The other day, there was a post I read on LinkedIn that started out, “I’ve stopped caring how well a candidate writes SQL.” It continued “Anyone worth interviewing can write the query. If they can’t, a model will, in four seconds”

I will just let that clear diss on on my chosen profession that “Anyone worth interviewing” can do my job and if not, they can just farm it out to a computer. I will also not get into the idea that these models can understand all the requirements and make sure that the code is, if not perfect, good enough. This will actually be the day I fear for humanity because the computer has actually learned to read minds in real-time”.

But we all have to be honest with ourselves. AI models do some cool things, for sure. And the pre-AI chatbots weren’t exactly chopped liver when it came to that task.

And the point of that post was simply that what was needed was people who wouldn’t just jump in and start coding. Which is a good point for sure.

But how many skills can truly be replaced entirely with software

I submit to you a few really silly examples that take the initial premise to levels that are semi-terrifying on face value.

The future of air travel

Airline X: “As an airline, we have stopped caring if a pilot candidate can fly a plane. The plane flies itself. We just need the pilot who can ask the passengers where they want to go and let them know when things aren’t going right”

Pilot: “Folks, we asked the autopilot how we could quickly get everyone some water because they are thirsty. Because we are running low on fuel and water, it was preparing to divert us immediately. Apparently we have enough fuel to get to the AAA airport, but since passengers are really thirsty so we are going to land in the LLL lake so you all can get a drink so you will stop asking for water over and over when we have run out”

Is the concept of AI running scenarios and providing input to the flying process bad? Not to my mind. I love the basic features in my car that watch the road and lets me know when I am not traveling correctly. And the GPS saying “you missed your turn” is a lovely thing.

But having override capabilities and a capable pilot/driver that knows the right answer will never be a bad idea. Like when the GPS tries to take you down Transylvania Drive at three AM for a “shortcut”. Um, no thank you, I will just go ahead and stay on Elm Street where it is safe.

Medicine

Hospital: At Merciless Hospital, we are realizing more and more we can use AI technology to handle the actual medical decisions, so our focus is on medical staff that is sympathetic to your needs and will treat you will respect.

A patient on a TV show comes into the ER with a lot of pain, so he had been sedated. We asked the patient “where does it hurt?” And he screamed that it was in his right leg.

The AI doctor scans the patient and says, “The wound you are showing me says that this person’s leg is very infected and needs to be removed immediately or they will likely die in the next hour or two”

Doctor to patient he has just reviewed after robotic surgery (driven by ACME Medical Systems), “Your right leg was removed to take care of the gangrene you had contracted. We are so sorry but we will take great care of you. Is there anything I can do to make your stay more comfortable?”

Patent, with a look on their face of true confusion, “what to do you mean right? I think I broke my ankle toe on my left foot. My right leg was fine. The infection is just makeup!”

Doctor to patient, “Sorry about that, you need to let your makeup artist know, because they are going to be really proud of the work they did. I will make sure you get extra pudding with lunch today to help you feel better.”

Are LLMs and medicine good partners? Very much so. I regularly ask an LLM to review my symptoms or even doctors’ orders. It can dispassionately look over that documentation and years of other results and offer suggestions. This is an amazing use of this technology and most of the time your doctor will be right. Sometimes they missed something, or the AI notes something about your individual case so rare that only a TV doctor in the last 10 minutes of a show would recognize and save your life James Bond style.

In a similar way , I was at the car dealer and they said something that sounded really wrong to me. So I took a screenshot from my mechanic’s video and asked to see what it would diagnose from the picture. It gave me the same diagnosis. Car repair isn’t life or death, but how often do we think “Are they right? Are they telling the truth? Are they just trying to make a buck?”

Law

This one is way too overdone, but scarier than finding your leg amputated by accident. Go read/watch Judge Dredd or even check out Robot Santa from Futurama. Santa is the funniest (unless you start to think of it as a possible reality) because he calls the people charging protection money as naughty along with the people failing to pay their protection money.

Two minds are better than one

This kind of especially goes for an AI “mind” because it can read a lot more than any human very fast. AI makes probabilistic guesses based on that information and gives you a possible answer. Ask it again, you may get a different answer. Call it wrong when it is right, and it might agree with you.

And it will sound confident every time

As computer programmers and/or engineers of any kind, we basically take jobs away from one group of people and shift them to different roles. The thing that AI has that humans don’t is speed. Point it at your organization’s files and it can dig through them faster than a huge group of humans could. It is probably going to tell you a lot of things that could be action areas.

Without an expert to validate a computer’s output, one day your new AI sql programmer (call this entity AI-1) may be adding code to funnel off money from your accounts, and paying off AI-2 and AI-3 to say the code does what it is supposed to do.

Summary

AI is a great tool, but don’t let it do everything. Even the programmers who build AI tools don’t know what it will output because it bases its “guesses” on what it has read. And much like our textbooks from school, there is no telling if students are reading the true truth, the sanitized truth, or a false narrative that completely distorts the truth.

So it is important to use them as inputs into the solution, not just find people who can ask the right questions and accept the answers.

But at least it will solve your problems very efficiently and cheap, right?