What AI Is Missing

There are instances in human history which we look back on and remember as a pivot point. A few examples:

• The printing press

• The telephone

• The industrial revolution

• The advent of personal computing

• The invention of the internet

AI will be remembered in much the same way as these events. And it may be even more significant in the long run.

You see “AI” everywhere. It’s in the news, in your social feeds, and it’s changing the way we work. It feels like every month a new cutting-edge model is released. And each new release drives a new wave of reactions ranging from excitement and hope to fear and despair. All of this has culminated in a rather disorientating landscape.

I consider myself to be “a creative” (graphic designer, web dev, photographer). The aforementioned “landscape” gives me the sensation of riding a roller-coaster. Highs of “This thing is so powerful! It’s so cool what it can do!” sharply contrasted with lows of “Now I feel like my skills are less relevant”.

While utilizing the tools every day, I have realized a fact that has set my mind (mostly) at ease: AI has no vision.

I don’t mean vision in the literal sense (AI is good at image processing). I mean it has no vision in a creative sense. In the programming world (and adjacent), there is a viral idea called the “one shot”.  A “one shot” occurs when an AI model, given a single prompt, accomplishes the goal of the prompt on one session. But these demonstrations of “capability” also reveal the same fact every single time: AI cannot create something new on its own – it requires a human with a vision sitting behind the wheel, steering it to the end goal.

This becomes more obvious when you scroll through a social feed like Linkedin. Hundreds of individuals/entities clamoring for attention by showing off the thing that they “built”. And, to their peers, the thing probably seems impressive. But, if you take a step back and look these examples, you will start to realize that they all… look and feel the same. And most importantly, none of them solve a problem that wasn’t already solved.

These models are insanely capable. Wired to the correct integrations/harness, they can accomplish amazing things. But they have no way of aiding a user with a problem when the user doesn’t understand the problem in the first place. They have no way of implementing a system which requires a vision of what that system needs to be. They need a visionary behind them with a goal, and they need to be steered in the right direction to meet that goal. Without the visionary, they churn out mediocre sameness to solve problems which don’t need solving.

Maybe I’m proven wrong, someday. Maybe the models get harnesses that allow them to collect a problem's context, and they become much better at understanding the human element that will interact with the solution. Maybe they become capable of vision.

But I doubt it. I believe that no matter how advanced AI models become, the output of an expert human in the driver's seat will always be superior (even if the model output is impressive on its own).

Do you agree? Disagree? Leave a comment below – I would love to hear your thoughts.