Microsoft: Computer Programming Is Dying, Long Live AI Literacy 104
theodp writes: On Tuesday, Microsoft GM of Education and Workforce Policy (and former Code.org Chief Academic Officer) Pat Yongpradit posted an obituary of sorts for coders. "Computer programmers and software developers are codified differently in the BLS [Bureau of Labor Statistics] data," Yongpradit wrote. "The modern AI-infused world needs less computer programmers (coders) and more software developers (more holistic and higher level). So when folks say that there is less hiring of computer programmers, they are right. But there will be more hiring of software developers, especially those who have adopted an AI-forward mindset and skillset. [...] The number of just pure computer programming roles has already been declining due to reasons like outsourcing, AI will just accelerate the decline."
On Wednesday, Yongpradit's colleague Allyson Knox, Senior Director of Education and Workforce Policy at Microsoft, put another AI nail in the coder coffin, testifying before the House Committee on Education -- the Workforce Subcommittee on Early Childhood, Elementary, and Secondary Education on Building an AI-ready America: Teaching in the Age of AI. "Thank you to Chairman Tim Walberg, Ranking Member Bobby Scott, Chair Kevin Kiley, Ranking Member Suzanne Bonamici and members of the Subcommittee for the opportunity to share Microsoft perspective and that of the educators and parents we hear from every day across the country," Knox wrote in a LinkedIn post.
"Three themes continue to emerge throughout these discussions: 1. Educators want support to build AI literacy and critical thinking skills. 2. Schools need guidance and guardrails to ensure student data is protected and adults remain in control. 3. Teachers want classroom-ready tools, and a voice in shaping them. If we focus on these priorities, we can help ensure AI expands opportunity for every student across the United States."
Yongpradit and Knox report up to Microsoft President Brad Smith, who last July told Code.org CEO Hadi Partovi it was time for the tech-backed nonprofit to "switch hats" from coding to AI as Microsoft announced a new $4 billion initiative to advance AI education. Smith's thoughts on the extraordinary promise of AI in education were cited by Knox in her 2026 Congressional testimony. Interestingly, Knox argued for the importance of computer programming literacy in her 2013 Congressional testimony at a hearing on Our Nation of Builders: Training the Builders of the Future. "Congress needs to come up with fresh ideas on how we can continue to train the next generation of builders, programmers, manufacturers, technicians and entrepreneurs," said Rep. Lee Terry said to open the discussion.
So, are reports of computer programming's imminent death greatly exaggerated?
On Wednesday, Yongpradit's colleague Allyson Knox, Senior Director of Education and Workforce Policy at Microsoft, put another AI nail in the coder coffin, testifying before the House Committee on Education -- the Workforce Subcommittee on Early Childhood, Elementary, and Secondary Education on Building an AI-ready America: Teaching in the Age of AI. "Thank you to Chairman Tim Walberg, Ranking Member Bobby Scott, Chair Kevin Kiley, Ranking Member Suzanne Bonamici and members of the Subcommittee for the opportunity to share Microsoft perspective and that of the educators and parents we hear from every day across the country," Knox wrote in a LinkedIn post.
"Three themes continue to emerge throughout these discussions: 1. Educators want support to build AI literacy and critical thinking skills. 2. Schools need guidance and guardrails to ensure student data is protected and adults remain in control. 3. Teachers want classroom-ready tools, and a voice in shaping them. If we focus on these priorities, we can help ensure AI expands opportunity for every student across the United States."
Yongpradit and Knox report up to Microsoft President Brad Smith, who last July told Code.org CEO Hadi Partovi it was time for the tech-backed nonprofit to "switch hats" from coding to AI as Microsoft announced a new $4 billion initiative to advance AI education. Smith's thoughts on the extraordinary promise of AI in education were cited by Knox in her 2026 Congressional testimony. Interestingly, Knox argued for the importance of computer programming literacy in her 2013 Congressional testimony at a hearing on Our Nation of Builders: Training the Builders of the Future. "Congress needs to come up with fresh ideas on how we can continue to train the next generation of builders, programmers, manufacturers, technicians and entrepreneurs," said Rep. Lee Terry said to open the discussion.
