正文 · AI 翻译
Professors are fobbing us off with Claude too...
Jay reporting from Aofei Temple
QbitAI | WeChat public account QbitAI
Jay reporting from Aofei Temple
QbitAI | WeChat public account QbitAI
AI is destroying my computer science degree.
AI is destroying my computer science degree.
The person who wrote this accusation is Katie, a second-year university student who has dreamed of studying computer science since childhood.
She has no prestigious-school pedigree and got only a B in high-school math. She had never received any formal training, either; the only “project experience” she could point to was a few rather rudimentary web games.
She had thought that, once she entered university, she could finally study computer science properly!
But who could have imagined—
Course outlines are churned out by Claude, final projects are graded automatically by LLMs, and even the caring emails from her adviser may well be automated scripts running in the background...
Faced with this absurd scene, she collapsed (just kidding):
Because of AI, my university seems to be going through an existential crisis.
Because of AI, my university seems to be going through an existential crisis.
Less than 24 hours after the blog was published, it shot to number 1 on Hacker News’s front-page rankings.
A computer science department in complete disarray
In her experience as a student, the school’s teaching policies have swung back and forth ever since the generative AI boom.
At first, of course, it embraced AI across the board:
Grading students’ assignments and writing feedback, generating course outlines, sending study reminders to everyone... Large models were put to work on all of it.
And the result?
Less than 3 months later, they found it was all hallucinations and junk content, and swiftly pulled back.
The entire teaching system has remained in a state of extreme contradiction—
You absolutely must not use AI!
Oh no, I misspoke: everyone must use AI!
No, wait... Maybe you still can’t use AI...
You absolutely must not use AI!
Oh no, I misspoke: everyone must use AI!
No, wait... Maybe you still can’t use AI...
Then something surreal happens.
In one module of the same course, the professor requires you to use AI for an assignment, saying it is an important part of your grade.
So you grit your teeth and pay for a Claude subscription out of your own pocket.
In the next module, you confidently upload code you worked Opus 5.5 hard to produce, only to find—
Wait, what?
What do you mean using an LLM is academic misconduct and costs me points???
To be honest, if it were only the rules changing from one day to the next, you could just about cope.
What upset the author most was that she had worked hard to get into university and was paying such high tuition precisely because she expected to learn systematically from expert professors.
And then you, a university professor, fob me off with Claude too???
Yes, the instructors sternly warn students against cutting corners, then turn around and hand their own requirements to Claude to generate teaching materials with one click.
At least assignments used to be carefully thought through by professors in light of teaching goals and reviewed by the teaching team. They were manageable in scope and built up step by step.
Now? She suspects these professors have not even opened the Markdown files.
They are full of large-model-generated filler. A 3000-word assignment brief asks for 15 submissions.
By her reckoning, the workload has tripled...
The result is that ordinary students with a weak foundation in math, including those who switched majors or entered through a different field, are forced to skim the material just to get through it quickly.
I came here to learn real things properly, damn it, not to skim the surface of 15 impossible tasks in a 3000-word assignment!
I came here to learn real things properly, damn it, not to skim the surface of 15 impossible tasks in a 3000-word assignment!
But that is not even the most frustrating part.
Many students spend more than 10 days working hard on it, yet, for reasons they cannot understand, they just cannot get it to run.
Then, 1 week before the deadline, the teacher suddenly sends an erratum email:
There were errors in 2 places in the outline. They have now been corrected; please resubmit your work.
There were errors in 2 places in the outline. They have now been corrected; please resubmit your work.
Give me back all those Tokens I wasted 😭
So why do this at all?
Katie says the professors cannot take all the blame. The real culprit is the administration.
Do you know what funders like to hear most? Learning at scale, preferably automated!
That way, they can squeeze as much profit as possible out of tuition without paying for additional staff.
Do you know what funders like to hear most? Learning at scale, preferably automated!
That way, they can squeeze as much profit as possible out of tuition without paying for additional staff.
If large models can automatically write questions and grade assignments in seconds, why spend heavily on teaching assistants or hire more lecturers?
Technology has become an excuse for the university to keep cutting teaching expenses without limit, piling so much work on lecturers that human attention alone cannot handle it.
Katie also says she hardly ever sees her adviser.
Their only contact comes when she has not logged into the online learning platform for several days and receives an email from her adviser—
She opens it anxiously, only to discover that her adviser did not write it; it is a triggered automated alert o (╥﹏╥) o.
Top of Hacker News
After the post was published, it quickly spread across the internet and reached number 1 on the Hacker News front page.
The comments also filled with people from the computer science community.
To my surprise, though, even among computer science people, not everyone was sympathetic.
The computer technology community split into 2 camps.
One camp consists of seasoned veterans already established in the industry.
From their perspective, the current chaos is just a brief growing pain in the evolution of a technological paradigm.
