Now I understand why at each windows 11 update, they introduce more bugs than ever
“Written by software” does not inherently mean AI.
Microsoft CEO Satya Nadella said that 20% to 30% of code inside the company’s repositories was “written by software” — meaning AI.
Nadella gave the figure after Zuckerberg asked roughly how much of Microsoft’s code is AI generated today.
I highlighted part of the article for you.
Code generators being a prime and ancient example
Does that mean that Microsoft shares are gonna crash?
And its all Teams.
Honestly, I’m not surprised.
Horseshit.
The current state of code generated by AI is sketchy at best. I often get plain wrong answers because the model tries to derive. It comes up with calls to functions and properties that just do not exist.
“You are right, I made a mistake. Here is a better answer.” Continues to give wrong answers.
Apart from that, apps that are glued together from AI generated code are not maintainable at all. What if there is a bug somewhere and you so not comprehend what is actually happening? Ask AI to fix it? Yeah good luck with that.
I do use AI for simple questions, and it works fairly well for that, but this claim by MS is just marketing bullshit.
“You are right, I made a mistake. Here is a better answer.” Continues to give wrong answers
The exact same wrong answer. Co-Pilot is especially bad for that. I’m practically giving up using it outside of vs code because the actual copilot AI is dog shit stupid m
I didn’t RTA, but if they mean ALL code at MS, that just can’t be true. They have legacy stuff going back decades, beyond just their windows platform. There’s no way 30% of all their code is replaced or newly created by AI.
They say that because they are selling it.
And yeah, my experience is the same. The most frustrating is when writing in a typed python, and it gives answers that are clearly incorrect, making up attributes that don’t even exist etc.
My brother said his superior asked him to use more AI auto complete so that they can brag to investors that X percent of the company’s code is written by AI. This told me everything about the current state of this bullshit.
“You are right, I made a mistake. Here is a better answer.” Continues to give wrong answers.
To be fair, the AI’s not wrong. It’s probably better, but just a teeny tiny bit so.
Honestly, AI is like a genie - whatever you come up with he’ll just butcher and misinterpret so you start questioning both your own sanity and the semantics of language. Good thing these genies have no wish limit, but bad thing that they murder rainforests while generating their non-sequitur replies.
This ^
“20%-30% of code inside the company’s repositories”
Now, if they had said “20%-30% of code written in the past 6 months…” I might buy that.
The repositories are going to have all the current codebase, likely going back years now. AI generated code is barely viable at this point and really only pretty recently.
No way 1/3rd of all current codebase is AI.
“Please move all comments from in-line to the line above, and add a separator line”
Even 20% of new code would be a stretch unless they count every first iteration of code written by AI that needs to be replaced by a human later because it was plain wrong.
Maybe they’re counting the six iterations of code it gives me as I tell it what’s wrong with each one.
20% of code = 20 lines of production code.
They mean new code, as per the article. And they mean code gen and IntelliSence
It would explain the constant barrage of system breaking bugs and RCE vulns
Turns out they have been using AI for decades.
Ah but it was nice for a minute imagining microsoft vibe coding windows…
“Auto complete generated 30% of characters”
Fixed it.
I bet they’re counting code written while someone had an AI plugin installed as “written by AI” and I bet that accounts for almost all of that 30%. On top of that, I’m betting that they made it mandatory to have such a plug in, and the other 70% is just code written before they mandated this.
I would be very surprised if 30% of their code lines had even been touched at all by anyone since AI coding assistants became a thing.
I wish this shot from The Terminator had the camera showing Sarah Conner’s face instead of Reese’s, because it’d be such an appropriate meme image on multiple levels for when someone makes a misleading claim about some current AI system.
I could see stuff getting small changes and them claiming that the entirety of the new version is “written by AI”.
Also, having 1/3 lines with obvious code that can be auto suggested correctly would make sense, but that is hardly code “written by ai” in the way they suggest.
Those are the easy time savings though, the safe easy stuff the developer doesn’t have to worry about anymore. (Giving them time do the gnarly stuff)
It is exactly the opposite, with simple, predictable auto-complete you didn’t have to worry about that anymore, with LLMs you always have to look at it in detail because every little thing could be just plain completely different and wrong.