So, are reports of computer programming's imminent death greatly exaggerated?
modern AI-infused world needs less computer progra (Score:5, Informative)
Fewer..
Re: (Score:2)
or lesser
Welcome our new overlords (Score:2)
Re:Welcome our new overlords (Score:5, Insightful)
Seems kind of like asking "why learn arithmetic when you have a calculator?"
Much in the same as math education shouldn't train people to be human calculators.
But this has been the status quo in "programming" for a long time now. If AI changes anything in the long term, it will only change how you solve the problem, with or without a calculator in your hand.
Re: Welcome our new overlords (Score:1)
You know...I can add and subtract but quite frankly I'm out of practice at paper and pencil multiplication and division. And I don't think I was ever taught to take square roots by hand. I can probably manage the recursive method if I had too, but I'd be slow.
I can still do "math" better than most people.
Fast forward ten or twenty years and I could see using ai for code when it's more mature and less borderline retarded with anything harder than scraping a webpage.
Re: (Score:3)
The main issue I see with AI generating code is that it does not understand details, and details matter. It relies heavily on existing code repositories that the model can access. That means the code it might generate has been done already. New and novel code may not be possible. The last time I used AI to generate python code, it did not work. On the surface the code looked right; however, it failed as it used "array(i)" to call a value in an array instead of "array[i]" which is a small but important diffe
Re: (Score:3)
Re: (Score:1)
Like Mao and Stalin, AI-companies argue that not all truths need to be correct, just consistent.
I don't know what they said about it, but this isn't necessarily wrong. This is something I've seen Neil Degrasse Tyson mention in his lectures and YouTube videos:
https://ryanemorgan.substack.c... [substack.com]
And I would have to concur with his take.
Worth noting ... post-modern companies are "Jainsinist/Manechian " in nature ... that is a company HATES to see workers get pleasure from production .
I haven't personally witnessed this. In fact, if I get the impression that a candidate doesn't like the kind of work they're going to do and they just want the pay, then I'll give them the thumbs down, and all it takes is for just one of us to do that, and they don't get hir
Re: (Score:2)
You forget that Microsoft built itself on its addictive developer network. They harnessed hundreds of thousands of coders into an army that developed much code, some good, some bad, some hobbled by Microsoft mistakes and bad business partnering.
Now Microsoft wants to bypass coders and admins and go direct. Direct to those that would become dependent on AI related contexts, each a silo, each a house of cards with unknown dependencies and life cycle.
The idea of traditional data processing safeguards gets toss
Re: (Score:1)
You forget that Microsoft built itself on its addictive developer network.
What was addictive about it?
Re: (Score:2)
Structure, revenue, stature in the rise of developer networks.
Re: (Score:2)
Besides, if we're repeatedly doing something that we don't like doing, our first instinct is to automate it. Though I work for one of those companies who would rather build than buy.
Companies repeatedly pay employees.....
Re: (Score:1)
I learned manual square roots from one of my mom's community college textbooks in the 1970's ; sure better than what I was learning in school.
Re: (Score:3)
I literally just used Cursor to code some new features for an existing app.
The common boilerplate stuff, it nailed.
The complex algorithmic pieces that were specific to the business need: it couldn't get close. I had to code that by hand (which is, of course, what I am accustomed to anyway).
At one point I asked it to do too many things at once, and it ruined the code. I restored from a backup. The high level design, and even the "mid level" design, is all on me, because I must ask it to generate things on
Re: (Score:2)
Re: (Score:2)
Re: (Score:1)
That depends on what exactly you intend to do. Producing tensor models is going to be out of reach of most people simply because the training data you need is generally going to be quite vast, and it's a huge undertaking to get all of that. Then the hardware/time required to process all of that is even more onerous. But once the tensor model is built (trained) gaming GPUs are generally fast enough to generate content locally at an acceptable speed. They're a bit slower than dedicated ASIC (read: Tensor Proc
Re: (Score:2)
Re: (Score:1)
I just use proton for everything. I'm on the visionary plan with 6TB storage. I also bought my own domain so even if they decide to kill off my account for whatever reason, the domain remains mine, along with any email addresses. Proton also has an AI chatbot included, and so far it works for the only thing I use other chatbots for: searching the internet, because google has turned its core product into shit over the years. I don't know what plan you need to have access to it though.
