Singularity
Few people seem to grasp the implications of AI and talk about it openly.
If my other posts were fact-based journalism, this is going to an opinion piece.
Few people seem to grasp the implications of AI and talk about it openly.
Most people who I have seen talk about AI and its future effects on the tech industry only talk about AI in its current state.
This is a mistake because AI is still in its infancy.
You are right, that AI cannot do the work of a software engineer right now.
The hallucinations can be fixed eventually, and its quality of work can improve, and it has been.
Record amounts of money are being spent on improving AI by companies like Meta, Google, and Microsoft.
When you consider that AI only requires an electricity bill and doesn’t require benefits, sleep, etc. it’s easy to see the business value compared to traditional employment.
Companies don’t employ us for a laugh; they employ us because they can make money off our labor and it’s the cheapest way they can make a profit.
I also think it's naive that companies are going to employ us without us being useful even if the government tries to make them.
Companies are really good at appearing to do the right thing while also not.
They will do anything in their power to not employ people unnecessarily, including lying.
AI could create a utopia if it is used properly, because now labor is outsourced to AI, as long as the common man is given the share of the profits.
I wouldn’t hold my breath though.
Will AI replace software engineers?
Hear me out, but probably many of them, yes.
There are real skills that go into engineering, but the output at the end of the day is code, which is basically text in a file on a computer for the most part.
Especially with the creation of tool calls, AI is likely going to be able to grasp codebases and do an okay job at developing, increasingly so.
Even today, Google has like 30% of the new code in their internal codebases coming from AI.
Some people say, my codebase is very large, and AI cannot grasp the entire context, therefore I will be needed as a developer.
I think over time AI will be able to work with large and larger amounts of context and this problem will fade into the background.
There is definitely business value to be had in doing this and it's a well-known problem.
Software engineers are very highly paid, if an AI agent (which gets better all the time) can do a software engineers job at 1/10th or 1/100th or even possibly less of the cost, companies will probably go with that.
AI will still likely need someone to check its work, at least for the time being, but this can be done with far fewer engineers than we have had and have now.
If it took 30 engineers to develop a feature for an app without AI, it might only take 15-20 now, or even less perhaps.
This can easily be understood as the automation of rote work that junior devs used to do, but that isn’t the only outcome.
Where are those 10-15 people going to go?
What about the new college grads looking for a role with no experience that the university system churns out every year?
There’s already been so many layoffs since 2023, especially now that the government is not hiring.
On top of these things, AI coding has come a very long way in a very short time and is overwhelmingly likely to improve as it is a major motivator for AI firms, as well as the overwhelming business value.
AI has been able to come up with minor mathematical breakthroughs at Google, I think it eventually will be able to create mostly original code that is specifically tailored to the business problems at hand over the next few decades.
There was an expert that spoke at my community college when I was a student who said that in 2045 you will be able to write product objectives for AI, and then AI will create an entire product with all necessary work done to make it a reality.
I am not saying every software engineer will be laid off in the next 2-5 years, far from it.
But I expect the workforce to shrink as more work can be done by far fewer people.
If it took 20 people to code an entire feature in 2023, maybe it takes on 10-15 now.
Of course, businesses could just start hiring people to create more features but seems unlikely because most businesses I have seen ruthlessly cut costs and aren’t incredibly skilled at creating new business.
As a side note, I have been following what the CEO of Amazon Andrew Jassy has been saying about AI.
While it can be interesting to listen to what CEOs have to say about AI, it is ultimately kind of useless because we don’t know why they are saying what they are saying and if they are telling the truth.
Andrew Jassy could just be talking to his shareholders and creating hype for Amazon as well as show that Amazon is forward looking about AI.
However, the layoffs have started if I remember correctly, though it could be for other reasons.
AI will shrink Amazon's workforce in the coming years, CEO Jassy says
Will AI replace recruiters?
I think there is a very high likelihood of it, yes.
AI once tweaked should be able to gather employees and hand them to hiring managers based on parameters written via a form, natural language, or some other means.
I can imagine LinkedIn or possibly Indeed putting in AI agents to work with recruiting data directly.
The other aspect of recruiting is taking calls, which I will touch on next.
AI recruiters could both feed candidates to a human who calls them, or it could one day do the phone interviews itself.
Will AI replace anyone who takes calls all day?
Most likely, yes.
AI is able to talk on the phone somewhat competently, more so all the time.
My local heating and air company has an AI receptionist who is quite good actually, the first few times my mom called she thought it was a normal person.
The AI can take notes on your AC problems and is programmed to reach out of specific people that you ask for.
It’s much more efficient to have an AI voice bot doing your calls than a human as it can take calls 24/7 and can be on the phone with as many people as necessary, as well as being potentially cheaper.
Tier 1 support is probably cooked as well as almost any kind of telemarketing in the next few years.
It's also highly likely that there will be a large uptick in spam calls once this technology become ubiqitous.
We’re basically waiting until they work out the kinks, so it could be a few years.
Outsourcing to AI is cheaper than outsourcing to India most likely.
This is a huge industry however and it is quite disruptive if it happens.
If you are someone who answers calls all day, consider another career while you still have a job.
On vibe coding / AI assisted coding
Vibe coding is likely the future as AI matures or at least has an increased role in software development.
AI is better than a beginner at coding now, but it will likely improve and when it works, it can code much faster than any human.
I remember how coding with AI was when it first came out in 2022 compared to now, and it has improved significantly.
Once AI has code quality, it will be faster than any human programmer.
As the tools improve, companies will likely use AI assisted means more and more to create features rather than relying on time consuming development processes and reducing the need for labor.
Goldman Sachs recently “hired” the AI agent Devin to work for them, possibly a publicity stunt.
Conclusion
This blog is called the singularity because it may be the most important invention in our lifetimes, perhaps barring the internet, perhaps surpassing it.
A lot is bound up in its success and failure.
It could be the automation of a large % of white collar work once the kinks are worked out.
If you want to see my take on what I think we should do in response to AI, see here.
A Brief Catechism for Investing in the Age of AI
The thing to keep in mind is that AI is still in its infancy and it will most likely continue getting better.
In its current state, it isn’t replacing many people, but that doesn’t mean it won’t in 2-10 years when the software is significantly better than it is now.
The performance only needs to be 80% - 90% of the average engineer and a fraction of the cost with a nice, convenient wrapping such as connecting through version control.
We lose as workers and as an industry if 40% of us get laid off due to AI automation.
The industry is already bloated with labor because of over hiring during COVID, and the layoffs don’t show signs of stopping especially with the government layoffs.
On LinkedIn for local in person jobs in my area, when I was a student in 2023, most jobs would have between 15 - 45 applies, now every job is 100+.
100+ could be a very high number, such as 500 - 1000 as 100+ is just the max value listed.
Nonetheless, it’s like multiple hundreds of % increase compared to 2023, as most jobs had 15-20 applys.
It was definitely competitive then, most of the people I knew in college were not getting internships and some of them never got jobs after graduating, but it's much worse now.
It’s a very negative thing that technology is a known industry that people associate with high salaries and prosperity, because everyone wants to work in tech now.
I have stopped recommending to people they go into tech; the likelihood that someone grind for a few months and then gives up with no job is quite high.
The people who will succeed at tech are not asking if they should go into tech, I think.
As a side note, paying $20 a month for AI is likely a teaser price, and I would be surprised if they continued to give us large free tiers as they currently do.
AI needs performance improvements to use less electricity, chips, and water, but it likely will see this soon, perhaps through increased prominence of distillations.