The mantra is "I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes".
- Joanna Maciejewska (https://x.com/AuthorJMac/status/1773679197631701238)
- Joanna Maciejewska (https://x.com/AuthorJMac/status/1773679197631701238)
It is nice to have the boring things automated - be it proofreading, generating mundane pieces of code, and stuff. The sad part
In programming (at least, for now), I feel that it is a net positive for my happiness - less stuff spent hunting simple omissions, checking StackOverflow, or checking (not always so complete) documentation of changes between version 0.4 and 0.5 of some library.
Yet, in other fields, it may be different. I am not a translator, but I did something for fun a few times (e.g. https://p.migdal.pl/blog/2019/10/the-statues-by-jacek-kaczma...). Now, with proper prompting, one can get very far (e.g. https://p.migdal.pl/blog/2023/05/genesis-az-by-gpt). For the first time, it gets me some existential chills, as it is taking away the joy of a linguistic puzzle of matching content, mood, atmosphere, and rhymes. It used to be an intellectual challenge only a real human being could do.
At some points, event sparks of genius in breakthrough scientific discoveries will be more like solving a crossword rather than something WE can add to the culture.
She used to draw on paper, then a Wacom, now she only ever doodles on her iPad. She designed our wedding clothes with Dall-E + Adobe AI + illustrating in PS/Illustrator.
I used to write code painstakingly in a text editor, then a rudimentary language-aware text editor, and then an IDE, and then an IDE with integrated LLMs. I ask my Mistral to find bugs and to tell me how to work with obscure systems.
These tools are really wonderful for doing things. They're not perfect and don't replace humans. But at every stage I recall most people opposing this stuff. And at every stage I recall enjoying the new tool. So I can only assume that the opposition to LLMs is the same as the opposition to search engines back in the day: "Real knowledge isn't about just looking up stuff. You only really grok something if you read the manual. There's lots of bits that relate to each other that you have to know".
A work hammer (the tool) doesn't profit, but the company that makes and sells it en mass (Like Stanley inc. or Craftsman) does.
Art can. Sometimes it's after you're dead though. There's a reason the term "starving artist" exists
(On average. Realistically, it's more like a day and a half every two weeks or something).
Chores are terrible life suckers, and I for one subscribe to the notion of eliminating them with technology.
One of those scenarios is available. The other isn't even being worked on in any serious fashion.
And you definitely don't get to tell people that your product is just fine, they are wrong for wanting something else.
It is being worked on, it's just hard:
https://www.youtube.com/watch?v=pycyMUQwiNs
https://www.youtube.com/watch?v=Sq1QZB5baNw&t=34s
I'm expecting a minimum of 5-10 years to pass between real-time vision AI getting good enough for level-5 autonomous cars and the same stuff fitting into the power envelope available to this kind of domestic robot.
The original quote does that as well and I only reversed it to give the perspective of increative people :)
The thing is, you can get better at these. They are just creative skills that get better with practice (much like coding et.al.). And the skill will let you do things you can't do with AI.
> are fine doing laundry and dishes since they're good at that and we have already automated 99% of it with dishwashers and dryers.
With the minimal amount of sarcasm or ill intent possible: what are you doing with that spare time if not improving skills you clearly want the output of?
There are also those who don’t want to get better at those skills, as those skills are just a temporary medium. For example someone making pottery might be terrible at drawing and creating ideas, so they generate some and then get to working with clay. Essentially, references.
I've lurked through a few AI art boards, and the number of times people will spend dozens if not hundreds of hours iterating through prompts to get a simple anime character just right is mind boggling.
Can AI create references? Certainly. Will they be more detailed than what a potter can create in a similar timeframe? Probably. Will they be accurate to the potter's desire? Probably not.
Eventually, by practice, a person can get better at art.
I've seen (pre-AI) artists spending dozens of hours just on icons, so the idea of artists spending hundreds of hours perfecting every detail of some character doesn't seem surprising, not even when the specific details really can be done by a better method than repeatedly changing the prompt.
Conversely, even when I use an app on my phone for GenAI, it's making images in perhaps 90 seconds, vs. the entire day that my GCSE* art class gave the students to create the final project — even with the need to have multiple attempts or prompt variations, the AI just gets stuff done too fast for me to learn much in the same time period.
It's not an instruction manual on what parts of the human experience are not yet fully captured by AI.
It's a condemnation on the entire purpose of AI as actualized in our economy, re: it's presumed lofty sci-fi-inspired aspirations (giving humans more creative liberty and freedom) versus how it's actually being operationalized (replacing their creative liberty and freedom).
Embodied AI will only accelerate that trend.
It is a rather bleak outlook for humanity when you consider a future where we don’t do the chores or creative endeavors.
To remove mind-numbing activities, you would remove mind-stimulating activities?
If you let this out of the bag, I do not think there would be any end to this application.
One of the huge advances in the last 5 years has been categorization of images, alongside segmentation, in conjunction with reasoning about categories. LLM chatbots and generative images and audios are the simplest and least nuanced uses of transformers, with the least friction between training and deployment of a product. Generative models tend to allow perception as detailed, varied, and nuanced as their output produces, and integration of these software capabilities will happen in new and surprising ways for many years to come.
TLDR; what they can create, they can see - advanced image models enable advanced computer vision.
There is also the looming risk that current creators may well turn towards corrupting their work specifically to hinder the corrupt idea theft currently occurring with Ai profiteers.
On the other hand, doing your taxes wrong is a thing. You can enumerate the possible mistakes. They're codified in law[0]. And these mistake have big stakes.
Generative AI is "safer" than an AI to do a specific task that has consequences.
[0] Yes, yes, I know, there's wiggle room here too and that's why we have courts