts in that field know why LLMs fail at key pieces.

I work in a field that is mostly related to law and accounting... My biggest fear is to be replaced by a person who pretends that the AI’s output is smart

Aaaaaah. I know this person. They're an accountant. They recently learned about AI. They're starting to use it more at work. They're not technical. I told them about hallucinations. They said the AI rarely wrong. When he's not 100% convinced, he says he asks the AI to cite the source.... 🤦 I told him it can hallucinate the source! ... And then we went back to "it's rarely wrong though."

I was writing some math code, and not being an idiot I'm using an open source math library for doing something called "QR decomposition", and its efficient, and it supports sparse matrices (matrices where many numbers are 0), etc.

Just out of curiosity I checked where some idiot vibecoder would end up. AI simply plagiarizes from some shit sample snippets which exist purely to teach people what QR decomposition is. It's actually unusable, due to being numerically unstable.

Who in the fuck even needs this shit to be plagiarized, anyway?

It can't plagiarize a production quality implementation, because you can count those on the fingers of one hand, they're complex as fuck and you can't just blend a few together to try to pretend you didn't plagiarize.

The answer is, people who are peddling the AI. They are the ones who ordered plagiarism with extra plagiarism on top. These are not coding tools, these are demos to convince the investors to buy the actual product, which is company's stock. There's a little bit of tool functionality (you can ask them to refactor the code), but it's just you misusing a demo to try to get some value out of it.

And to that end, the demos take every opportunity to plagiarize something, and to talk about how the "AI" wrote the code from scratch based on its supposed understanding of fairly advanced math.

And in coding, it is counter productive to plagiarize. Many of the open source libraries can be used in commercial projects. You get upstream fixes for free. You don't end up with some bugs or worse yet security exploits that may have been fixed since the training cut-off date.

No fucking one in the right mind would willingly want their product to contain copy pasted snippets from stale open source libraries, passed through some sort of variable-renaming copyright laundering machine.

Except of course the business idiots who are in charge of software at major companies, who don't understand software. Who just failed upwards.

They look at plagiarized lines and count them as improved productivity.

I have the same worries in engineering. We had a presentation of some AI "consultancy" firm that was telling us that now is the time to stop hesitating and start doing with LLMs and gave some examples of companies "they" found in regards to our industry. When i asked, if they know any company that is willing to take the legal risks if their designs turn out hazardous, there was the sound of crickets. And just with that, LLMs are completely useless for any design tasks. If i still have to check the design to be in adherence with all relevant laws, norms and other standards, i might just do the design myself.

That is not to say, that there wouldn't be useful tools that fall into what is called "AI" these days. But these tools are designed for specific purposes, by people who do understand the specific purpose and its caveats.