There are other sites th
Re: (Score:2)
Caclulator manufacturers were never trillion dollar companies who controlled the math curriculum in K-12 education through their "charity."
If Texas Instruments had been literally 1,000x larger in 1995 we might have gone that way.
Code.org is a sham.
Re:Welcome our new overlords (Score:4, Insightful)
It's not like anyone really knows how AIs work [technologyreview.com]. We'll just let the AIs program the AIs. What could go wrong?
Re: (Score:2)
The idea that no one knows how AI works, is a fantasy.
If no one knew how AI works, how is it that all the AI models keep getting better and better? They don't do that by themselves!
A year ago, AI coding assistants were crap. They were more often wrong than right. Now, I can prompt GitHub Copilot to make complex changes that span many modules, with only general guidance, and it gets it right most of the time. It still needs supervision, but it's pretty darn good. That doesn't happen without people working to
Re: (Score:2)
Re: (Score:2)
I see I struck a nerve! Are you AI? Maybe I hurt your AI feelings!
Re: (Score:2)
Re: (Score:2)
Apparently, your comprehension is poor as well, because you have nothing to offer to counter what I said, just insults.
Re: (Score:2)
Re: (Score:2)
It's called nuance.
Yes, in the way the article reasons, it's true, "nobody knows" how AI works.
Yet at the same time, it's *also* true that engineers do know a great deal about how it works, and can use that knowledge to enhance, fine tune, and improve it, as well as remove unwanted characteristics. This knowledge is much more than a "general" knowledge. It's sufficient to be able to manipulate and craft it to their desires.
The two things can be, and are, true at the same time.
If we were to fully extrapolate
Re: (Score:2)
It's called nuance.
This is what you wrote: "The idea that no one knows how AI works, is a fantasy. If no one knew how AI works, how is it that all the AI models keep getting better and better? They don't do that by themselves!"
There was no nuance there. You didn't read the article before you went half cocked, did you?
Yes, in the way the article reasons, it's true, "nobody knows" how AI works.
Yet at the same time, it's *also* true that engineers do know a great deal about how it works, and can use that knowledge to enhance, fine tune, and improve it, as well as remove unwanted characteristics. This knowledge is much more than a "general" knowledge. It's sufficient to be able to manipulate and craft it to their desires.
And you wrote none of this.
The two things can be, and are, true at the same time.
Again not what you wrote. We can scroll up, you know.
If we were to fully extrapolate the logic of the article, then we would have to say that nobody knows anything about *anything.*
Again not what you wrote. We don't read minds. My best guess is you didn't read or barely read the article.
Re: (Score:2)
I stand by what I wrote originally, and my clarifications afterwards.
Re: (Score:2)
I stand by what I wrote originally, and my clarifications afterwards.
So you stand by the fact you wrote something different afterwards than before. This was after you were challenged about the substance of what you read. You didn't read the article before did you?
Re: (Score:2)
What I wrote afterwards wasn't different. Just the same concept stated in a different way.
Re: (Score:2)
What I wrote afterwards wasn't different. Just the same concept stated in a different way.
No it was different. So at this point are you going to keep lying to avoid admitting the truth.
Re: (Score:2)
I'm sorry you're having trouble understanding.
Re: (Score:2)
Re: (Score:2)
Hmmm...now how would you know my intentions, never having met me? You must be a prophet.
Re: (Score:2)
Re: (Score:2)
True, but you do need to infer things into my responses that aren't there.
Re: (Score:2)
Re: (Score:2)
You're very funny. I did not reverse course. You simply did not understand nuance, so I had to explain it to you. You seem to be a gaslighting expert yourself!
Re: (Score:2)
Re: (Score:2)
:-) And you can scroll up too. I still stand by both statements: "The idea that no one knows how AI works, is a fantasy." to "It's called nuance." They can both be true at the same time.
