Comment by bufferoverflow | Hacker News Reader

Comment by bufferoverflow | Hacker News Reader

I used to think creative jobs (arts, writing), programming and science will be last to automate, but this year completely destroyed my confidence in that.

Stable diffusion, DallE, Midjourney produce absolutely beautiful paintings, drawings, renders, photos of sculptures. We have AIs that write movie scripts, book plots. We are starting to get AIs that are okay at making music. We have transformer models that generate textured 3D models.

On the programming side we have Copilot, which can't yet replace a programmer, but it's definitely a first confident step.

On the science side we have AI tools that fold proteins, predict material properties of novel materials, the wet lab work is getting automated too.

I find it hard to name an area of human work that can't be automated. And not even work, any human activity.

It looks like all the problems that can be solved in software, will be solved first. Robotics will be lagging, but not very far behind. Once we perfect artificial muscles, it's game over. The rest is software and scaling everything up.

I feel like this take is a lot like saying that the invention of a car is going to obsolete the profession of carriage drivers. The job of a carriage driver isn't to drive a carriage, it's to drive a vehicle. It's the same with all of these new technologies you've listed: they alter the job description, but don't remove the job.

Yeah. Look at the millions of words written on tweaking prompts.

And while SD is very impressive, it’s going to be a long time before it’s producing high res imagery, at least without ready access to lots of data center grade hardware)

There are two opposing forces that determine the net change

- Greater productivity. A single carriage driver can now drive a bus, so fewer carriages are needed

- Lower costs/ higher demand. Everyone who walked can now afford the bus, so the number of drivers goes up even if fewer drivers are needed per traveler.

Except we are developing artificial minds that are better than humans at particular tasks.

That's a bit different than inventing a car (which at that time didn't even promise self-driving, so your analogy is doubly bad).

Mundane writing jobs might be automated soon, perhaps. But truly creative jobs?

Do you honestly, really would want to read a 1000 novel written by an AI even if I assure you that is as good as what an average human could do?

Now we are all wowed with Stable Diffusion and what not, ok. But give me a library stand with 999 decent AI books and 1 human book, I pick the human.

I want to talk to the human behind the pages, feel the human behind the pages. I don't want a random number very efficiently generated by a machine. Who cares!

"Who cares!" Most people ?

Its the content of the book that matters, not how or who made it.

The human behind the pages is a nessecery evil to get to the book, same as food or furniture.

If something is entertaining enough to capture people’s attention, and is also infinitely abundant, it will probably undermine the market for higher quality, higher cost alternatives. AI doesn’t have to be as good as traditional media to destroy it - it’s more likely to be competing with Tik-Tok for attention.

But in the long run it probably _will_ be as good as traditional media, and instantly customisable to your tastes to boot. I suspect it will be a difficult pill to swallow, but we are likely over-estimating the depth and complexity of our souls, just like every other AI task. Some people will still insist on human-authored works and swear they can tell the difference. Money will still be made putting humans on stage to perform. Older generations will insist everything is worse and will possibly be right. I’m sure there will be several billion deprecatingly self-referential AI retellings of Prometheus and Pandora to show us what a bad idea it all was.

You can confidently bet that James Patterson's team is already using GPT3. For better or worse, he has been the Henry Ford of novel production and modern technology like GPT3 is exactly the force multiplier the team would want to adopt.

Plot twist: In the future we will can talk to the computer behind the pages...

Are you confident that you will be able to identify a book written by a human among 999 written by AI? Give it a few years, and you won't be able to.

So I am also very torn about this. A big part of art is actually knowing the "history" of the author? (Just for this I am not quite lumping coding into Art just yet as I still code to solve hard problems - bit like fixing up a car - than having to impress others).

I am big into Carnatic music (a form of south indian classical music with a ton of intricacies and improvisation). The actual songs that most performers perform on stage had actualy been written by a handful of folks 3-5 centuries ago. And without any judgement, the style has a lot of homogenity to it and several folks have created compositions (in the past 50-100 years) which mimic a lot of the original style and yet acceptance of them in public performances is limited ("oh its another X wanna be").

Now imagine the same if created by AI.

Reply on news.ycombinator.com