talent bifurcation
i've been recruiting project members and officers for ACM AI throughout this quarter and am getting concerned about the level of apathy in many underclassmen. it seems to have become the norm to present to others work that you can't even explain, or to point to your AI agents when asked about anything. here are some examples:
for a projects officer position, we ask candidates to come up with some rough idea of a project they'd like to run, and walk us through how it'd be structured, the stack they'd use, and what members might learn. but most candidates showed up to the interview with a painfully AI-generated slide deck (too much text, everything in a card/pill, gradient colors, "catchphrases everywhere"), some even reading a clearly AI-generated script off their phone. when we interrupted their monologue to probe further on certain details of their project or ask about claims they were making, they looked completely dumbfounded and floundered trying to make up details. in one particular case, a candidate claimed that their model would have 7 features. when I asked what the features would be, they told me that they're not sure what a feature is (and yet they were proposing a complex NLP automated trading pipeline).
when I recruited people for my project this quarter, I did my best to screen for people who would truly put in the work to understand their notebooks, but this has proven to be pretty difficult even with some interviews (although some were obvious - a few people left AI preambles in their notebooks). even with a smaller, hand-picked team, i still have members who are unable to even explain their pipelines or run their test suites when I ask them to. some clearly never bothered to look at or interact with their code at all before showing it to me. i find it somewhat alarming that most people are not embarrassed/ashamed to say "I don't really know, Claude did everything for me" when they're unable to explain their work, and since they're not embarrassed by it, their behavior does not change.
this is a pervasive issue in the population of new "programmers" who have never known a world where you didn't code with AI. it seems that they don't really feel responsible for taking ownership of their work. this doesn't make much sense to me - how can you expect someone else to read/understand something that you haven't even bothered to yourself? clearly we are in a new age where you can't expect people to be able to walk you through their code line by line anymore, but at the bare minimum you should not become a chinese room between an AI agent and someone interested in your work. I wonder if they understand that they haven't shown what they contribute beyond the AI's output, so there's zero reason to hire them. when I ask some of these students why they built something that way, I get "i don't know, i didn't build it." if you're just a proxy to AI, why wouldn't I just cut out the middleman and talk to AI myself?
now is a really interesting transitory period for the club. the senior officers who have graduated are truly smart, have a deep understanding of their domains of expertise, and have enough credibility that you would respect their opinions on certain topics. meanwhile, new talent has completely put the cart before the horse by first prioritizing "building a personal brand" and making sure they seem like they know a lot, before putting in the requisite work. the caliber of incoming talent just does not nearly match those who are graduating - even if you try to forecast their growth and learning throughout college, I think that their mindsets towards learning, being curious, and building projects for the experience of building them are stunted in a way that may prevent them from developing the understanding that our graduating officers have.
this also makes me wonder how much of the difficulty people describe in recruiting comes from a tough market, and how much comes from being unable to explain their own work. if this is a growing share of new grad engineering talent, I don't think it is crazy to predict that software engineering will quickly lose its luster as a field of curious people thinking analytically about interesting problems, and instead become filled with people who realized that they like high salaries, and know how to talk to AI and say buzzwords to others. I wonder if and how this talent collapse will spread to other knowledge work industries, and if anything is really safe. of course there will still be people who are truly talented, but i predict that that slice of the pie will continue to shrink. i think the places best positioned to retain them will be those with a meritocratic, low-ego, intellectually driven culture, where people are valued for the quality of their ideas and work, and have colleagues who push them to keep learning.
but even beyond all that, it makes me saddest to see a lack of curiosity, intrinsic motivation, and true interest in their work. you can tell that many people are building projects to tick off boxes, not because they enjoy the process. they would never go out of their way to learn an unfamiliar concept, but they would pretend to know it by reading an AI summary of it to your face. i just really want to find and work with low-ego people who still love to learn and care about the work they do, not about everyone knowing how much they do.
timestamped with opentimestamps: source · proof