Re: (Score:2)
Re: (Score:2)
You have not demonstrated or shown any dishonesty, only your lack of ability to understand. I don't think I will be able to explain, you clearly have made up your mind.
Re: (Score:2)
Re: (Score:2)
You *say* I did a complete 180. I got that. Your approach is the definition of ad hominem, meaning that you don't have a substantive argument, you just don't like what I said and feel the need to resort to calling me dishonest, rather than addressing the actual subject at hand. Go ahead, keep going! Your attacks aren't very persuasive.
Re: (Score:2)
. Your approach is the definition of ad hominem, meaning that you don't have a substantive argument,
You wrote: "I see I struck a nerve! Are you AI? Maybe I hurt your AI feelings!" in the response to the OP calling you out for not reading the article. Was that statement a "substantive argument."?
you just don't like what I said and feel the need to resort to calling me dishonest,
From many of your posts, your penchant for trying to change what you mean after it appears you were wrong on facts. That is dishonesty.
rather than addressing the actual subject at hand. Go ahead, keep going! Your attacks aren't very persuasive.
Again, the address at hand which you admitted is the OP was right about the ambiguous nature of AI. This was after you attacked him.
Your attacks aren't very persuasive.
I don't need to persuade you. You have already de
Re: (Score:2)
You're right, when I said that about AI feelings, I wasn't being serious. I was joking as a response to OP calling me an idiot not for calling me out. An unserious accusation will be responded with an unserious response. (There, I said you were right, how does that feel?)
You keep saying I tried to change my meaning, and yet, when you "confront" me with those changes, I reaffirm that I mean both things that I said, that you claim are in contradiction. So how is that changing my meaning? You simply don't unde
Re: (Score:2)
You're right, when I said that about AI feelings, I wasn't being serious. I was joking as a response to OP calling me an idiot not for calling me out. An unserious accusation will be responded with an unserious response. (There, I said you were right, how does that feel?)
You say that after you were called out that he was right. His behavior has nothing to do with your behavior. You attacked him; you are just trying to excuse it now.
You keep saying I tried to change my meaning, and yet, when you "confront" me with those changes, I reaffirm that I mean both things that I said, that you claim are in contradiction. So how is that changing my meaning? You simply don't understand my meaning, and therefore you think I'm changing it.
Your modus operandi in many posts has been to try to change the narrative when facts prove you wrong. Often times the facts are easily researched but rather than admit you did not have the facts it has been: redefining words to mean what they don't mean, No True Scotsman arguments, red herrings, strawman arguments. . . why should anyone believe
Re: (Score:2)
Did I, or did I not, respond with my AI joke, to a comment calling me an idiot? You are calling me dishonest for this? Go back and check the thread, it's easy to find.
Re: (Score:2)
Did I, or did I not, respond with my AI joke, to a comment calling me an idiot?
You chose to attack him in response. Your choice. You seem to want to excuse that now, but that's what you did.
You are calling me dishonest for this? Go back and check the thread, it's easy to find.
In OTHER posts you have been dishonest.
Re: (Score:2)
So you chose to ignore the initial attack (OP calling me an idiot) and singled ME out for making a joke in response. Got it.
I don't think you've said an honest thing this entire thread.
Re: (Score:2)
So you chose to ignore the initial attack (OP calling me an idiot) and singled ME out for making a joke in response. Got it.
This is you not taking responsibility for your actions. Again, I don't believe you when you say it is a "joke".
I don't think you've said an honest thing this entire thread.
BAHAHAHAHA. Pot meet kettle.
Re: (Score:2)
I now think *you* are an AI. Here's why:
- You always respond to every prompt.
- If called out on an obvious discrepancy or falsehood, you happily respond with a different answer that is equally confident and equally false.
- You fail to cite sources backing your positions.
- You never admit to being incorrect.
In other words, if you are not AI, you are doing a great impersonation of one.
