Today, I am going to be talking about AI and, in particular, why both the hype and the FUD over it are overblown.
The main reason for this is the way that current AI is created. In case you are unfamiliar with how it works, here is a quick primer. To train an AI, two things are necessary: a task and a trainer. The task can be anything from the creation of original video to the creation of boundaries of a particular item of interest (for example, “find me all the cats in this image”). The AI then works its magic (usually using neural networks) and tries to solve the task at hand. Then, the resulting creation is given to the trainer, who decides if the AI did a satisfactory job or not.
What this means is two things: 1) the AI is only as smart as necessary to fool the trainer into believing it has accomplished the task, and 2) there is still the potential for a fully-trained AI to fail to fool the trainer.
When people say they are worried AI will replace them as programmers, they mean that they think AI will fully produce error-free programs without the intervention of humans. This is false on two levels: First, the *trainer* must understand programming in order to train the AI. Second, the user of the AI must understand when the AI produces what amounts to sophisticated-looking nonsense. Contrary to popular opinion, this will (probably) mean an INCREASE in demand for trained programmers, as non-experts will be using AI to produce code that they themselves don’t understand and has subtle bugs not immediately obvious to their understanding.
I think if you’ve used GitHub Copilot before, you will have encountered this problem: You are writing code, and Copilot gives you a suggestion that looks correct at first glance but is actually either not at all correct or has some subtle bug. Do you think someone not trained in programming would be able to suss out this type of error? If not, who do you think they will turn to in order to assess the situation?
In conclusion, at least in its current incarnation, AI will probably not replace programmers writ large. Maybe one day, it will. Maybe. But until that day arrives, smart humans will always be in demand.
