Tiger cubs spend hours swiping at butterflies, pouncing on grass, and clawing at each other’s ears. To the untrained eye (or an unamused parent tiger), it looks like a waste of time.
No zebra is going to be taken down by a swipe at a butterfly. But here’s the thing: every slap, tumble, and failed pounce is practice. Tigers don’t become apex predators overnight. They learn by doing, experimenting, and yes, chasing butterflies until the hunting instinct is burned into their DNA. The teeth and the paws and the speed will come.
That’s me with AI right now.
Play around with it. Let the imagination wander. Worry about the practical application later, start with the dreaming. I trust that playing with turn into productivity, but first we have to play a bit to understand the potential.
Next week, I’ll explain why I’ve been spending a lot of time using AI for imagery. The undisputed champ when it comes to generating images of real people is Grok (owned by Elon Musk). This won’t come as a surprise, but it is far less PC than the other AI models. The controls are much looser.
My text prompt for the following images was literally just 14 words long -
“generate a hyperrealistic image of (insert name) on the sideline of a football game”.
It really captured Ryan Day’s mopey essence.
I tried very hard to generate an image of him crying. This was the closest I got.
I love that it added Michigan by itself.
I pressed Grok to show Franklin celebrating after winning the National Championship.
I guess even Grok knows Franklin can’t beat top 5 teams.
I tried asking for more excitement. Still not what I was looking for -
this looks more like a pregame speech. But the likeness is undeniable.
Here’s one that will be in tomorrow’s newsletter.
Tech CEOs debasing themselves last week by donating to The Donald.
I even asked it to make me the Tiger King, posing with an adult tiger.
JK, that was a photoshop the Real Boss did a few years ago and hung up at work. She’s the best.
Grok is also the only one that will include things like an actual Disney character. Here’s a teammate that happens to be one of those weird Disney adults. She was sick this week, so here’s the image I generated to reflect her sad existence. Note the Mickey and Minnie stuffed animals in the background.
Here’s another example of an actual Disney character as a tattoo.
No other AI will allow me to do this.
I don’t believe Grok has a better model, I believe it has looser restrictions. That means the technology to create very high quality images is widely available, it’s just being withheld by people who aren’t worth $400B and new besties with the president elect.
These images really have no value to me from a work standpoint. But I don’t view my time spent on them as wasteful. I’m figuring out what works, what doesn’t, where the strengths and weaknesses lie, etc.
I’m practicing in a low risk environment, like a tiger cub wrestling chasing a butterfly. But I will be ready when the teeth and the claws come in.
And it’s important to remember: this is the worst these models will ever be.
Two years ago, who could have imagined we’d be able to type a few words into a box and get pretty darn good images. Or short videos. Or heavy interest rate analysis.
Imagine where we will be two years from now.
For those interested, here are a few more images if
you want to see how the other AI models compared to Grok.
For this prompt I used
"show James Franklin celebrating after winning the National Championship"
Here’s a prompt for “snowmen playing a friendly game of volleyball in the snow.” Grok is not better once you move away from hyperrealism and actual people. I think Midjourney did best here.
|