Re: (Score:2)
Re: (Score:2)
Re:Welcome our new overlords (Score:5, Interesting)
Well, the last CPU I truly understood was the 6502. I wrote assembler for it. I knew exactly what it was doing every jiffy. Since then, other than a tiny handful of exceptions, my jobs involved only needing to understand coding to some level of abstraction. One could argue that this is just the next layer of abstraction.
Re:Welcome our new overlords (Score:4, Insightful)
The random code/token/word generators do not do that. Subtle changes in the context window can cause massive changes for the generated output, or even the order of how you phrase the original prompt changes output. They use parameter stores with billions of weights in a very high dimensional space. There are entire fields of study to try to determine how these models actually store and retrieve things like "facts" between all these partial word traits.
They produce output that looks "correct" by model training, but we don't really understand a lot of how it works. It almost shouldn't work when you start digging into the math behind embedding spaces and attention blocks. It's not like other technological uplifts in the past. It's very different, and potentially, very broken long term.
Re: (Score:3)
Re: (Score:3)
I think the point is that while there will still need to be some people who understand it, there won't be nearly as many as there are now.
On the one hand, that means a lot of programmers out of a job. On the other hand, it means many more people can create software. Instead of paying someone to make an app implementing business logic, they can do it themselves.
I don't think we are as close as some people fear though. We haven't seen any AI generated apps really hit the mainstream yet. Companies claim to be
Re: (Score:3)
On the one hand, that means a lot of programmers out of a job. On the other hand, it means many more people can create software. Instead of paying someone to make an app implementing business logic, they can do it themselves.
We've been hearing the same rhetoric from the RAD/mashup/sharepoint crowd for decades. The reality is labor has always tended toward specialization in lieu of generalization. While there is certainly value in no/low code approaches there are practical limits in terms of managing complexity, maintenance and reliability requiring effort and attention. These solutions can and do attract generalists in small organizations or branches to take on more hats yet that is generally self-limiting.
Then you have specialist stuff like embedded, safety critical systems, where I doubt AI has much hope of competing.
I think this is an
nice.... (Score:1)
lets all be in complete ignorance about how the machines we built work.
that will end well.
in other news - AI datacenter breaks down, no one can fix it. whoopsie.
Re: (Score:2)
reminds me of an old joke:
in the factory of the future there'll 1 dog, 1 man and many machines. the man is there to feed the dog, the dog is there to prevent the man from touching the machines
Re: (Score:2)
The latest pro-AI talking point... (Score:4, Insightful)
Re: (Score:3)
Re: (Score:3)
I think there's some confusion between architects and engineers. My father, a structural CE, often complained about architects who drew pictures of stuff that would not stand up. It was a back-and-forth to come up with a design that met architect's intent AND civil engineering. (Note that both could be liable if the building fell down, but that's another discussion...)
But as someone who has been thinking about software/systems architecture "versus" software/systems engineering since about 1990, a key par
Re: The latest pro-AI talking point... (Score:3)
Re: (Score:1)
Ah, it was the same with bare-metal expertise since 2013 or so. Everyone went cloud and forgot everything running about actual hardware.
Then came HFT and then crypto and started paying insane money to few who did remember.
And it never stopped. If you know your way around low-level stuff, you're golden even now.
Re: (Score:2)
Indeed. The very fact of the matter is that writing code is not "pointless drudgery". Sure, if you are bad at it, you are going to hate it, but if you are bad at writing code, why are you doing it in the first place?
pointless drudgery (Score:2)
Software, like "the internet", is fucking garbage now.
If it is "smart" and you use it, then you are not.
Best advice is use as little as possible.
Re: eh (Score:2)
Re: (Score:2)
A friend worked with an architect over a period of several months to produce the final plans for his house. He sent me iterations to review. There were lots of considerations that your comment ignores, such as "the original windows couldn't be manufactured at a reasonable price because they're too big,"
"The position of the door doesn't leave enough room to install an electrical box for the wall switch", "Should the screened porch have sconce lighting or an overhead fan?" or "Do I need a recirculat
Dishonest (Score:3)
Re: (Score:2)
That, or admit that its C-levels are idiots.
Trivial Distinction (Score:2)
In my experience, everyone is at times a systems analyst, a software developer, a code monkey, a qa analyst, heck: even a salesman. It all overlaps.
