I've been emersing myself in AI lately trying to grok it all. Back in time a million years ago, when the Earth was young and dinosaurs walked land, I has started out wanting to focus on AI in Uni. The promise of AI has always fascinating to me since the first program I wrote on a TI 99-4a (had to save it to cassette tape with a tape-recorder!)
But back then AI sucked. It wasn't AI. It was Expert Systems written in LISP. Belch. There was nothing inspiring about that. That was a phone menuing system with delusions of grandeur. So I became disillusioned and dropped it and just moved on to generalistic programming to get an industry job. Now I feel nothing but excitement about the AI developments I'm seeing.
It's not perfect now and that causes too many people to pooh-pooh it. What I see is good bones. I see an approach we are just at the very beginning of. You are seeing a toddler now. An infant stumbling on wobbly legs. So sometimes it's easy to underestimate now. There will be many who suffer because they underestimate the tektonic shift.
GPT has many powerful capabilities. But I am starting to get a feel for what it is best for. It is good with words. Planning. Design, documentation, high-level designing out and tech stack choice. Good for back and forth Socratic concept exploration.
I've been a lot more satisfied with Github Copilot in coding. Even when using the GPT 5 model. They must have wrapped better code optimization into it. When carefully constrained and seeded, it can write better code than me. In fact in a lot of my coding now I have the Copilot maximized me than the code window. If I have to go get my hands dirty on code, I either didn't spec it right or the model isn't mature i that case yet. But i is getting better so fast.
That is the thing that is going to kill people. People don't intuitively grasp exponential growth. It is not intuitive for Humans.IT always catches most people off guard. Having been following it for a year or so, what I see is the rate of change. A year ago, AI was crating hands with 8 fingers. Now it can make indistinguishable images.
A year ago AI voice was crap. Now it's merely not perfect. A lot of people like the hate on AI voice, but that is irrelevant. That is the way it is going and by next year you won't be able to tell human from AI voice and you will be able to dial in a Morgan Freeman voice with a touch of Hal Holbrock.
3d modeling is painful now but it is improving exponentially. Exponential growth is the important factor. By the time you as a human can get you head around it, if you start NOW you might learn fast enough to be ready to just on the back of the dragon and ride once it reached adolescence. This is going to move too fast to take a normal learn methodically and take your time. The train is already starting to move to fast to just step up on. You need to get a running start and just jump on and hang on for dear life and go along with for ride.
3D creation is the next big step. AI voice is already almost getting there.
ChatGPT can do image creation well, but it's 3d capabilities make me want to cry.
There are other dark horses out there to watch. There is a huge amount of money in solving this problem for the game and film industry. The problem WILL be solved. Do not doubt that. And faster than you might think. I image by next year it will be common.
Right now the best I've see is a model called Sparc3d.
I tried for hours to get GPT to make me a propellor. It was painful to see. He almost understood but just could work it out yet.
Sparc3D took a long time processing but nailed it on first try. I was very impressed. IT EVEN BUILT THE SIDE THAT WASN'T IN THE IMAGE. Grok that.
All the way to Blender no problem:
There will be others coming. But right now, that is the one I am watching closely.
https://hitem3d.ai/?sp_source=huggingface/Then something like Substance Painter to put the lipstick on:
I saw a great quote: "AI won't replace human workers. It will only replace human workers who can't implement AI into their workflow."