I have worked in this profession for more than 20 years, and I learn something new every day. Over the past few years, as AI technology has developed, I have learned faster and gained a deeper understanding of the systems I work on.
I have worked in this profession for more than 20 years, and I learn something new every day. Over the past few years, as AI technology has developed, I have learned faster and gained a deeper understanding of the systems I work on.
In short, AI is ultimately just a useful tool. If you use large models well for Socratic guidance, the learning resources available today are vastly better than those they had back then.
AI will not replace education and traditional ways of learning.
Instead, AI will improve your ability to understand and remember. As a technically minded, self-taught professional, what I have always wanted is to keep learning.
AI will not replace education and traditional ways of learning.
Instead, AI will improve your ability to understand and remember. As a technically minded, self-taught professional, what I have always wanted is to keep learning.
So quite a few experienced veterans even feel a little envious of newcomers:
Becoming a senior engineer does not really depend on how long you have worked in software development. I have seen some outstanding 21-year-old developers whose programming skills even surpass those of people who have worked for more than 20 years.
I think the reason they perform so well is that they have a solid foundation in computer science.
Becoming a senior engineer does not really depend on how long you have worked in software development. I have seen some outstanding 21-year-old developers whose programming skills even surpass those of people who have worked for more than 20 years.
I think the reason they perform so well is that they have a solid foundation in computer science.
But that is, after all, the story of a minority.
At the other end, the vast majority of recent graduates feel only suffocated.
One commenter posted the findings of a quantitative study:
Stanford University researchers found that the employment trend for junior workers they predicted last year is continuing and widening.
Specifically, employment among workers aged 22 to 25 in industries most exposed to AI is 19% lower than among their peers who are less often exposed to AI disruption.
Stanford University researchers found that the employment trend for junior workers they predicted last year is continuing and widening.
Specifically, employment among workers aged 22 to 25 in industries most exposed to AI is 19% lower than among their peers who are less often exposed to AI disruption.
Some industry professionals who have long been involved in hiring candidly confirmed this reality:
Indeed, giving a senior engineer a top-tier Coding Agent is more efficient than spending time training a fresh graduate from scratch.
I really do not know what advice to give developers who are about to graduate.
Indeed, giving a senior engineer a top-tier Coding Agent is more efficient than spending time training a fresh graduate from scratch.
I really do not know what advice to give developers who are about to graduate.
Spend 4 years studying computer science and end up unable to find a job...
So what, exactly, is the point of a CS degree?
Rediscovering the warmth of programming
As for that question, I think the answer may be in Katie’s blog post.
At the end, she wrote this:
I do not care whether computer science and math have reached a dead end! I still want to learn!
Okay, I do care about my future career, but that does not mean I want to forget everything I have learned or stop exploring.
There are still so many interesting directions to explore, some related to AI and some not. But the flood of headlines saying things like “software engineering is over” and “use AI or be left behind” is suffocating.
Can a girl not study statistics simply because it is fun?
I do not care whether computer science and math have reached a dead end! I still want to learn!
Okay, I do care about my future career, but that does not mean I want to forget everything I have learned or stop exploring.
There are still so many interesting directions to explore, some related to AI and some not. But the flood of headlines saying things like “software engineering is over” and “use AI or be left behind” is suffocating.
Can a girl not study statistics simply because it is fun?
After all this, Katie’s answer is—
I could keep going, but I will not. It gives me indigestion.
Maybe I will go back to OpenProcessing and make some sillier games.
At least now I really do know how to do some programming.
I could keep going, but I will not. It gives me indigestion.
Maybe I will go back to OpenProcessing and make some sillier games.
At least now I really do know how to do some programming.
Then she posted a few simple web projects she had made on OpenProcessing years earlier:
A simple space shooter with catchy music;
A Moogle that says “meow” when you click it;
And an interactive demo where sounds picked up by a microphone scare little fish away.
To be honest, they are very rough.
Any one of them is miles behind what Astra and Fable can now produce from a single prompt.
They might not even attract as much attention as a new model’s pelican bicycle demo...
But the moment you open one, you can feel something a model could never generate—
Life.
It is what many programmers were like when they first fell in love with writing code.
It was not about grinding through hundreds of LeetCode problems on a job-search assembly line, reporting in a weekly update how many Tokens had been saved on API calls, or catering to performance metrics dreamed up on a whim by middle managers at big tech companies.
Before all of this was bent toward practical gain, simply typing a few characters into a black terminal with white text and watching the pixels on the screen move and flow as you wished was enough to keep you up all night with excitement.
With our own hands, we built a supermachine that seems to know almost everything.
Yet, without quite realizing it, we seem to have shut out humanity’s most basic desire to explore and become absorbed in something...
Original blog post: https://lordtoots.bearblog.dev/ai-is-destroying-my-cs-degree/