You’re using today’s AI wrong

To be clear, I don’t mean how your business might use complex machine learning models to correctly flag mistakes in large datasets; I’m talking about the use of models like GPT or MidJourney right now. When I look at the type of examples I see all over LinkedIn or TikTok, I see the same approach over and over again, with very few people realizing the real utility of these models.

Let’s take GPT as an example. Sure, you can use it as you would a slightly more advanced version of Alexa or Siri, but if the only reason you use it is to be told that nine plus ten equals twenty-one, then you are wasting the potential that is brought on by the massive dataset it is trained on.

I can understand why people hear it is a “language model” and expect to use it to write their copy, create original content, generate ideas, and find topics to write about — all the things you might first think of when you describe something as a “language model” — but in reality, there is a lot more to utilize here.

A model like GPT is so innovative because of the scale of data it has been trained on. Think of it like an AI that has read and learned from a massive amount of the internet. Did someone write an article about how to best pick your haircut? GPT has studied it! Is there a guide on Medium teaching you how to color footage in Final Cut? GPT has studied it! Published a research paper on what makes for an effective brand name? GPT has studied it! This is where the power of the language model lies, and the same applies to MidJourney for visual content.

Less than two weeks ago, I started working on a new business, and despite having had the chance to use GPT3 since 2020, I completely revised my approach to the best way to use a model like this.

Imagine you need to design a logo, as many founders have at some point. Normally, you spend way too much time thinking about colors, style, layouts, and a million other things. Alternatively, you could go on Fiverr, find someone whose style you like, and explain a bit about your business.

I was very close to doing that this weekend, until I decided to run a little experiment. I asked GPT about it and saw what it suggested, then used some of those foundations to get MidJourney to give me some visual inspiration. By writing out almost exactly the kind of things I would have to tell the designer on Fiverr anyway, ChatGPT quickly came up with suggestions. After only a couple minutes of back and forth, I got exactly what I was looking for. Taking that and throwing it into MidJourney, I ended up with drafts of designs which I would be more than happy to receive from a graphic designer I was paying.

The last step did require me to remake the designs in Illustrator so it was a vector graphic and included the name and tagline (which GPT came up with, too). But what would otherwise take hours of my time and at least two days of back and forth with a designer, I had a solution for in less than 45 minutes.

This is when it hit me: I had been thinking about these AI solutions all wrong. They aren’t simply just a helpful tool to accelerate the process; to some extent, they are like actual employees. I remembered how, in my previous business Fair (which I shut down last year in June), when we were talking about redesigning the logo, it involved three people, took several days, and many many man-hours. Meanwhile here I was, with a design I was quite happy with, in less than 45 minutes later.

That’s the real innovation with AI like GPT or MidJourney: it’s no longer just a helpful tool; when used correctly it’s almost like an actual employee.