My long term fear with AI is that by replacing entry level jobs, it breaks the path to train senior level employees. It could take a couple of decades to really feel the heat from it, but could lead to massive collapse as no one is left with any understanding of how existing systems work, or how to design replacements.
> It could take a couple of decades to really feel the heat from it, but could lead to massive collapse
When you consider how this interacts with the population collapse (which is inevitable now everywhere outside of some African countries) this seems even worse. In 20 years, we will have far fewer people under age 60 than we have now, and among that smaller cohort, the percentage of people at any given age who have useful levels of experience will be less because they may not be able to even begin meaningful careers.
Best case scenario, people who have gotten 5 or more years of experience by now (college grads of 2020) may scrape by indefinitely. They'll be about 47 then and have no one to hire that's more qualified than AI. Not necessarily because AI is so great; rather, how will there be someone with 20 years of experience when we simply don't hire any junior people this year?
Worst case, AI overtakes the Class of 2020 and moves up the experience-equivalence ladder faster than 1 year per year, so it starts taking out the classes of 2015, 2010, etc.
> Worst case, AI overtakes the Class of 2020 and moves up the experience-equivalence ladder faster than 1 year per year, so it starts taking out the classes of 2015, 2010, etc.
This is my bet. Similar to Moores law. Where it plateaus is anybody’s guess…
Juniors and offshore teams will probably be the most severely impacted. If a senior dev is already breaking off smaller tightly scoped tasks and fixing up the results, that loop can be accomplished much more quickly by iterating with a LLM. Especially if you have to wait a business day for someone in India to even start on the task when a LLM is spitting out a similar quality PR in minutes.
Ironically a friend of mine noticed that the team in India they work with is now largely pushing AI-generated code... At that point you just need management to cut out the middleman.
Management will cut down your team’s headcount and outsource even more to India ,Vietnam and Philippines
Management did all that at companies I've worked for for years before 'AI'. The big change is that the teams in India won't 200 developers, but 20 developers handholding an AI.
The worst case for such a cycle is generating new jobs in reverse engineers. Although in practice with what we have seen with machinists it tends to just accelerate existing trends towards outsourcing to countries who haven't had the 'entry level collapse'.
We've already eliminated certain junior level domains essentially by design. There aren't any 'barber-surgeons' with only two years of training for good reason. Instead we have surgery integrated it into a more lengthy and complicated educational path to become what we now would consider a 'proper' surgeon.
I think the answer is that if the 'junior' is uneconomical or otherwise unacceptable be prepared to pay more for the alternative, one way or another.
I’m actually worried we’ve gotten a kickstart on that process already. Anecdotally it seems like entry level developer jobs are harder to come by today than a decade ago. Without the free-money growth we were seeing for a long time it seems like companies are more incentivized to only hire senior developers at the loss of the greater good that comes with hiring and mentoring junior developers.
Caveat that this is anecdotal, not sure if there are numbers on this.
This is what I fear as well: some companies might adopt a "sustainable" approach to AI, but others will dynamite the entry path to their companies.
Of course, if your only goal is to sell a unicorn and be out after three years, who cares... but serious companies with lifelong employees that adopt the AI-first strategy are in for a surprise (looking at you, Microsoft).
That said, the first thing that jumps to my mind is cars. Back when they were first introduced you had to be a mechanically inclined person to own one and deal with it. Today, people just buy them and hire the very small number of experts (relative to the population of drivers) to deal with any issues. Same with smartphones. The majority of users have no idea how they really work. If it stop working they seek out an expert.
ATM, AI just seems like another level of that. JS/Python programmers don't need to know bits and bytes and memory allocation. Vibe coders won't need to know what JS/Python programmers need to know.
Maybe there won't be enough experts to keep it all going though.