I can read way faster than I can type though. You still check it, but it’s pretty good as that kind of stuff once you have an example for it to follow.
Reading code is usually orders of magnitude slower than writing code. Sure, typing might be slower than reading but to check if it is what you intended you have to understand it too.
Well, I’m generally very anti-LLM but as a library author in Java it has been very helpful to create lots of similar overloads/methods for different types and filling in the corresponding documentation comments. I’ve already done all the thinking and I just need to check that the overload makes the right call or does the same thing that the other ones do – in that particular case, it’s faster. But if I myself don’t know yet how I’m going to do something, I would never trust an AI to tell me.
I’d guess a lot of the people writing the code don’t even have it turned on, it’s just installed because management said it had to be, because management wants to be able to tell investors they’re “innovating work flows”.
Every few months I turn it on for a few days just to see if it is better.
Then I go back to the old AST based autocomplete that actually knows something useful about my code.
I am a small sample to confirm that’s exactly the reason in my brother’s company.
And in my company we’re pressured to make X prompts every week to the company’s own ChatGPT wrapper to show we’re being productive. Even our profit shares have a KPO attached to that now. So many people just type “Hello there” every morning to count as another interaction with the AI.
Power move by the zucc by first asking how much genai is used at Microsoft then refusing to answer his own question at Facebook 😂
Would be interesting to see how they measured that metric. Are they tagging individual lines as AI generated?
What those lines are too would be interesting,AI as auto complete is less dangerous than complete generation, but probably also less dangerous.
Most probably Microsoft has set objectives for how much LoC are from LLMs and developers invented numbers to match that metric (because they probably have things more important to do than counting LoC)
AI as auto complete is exactly what I was thinking.
I’ve seen lots of cases where AI appears as an auto complete suggestion and I can just hit <TAB> and it finishes the current line. It’s essentially filling in the boilerplate text. Heck in some cases it isn’t even right, but it’s close enough that I can change a few values.
I also want to point out that this isn’t particularly new technology. This existed before AI. It has perhaps expanded more, but it isn’t a revolutionary improvement, it’s an incremental one. So when we talk about usefulness, I think it is actually more useful.
Now if it could do all the magic planning and thinking, that would be more useful, but we’re not there yet.
Well, that would explain a lot.
I’m also guessing that at “up to 30%” of the company’s leadership decisions are being made by AI too.
“Up to” can also be 0%. Why is there even a need to say “up to” here.
So you are saying up to 100% of their code is written by AI?
Up to 100% of all Windows code was written by a Macaque monkey on meth.
So you are saying up to 100% of all Windows code is Mac code?
He used the word written by software. This is ambiguous and doesn’t mean AI, for example, using annotations for variables and generating the getters and setters would count. Right click and create function body for interface function definitions also.
They’re exaggerating to pretend their AI is more useful than it is.
Don’t forget code generation for stuff like bindings or database schemes
Intellisense in visual studio has also been really good for over a decade. Which is technically also written by software and not me.
I mean, really good intellisense is a great improvement, but it’s not replacing devs any time soon.
People have been using annotations to generate code since I rode my dinosaur to work.
I hope that that’s inclusive of something like lines of documentation.
??? No it’s not! Can investors sue because this is such an obvious lie? Pls I have 0.3 Microsoft shares
Are they including code generated to test their own models capability maybe?
So this explains why Microsoft Swiftkey is total dogshit now. Also why the Outlook barely works.
Its unbelievable.
I used to be able to swipe freely on SwiftKey, and now I can’t really do it without being extra careful and mindful of not spelling the wrong thing. Idk what Microsoft did to the product but I wouldn’t call it an improvement.
Idk what Microsoft did to the product but I wouldn’t call it an improvement.
I think the article we’re looking at here isn’t really hyperbolic. They got AI to write all their code and broke the Keyboard.
Just FYI, if you can live without swipping, I recommend FUTO keyboard., it is basically Swiftkey but it actually works and doesn’t come with Microsoft’s spyware built in.
It’s what I use now, and I’m really happy. Don’t be fooled by it being in Alpha, because it works flawlessly.
Thanks for the suggestion, I’ll check it out
That’s… not something to be proud of.