Gordon Brander writes: The thing about computers is that they can do anything you wish, so long as you specify your wish in exacting detail. However, LLMs relieve this constraint. An LLM can extrapolate what you mean (more or less) from just a few words. So computers can do vibes now.
It used to be that computers could only ever do exactly what you told them to do, nothing more.
If they did something unexpected (a.k.a. a bug), it was because you failed to 1) anticipate it could happen, or 2) specify with greater detail that it shouldn’t happen. But in this new world, computers can extrapolate your intent.
LLMS can act like a fill-in-the-gap machine, which alleviates you from the pedantic demands of coding from years past. There’s going to be a lot more “Oops, I didn’t mean to do that” in the future.
This should be what separates professional practitioners from hobbyists.
A good practitioner recognizes that leveraging some foresight is part and parcel of the job in order to mitigate unintended consequences.
Meanwhile, a hobbyist or amateur doesn’t even have the capability or experience to understand the assumed intent of an LLM until it’s too late and users have been affected.
Software we use everyday, software that many aspects of our lives increasingly depend on, needs less unintended consequences. Less “Oops, should’ve thought of that.”