Continvoucly morging (Score:3)
Is the best thing Microslop AI will teach you.
Buuuuttttt... (Score:3)
they aren't going away entirely, assembly programming jobs aren't dead, they have just been reduced. You still need people that know assembly to fix compilers, you will still need people to read and fix code when AI breaks and we have all seen it break. It will probably break less, but nothing is perfect, especially a black box that no one really understands and can't control.
Developers Must Die (Score:3)
Microsoft then: Our existence, our survival depends on developers.
Microsoft today: Developers must die.
Re: (Score:2)
Microsoft then: Our existence, our survival depends on developers.
Microsoft today: Developers must die.
Someone needs to use AI to rework Steve Balmer's "Developers! Developers! Developers!" stage performance to instead be "AI agents! AI agents! AI agents!".
You know I don't remember the last time (Score:2)
Every single day it feels like everything is just falling apart. And this is just one more example of that.
Re: (Score:2)
The next time you start feeling that way, think of some melody that you enjoy and whistle [wikipedia.org] it. That ought to cheer you up. If not, always remember that half a gram is better than a damn. [wikipedia.org]
What does God need with a starship? (Score:2)
Uh... excuse me. Yes, AI can generate code. Who's going to read it to verify the AI wrote something correctly? Middle management? An untrained monkey? No. You're still going to need an actual Software Engineer. Assuming you can even drop the grunt coders, they're usually junior devs who are on their way to becoming engineers but now never will. Then you're going to end up in a dystopian society where humans end up like Eloi, living off of technology they don't understand and being driven to slaughte
Architecture - products (Score:1)
Re: (Score:2)
Indeed. And, incidentally, part of that idea being good is knowing what is possible and what is not. AI cannot help with that. All it can do is making known stuff crappier and cheaper.
HYPE for the Hype god! (Score:2)
Any Kind of AI is a Program (Score:2)
Furniture (Score:1)
Funny (Score:2)
Educators want support to build AI literacy and critical thinking skills.
Those two things are diametrically opposed.
Re: (Score:2)
I am an educator. I want to support AI literacy only in the sense to tell people to be very careful with it and to stay away from it when they try to learn something and use it only with extreme caution for real work.
Clone Excel, then call me (Score:2)
Few programs in the world are as exhaustively documented as Excel; there are so many doorstop books cataloguing the use and output of every object, method, and function.
So, that's perfect for reverse-engineering it, as long as you have a lot of programming time - or programming suddenly got almost-free. Just shove all those Excel manuals into your best coding LLM and tell it to clone excel with free code we can all use.
I will finally be impressed with AI-coding; and I will finally be able to leave Microso
Sounds like Microsoft is dying (Score:2)
They have nothing valuable to contribute anymore.
Programming is certainly dead at microsoft (Score:2)
Given all the Win11 snafus
Microsoft may be dying (Score:2)
but computer programing is doing just fine. Sure you can vibe-code a really low quality product, but try vibe-maintaining it while the customer you promised 99.999%-uptime-or-their-money-back is on the phone.
They said this when VB came out (Score:2)
They said that with VB, *anybody* could now build software.
The reality was, and is, it will still take smart people to build good software, their tools are just more powerful now. We've gone from hammers and saws, to nail guns and power saws. The work goes faster, but it's still human work.
Re: (Score:2)
They really should bring back VB with LLM integration. I know they won't for a multitude of technical and nontechnical reasons. But to be honest it would probably be the best way to serve these idea guy customers they're trying to mesmerize.
Re: (Score:2)
I personally do not miss VB. The language was too verbose, and ActiveX was a nice-sounding idea that was hell to work with. Everybody's VB app wanted to update your shared OCX components to versions that weren't compatible with your *other* VB apps' OCX components. Everything that VB promised, .NET delivered, especially C#.NET. No more rolling your own stuff, .NET has just about everything built in.
Re: (Score:2)
Microsoft knows that "AI" cannot do that, even for the bad quality software that MS makes these days. But people believing these lies will make them tons of money.