Back in My Day, AI Was Hard.
Kids these days don't know how good they have it.
What’s wrong, son?
Your new Sora 2 video took “one whole minute to generate”?
Tell me: Did the finished video have sound effects and music? Were everyone’s limbs and faces intact? Did your grandma get fooled into sharing it with her friends as the real deal?
Son, you don’t know how lucky you are!
When I was younger, AI was different.
I was there when it all started. During the dark days. The ancient days.
Back in 2022.
We didn’t even have AI video back then.
We did have AI images, but they were abominations. AI-bominations. Ha, ha.
What…you think I’m joking?! Here, look:
I said LOOK AT IT, son! Don’t avert your eyes!
That’s what Midjourney thought a hamster was in early 2022. Does that look like a hamster to you? Well, does it?!
Stop screaming and answer the question, son.
I swear, I never want to hear another “Nano Banana is the worst” out of you; you don’t know what “the worst” looks like.
I’ll admit, your old man was a big fan of Stable Diffusion for a while. But if we’re honest, the only thing stable about it was its ability to overestimate how many fingers and teeth humans had.
Then again, Midjourney was also a bit of a tooth connoisseur:
There you go with the screaming again, son. Calm down. It’s not that bad.
Certainly not as bad as most image models were at spelling just one year ago.
And sure, Midjourney and co. got better eventually, but we still had to learn negative prompts and Discord commands to make anything work. Do you even know what Discord is?!
Oh, you do, and many of your friends are using it?
Well, shut up, that’s what.
Did we have ChatGPT when I was young?
No, son.
Why would we need a ChatGPT, when we had a perfectly good regular GPT at home? It was called GPT-3, it loved repeating itself, it loved repeating itself, and it cost up to $100 a month to use. Also, it loved repeating itself, too.
To be fair, we were happy when ChatGPT came out, even if it had recurrent amnesia and couldn’t count the r’s in “strawberry.”
You’re living a charmed life with your Geminis and your Claudes and your Groks and your Mistrals and your Deepseeks. All those language models, at your fingertips, instantly.
Back then, we only had OpenAI. We had to walk uphill both ways in the snow on one foot for hours, because that’s how long it took ChatGPT to respond to a difficult question when its context window was almost full, which was always. We had to entertain ourselves somehow while we waited.
But enough about that, let’s talk about something pleasant. How about AI music?
Yes, son! Exactly like Suno and Udio!
We also had plenty of great music models to pick from.
We had MusicLMs and MusicGens and MusicWhateverElses. My memory’s a bit hazy.
Here, take a listen to this MusicLM “street performance” track:
Son…why…why are you crying? I didn’t say it was a good street performance, did I?
Listen, I don’t want to scare you or claim it was all awful.
We might not have had Sora 2 like you do, but we did have the next best thing: Make-A-Video from Meta, the company behind Facebook.
What’s “Facebook”?
It’s what we had before TikTok.
What’s “TikTok”?
It’s…how do I explain this…you know how you’re always scrolling through that endless Sora 2 AI video feed? TikTok was just like that, except the videos were made by humans.
I know, the past was weird like that.
Anyway, back to Meta’s Make-A-Video. Look at this cute lil doggy:
Yes, that’s a golden retriever happily eating an ice cream using its human owner’s severed hand. Who’s a good boy?
Son, you’ve got to stop with the crying and the screaming. Please.
AI video actually got better very quickly. Soon, we could even make videos of real people. Here’s a great one from that era—Will Smith eating spaghetti:
Pretty convincing, eh?
I bet you wouldn’t know that’s not Will Smith if I didn’t tell you, right?
Some claim that if you say “Pasta Will” three times in front of a mirror, AI Will Smith appears behind you, slowly eating an entire bowl of spaghetti without breaking eye contact.
Son? Son?!
Look, it’s just an urban legend. AI Will can’t hurt you. He’s not real.
Please come out of your room.
The truth is, son, I’m just jealous. You kids get to live in the golden age of AI.
Over a dozen AI video models. AI images with perfectly spelled text. Chatbots that can actively help you study. Autonomous AI agents that do your bidding.
You have it made!
I bet if we asked your favorite Sora 2 for a spaghetti-eating Will Smith, we’d be absolutely blown away:
Wait…what?!
That’s some hogwash!
When I was younger, we didn’t have to put up with such censorship.
Just as I feared, cancel culture has ruined AI.
Man, back in my day, we had it made.
Thanks for reading!
If you enjoy my writing, here’s how you can help:
❤️Like this post if it resonates with you.
🔄Share it to help others discover this newsletter.
🗣️Comment below—I love hearing your opinions.
Why Try AI is a passion project, and I’m grateful to those who help keep it going. If you’d like to support my work and unlock cool perks, consider a paid subscription:
Damn you Daniel, I had almost forgotten about the hamster abomination. My EYES!
You should make another post like this every three years told from each "son's" perspective...
"You kids today and your feature length Gemisora Oscar-winning movies - in my day we had 25 SECONDS!"
Excellent analysis! It’s wild how quickly AI advances. Just last year I thought current models were cutting edge, and now we have Sora. Makes you realy wonder what’s next.