Saw this article in the New York Times yesterday. It's been making its rounds. There was some discussion in some of the CS Ed related online groups, my friend Alfred Thompson posted his take here, and when my daughter, a professional SWE at Meta for close to a decade came over for dinner yesterday, she brought it up.

I haven't posted in a while so I figured I'd add my two cents.

The gist of the article is that with recent AI advancements, CS Education has to change.AI tools could make programming a less needed or at least a less in demand skill, students can lean on AI tools to complete class assignments, and tech companies are hiring less and looking for more and more experience in entry level.

Looking at the industry sign, things look bleak. It's hard out there. The NY Times article says that "many students say they send out 100 to 200 applications" which sounds crazy but really, even in the good times, many applicants, particularly if you didn't go to CMU or MIT which only accepts students they deem to be premade for success and then the school's reputations carry them to jobs, you were already sending out that many resumes.

It sucked then and it sucks now.

It's a broken system but it's been broken for a long time.

Of course, I'm not arguing that jobs are less plentiful now, they are but I suspect that things will correct in time. Alfred seems to share my beliefs here, or at least questions if this downturn will remain.

AI is great for taking what's already in the corpus of existing code, mashing it together, and spitting it back. The result is that it can save a lot of time for a seasoned programmer who can prompt the AI effectively and then take the results, debug it, and then adapt it to their needs .

Someone with no programming experience can't do that. I'm also skeptical of AI's ability to generate new and innovative solutions.

A friend of mine recently described relying on AI for code is like relying on a drunk grad students. That got a chuckle but there's some truth there. Hiring a drunk grad student as an engineer and blindly relying on them is probably not a good idea but having a talented drunk grad student on staff can, if used correctly make your company or team more productive.

On the industry side, another point that my daughter brought up is that companies still want mid level and upper level software engineers. If they're not hiring any entry level people, how will those mid level and upper level engineers ever come to be? Colleges have enough challenges making entry level engineers.

My take is that at some point, tech companies will realize that they still need entry level software engineers. They might need fewer or they might need them in different capacities, but they'll still be needed. Meanwhile, they'll refine how they use AI - tools to assist their software engineers, tools to allow their non-programmers to get some programming done in non critical areas, and to support other business efforts.

Sure, it won't be like the days when every two bit code school could place a "grad" with 3 months of Ruby on Rails experience into a super high paying job but kids with a strong background who are passionate about the field will still be in demand.

On the teaching side, sure, there will be changes. The NYT article talks about changes CMU has experimented with. They allowed AI in intro courses but then realized that students didn't really understand or learn what they were supposed to learn.

This is going to be a tricky thing to adapt to but nothing new. It's not really different than a student in a math class going to a teacher or professor for help, having the teacher go through a problem or proof and the student honestly thinking they understood things. Later, when they got home or on the exam, they realized they could follow the teacher but they didn't really understand the solution.

Now, some instructors understand this gap and work to teach students at a deeper level, some don't. AI is the same but we don't have self-aware AIs that can make the decision to try to do better.

It's also like when calculators came in. Teachers adapted, things were lost, but also things were gained.

It's going to take probably a few years but if they're allowed to, teachers and professors will figure it out. I saw a video the other day from a writing professor. He was talking about the process he's been going through to adapt to AI. The same will happen in CS.

Well, not universally. Some places will stick to the old ways. I mean, many places or at least CS profs still do things like they did 20 or 30 years ago which was bad then and is bad now. Some will follow the mainstream - what's dictated, say, in K12 by the College Board. I'm anticipating that will be mediocre at best.

Still, others will actually come up with wonderful ideas and they'll implement them.

It'll be a mixed bag but it won't be different than other innovations or other subject areas. It just seems a little scarier now because the changes have come so rapidly and from all sides - the students, the schools, and the world the students are graduating into.