Why I'm not worried about AI
In today’s article, I talk about the reasons why I’m personally not worried about AI and what actually concerns me
There’s been a lot of scaremongering recently talking about how AI is going to replace software engineers, I’ve heard a lot of people worry that their roles are going to become non-existence and AI will replace us all.
That just simply isn’t true, at least by today’s standards.
In today’s article, I talk about the reasons why I’m personally not worried about AI and what actually concerns me...
How it all got started…
A month or so ago, a product called ‘Devin - The First AI Software Engineer’ got showcased in this demo:
And it kind of sent the internet into a frenzy, believing that AI could effectively replace what we do as software engineers, everyone was writing about it, everyone was creating videos about it and there was a lot of fear.
Until…
It was shown to be a lie.
This video got released shortly after…
Essentially debunking the demonstration and talks about how Devin does a lot of things a Software Engineer typically wouldn’t do.
The demonstration made it look like Devin was solving bugs within the repository, but actually what Devin was doing was solving problems with the mess they had created, and making even more of a mess of it in the process.
As well as that, it didn’t follow basic instructions that were in the repository README that would have solved the issues it ran into.
Maybe it is like a Software Engineer after all? LOL
I really recommend you check out the video if you want to know more about this, but otherwise, lets get into the areas I think AI isn’t there with yet…
Critical Thinking
So one of the things that AI can’t really do right now is critical think, that’s the ability to look at lots of data and determine what is good and what is bad.
It has to be trained on that.
Take for example when new libraries come out that end up being the standard way of how to do something.
AI won't automatically prioritise them unless it's explicitly trained to do so.
Instead, it tends to stick to the most common solutions, even if they're outdated or incompatible with the latest versions.
It would likely use the most common instance of that problem being solved.
Which might not even work with your latest version.
Which leads me to…
It’s only as good as the data
If your AI is trained on bad data, you’re going to get bad data out.
Unfortunately, there's an abundance of unreliable data out there.
Keeping the training data up to date and ensuring it reflects what's considered 'good' is a mammoth task in itself, and it's easy for it to become outdated quickly.
Software engineering is one of those fields where things that were relevant a couple of years ago, are now outdated as the new versions of the languages and frameworks come out.
AI is only as good as the data you put in.
Gathering requirements
I would say the most challenging thing about being a software engineer is figuring out the problem that the customer actually wants solving.
And sometimes they don’t even know.
They’ll come to you confident, telling you that they really want a big red button that they can easily press that will send out alerts regarding their latest deals.
When what they really want is the ability for alerts to be sent out automatically when new deals come into place.
The ability to push back and understand the why in why someone wants something, and give them the solution they actually want, is the hardest part of building a product.
Customers do not need a yes person, they need someone to challenge them and actually help them get to the root cause of their issue.
AIs often comes across as over confident in it’s solutions, and doesn’t really seek to understand the why. A lot more needs to be done on this area.
In my experience, every customer that I have worked with has always had things that I’ve had to do that typically are not found on the internet or in some code repository somewhere. I’ve often had to just figure out by reverse engineering the code, debugging the code and/or altering some configuration to try it out.
AI typically relies on predicting responses based on its training data. It currently lacks the ability to reverse engineer code or innovate new solutions without a lot of input on requirements.
I’m concerned
One of the things that massively concerns me about the use of AI is the attitude I’ve seen some companies having…
I do worry that companies are using AI as a reason to not hire new talent into the industry, and instead using seniors and AI alongside to have better productivity gains.
If they do this, I worry that we will end up with a skills gap in the next 5-10 years when current seniors/leads have either moved to new roles or retired.
This would mean we have a gap of talent of those able to work with AI and debug issues that have come from working with other developers and AI generated code.
AI is not a replacement for hiring new talent into the industry. Hiring new talent is crucial for longevity of the tech industry.
Productivity gains
All being said, I truly believe that AI will make us be able to do so much more, and reduce the need for us to do repetitive, boring and tedious tasks.
I love that I can use GitHub Co-pilot Chat to ask advice on something wrong with my code, or I can use GitHub Co-pilot to help me write documentation and test cases, and yes whilst I need to guide it a lot, it can typically learn from my previous code the format I want my new test cases and documentation to be like.
This saves me a tonne of time, and if you don’t use AI as part of your role, you will find yourself being left behind.
And that’s all from me!
Ultimately, if you use AI as part of your everyday role and focus on areas that AI can’t do, then you will continue to succeed and thrive in the industry.
Just like anything in tech, AI is a tool that you can use to your advantage, and rather than be scared of it, figure out how to wield it!
AI will not replace software engineers.
Software engineers using AI will.
If you found this article useful, please like, restack and subscribe, and also check out the video where I talk about this too:
Articles I enjoyed this week:
By Irina Stanescu - I honestly love everything Irina writes, it’s always so refreshing and insightful.
Spot on. AI is nowhere close to being as good at reasoning and problem scoping & solving as a human dev. But it's also true that some CEOs won't care or understand that and replace devs or neglect to hire new devs anyway. A lot of business decisions are based purely on hype and snakeoil.
The only thing I would add is about the phrase "AI won't replacr developers. Developers who use AI will" - I think that plays a little into a narrative being pushed by AI hype merchants that is not true. In reality, if you are a great software engineer you can refuse to use AI crutches and be just fine and just as kick ass as ever. These tools are nowhere newr good enough to be considered essential or advantageous for anyone except for perhaps very junior developers while they learn or experienced devs who for some reason are trapped in a job where they only have to write very mediocre, repetitive code solving problems that have already been solved by thousands of other developers on github (where the LLM model is sourced).
A great read, Jade!
I totally agree with that too. We could use the tools as engineers, it can make our lives a lot easier. But it's hard for us to be 100% replaced by them.