In 1995, a then-little-known company launched with the mission to sell books on the internet (which, at the time, was also a little-known object floating somewhere in space). By 1997, that company had reached over 1,000,000 online transactions. Its chief competitor at the time was still alive and well, albeit beginning to notice the growing threat of e-commerce on its target market.
That company of course was Amazon, and its chief competitor was Barnes & Noble. Fast-forward to 2023, where we know of Amazon as something much more than an online bookseller. But we also know that we can’t replace the experience of going into a real, physical bookstore to smell that “new book” smell, perusing the shelves until we find exactly what we’re looking for (and probably leaving with a few extra copies of favorites to help anchor that empty shelf in the office).
It’s no stretch to draw parallels between this scenario and the one software creators and industry onlookers both face today with the rise of AI tools like OpenAI’s ChatGPT. Thousands of tools using the same AI technology have been created in just the last few months alone, most with the advertised intent of making automated decisions or creating pages of marketing copy or lines of code, all with just the press of a button. In fact, as I’m writing this draft on Notion, the first thing it suggested to me was to write a blog on this topic, using its own AI toolset.
ChatGPT is a language learning model, and as such it will be very happy to show you how much it’s learned from all of the prompts it’s been given. It can generate poems, speeches, even give you information on people in the news. But that information isn’t always correct, and the other content it generates, although it does so quite quickly, doesn’t exactly read like Maya Angelou.
As tempting as it is to use this feature and save some time, I find it takes away a lot of the most enjoyable aspects of writing - the creative process - that make such a task enjoyable in the first place. In the same way for writing code, GitHub’s Copilot tool promises to “code faster” and enables the developer to “do what matters most: [build] great software”.
Will AI replace developers?
What these marketing materials don’t tell you is the level of complexity of the prompt (or, for developers, code) that the AI is able to write. Sure, it can generate for-loops and write lots of standard or repeated “boilerplate” code, saving some keystrokes in return for more quality brain cycles, indeed precious commodities for busy developers. But there is a functional ceiling to this technology, reached quickly when compared to the complexity of the problem at hand. That complexity can turn this handy albeit simple tool into an expensive Clippy clone. And if that’s what they’re going for, where’s my anthropomorphic compiler with the friendly eyebrows?
Until this kind of technology can advance beyond the impossible limits of human innovation and creativity, there will always be a need for humans (developers, writers, cashiers, you name it) to monitor and guide the technology that, admittedly, can make our lives easier. But we remain, at present, quite far from that threshold.
AI in Education & the Future Workplace
As an educational tool, AI can be quite useful when used properly. Used poorly, it can also become a crutch. Much like the arrival of the abacus in eons past, personal calculators changed the way mathematics was taught in schools. Students learning multiplication or long division had to be forced to learn “the long way” first, before they started algebra and were able to do those calculations “by hand”. In my school, at least, calculator-users were always caught and shown the error of their ways when those students (who were totally not me) couldn’t do basic multiplication in front of the class. Public shaming aside, the ability to perform tasks that could be made easier or quicker by new technologies is still useful in almost any industry. Now, with the advent of AI, some development tasks can be automated in the same way.
But, what is the virtue of a calculator to a 8- or 9-year-old student that doesn’t know what the calculator does to get that answer? What if the batteries run out during a test, or the dog eats your AI? Similarly, most technical job interviews may not tolerate potential hires constantly referencing ChatGPT, or let the code interview continue with Copilot running. Indeed, some workplaces already can’t afford team subscriptions to Copilot, either financially, legally or morally. Jobs that utilize highly sensitive material or include working with personally-identifiable user data shouldn’t rely on tools like ChatGPT (hello, Samsung), since it is a language learning model, and anything put into it can end up as results on someone else’s query. So, while useful and quick, it’s a long way from secure or reliable - essentially, as one coworker put it, language models are “consistently confidently incorrect”.
Is learning software development still worth it?
Emphatically yes! Calculators didn’t put mathematicians out of a job. Indeed, it solidified their work and pushed it further ahead than anyone could have imagined. If you’re worried about AI taking over your job, or if you let that fear stop you from being proactive about learning the exciting possibilities (or shortcomings) of this new technology, then all the more reason to become a developer now. As with any era of automation in history, jobs are both created and changed when the workplace inherits new technologies. I would encourage anyone interested in AI now to take advantage of this moment, and put themselves on the cutting-edge of this technology by becoming involved in software.
Want to take a closer look at how Turing School of Software and Design can transform your career? Sign up for an online Try Turing session to learn more today!