The quip / aphorism is a commentary on that a few years ago, AI was about being on the cutting edge - the things that were going to be seen next year.
The work of AI researchers were about envisioning the future of what could be.
Well, now what we've got is something that consumes everything and outputs aesthetically pleasing means of the content that it has consumed.
However, that "consumes everything" means that its always reactive and working from data that is in the past.
There is no creativity in it. This doesn't mean that it won't inspire a human, and it does employ a lot of fancy tricks too (I'm seriously impressed by its ability to create limericks of made up words with a coherent definition).
However, the AI can never be "ahead" of the curve because it can only produce what it has consumed.
This isn't to say a lot of the work that we do in software development isn't doing what has been done again and again and again (see also CRUD), but the moment you're given a novel problem or need to maintain an existing solution and encounter a new bug, the "can an AI do it" rapidly fails at those tasks as it lacks the domain knowledge necessary.
> ... Anyway, Matrioshka brains, the end product of a technological singularity – they're bandwidth-limited. Sooner or later the posthuman descendants evolve Economics 2.0, or 3.0, or something else and it, uh, eats the original conscious instigators. Or uses them as currency or something. The end result we found is a howling wilderness of degenerate data, fractally compressed, postconscious processes running slower and slower as they trade storage space for processing power. ...
(the part that reminded me of it is the GP post's comment... and the intro:
> Manfred pauses in mid stride, narrowly avoids being mown down by a GPS-guided roller blader. This is getting weird enough to trip his weird-out meter, and that takes some doing. Manfred's whole life is lived on the bleeding edge of strangeness, fifteen minutes into everyone else's future, and he's normally in complete control – but at times like this he gets a frisson of fear, a sense that he might just have missed the correct turn on reality's approach road. "Uh, I'm not sure I got that. Let me get this straight, you claim to be some kind of AI, working for KGB dot RU, and you're afraid of a copyright infringement lawsuit over your translator semiotics?"
)
It takes an interesting path of the acceleration (note that I don't agree with the essence of the slope but instead believe we are seeing the product of a lot of sigmoid curves with the slope near a maximum - but as the easy stuff gets done, the (heh) second derivative gets negative as harder and harder problems need to be solved https://towardsdatascience.com/derivative-of-the-sigmoid-fun... ). We are seeing the seeds of ideas sown decades ago sprouting in rapid succession - and get to the maximum where things are neat and everyone is going "whee!" but the next step of maintaining that growth will take time and much more effort than sowing the ideas.
Lots of people are working on the first half of the sigmoid curve... but people lose interest or resources when it comes to the second half and so you get a lot of things that are abandoned at or near where things started getting hard.
The work of AI researchers were about envisioning the future of what could be.
Well, now what we've got is something that consumes everything and outputs aesthetically pleasing means of the content that it has consumed.
However, that "consumes everything" means that its always reactive and working from data that is in the past.
There is no creativity in it. This doesn't mean that it won't inspire a human, and it does employ a lot of fancy tricks too (I'm seriously impressed by its ability to create limericks of made up words with a coherent definition).
However, the AI can never be "ahead" of the curve because it can only produce what it has consumed.
This isn't to say a lot of the work that we do in software development isn't doing what has been done again and again and again (see also CRUD), but the moment you're given a novel problem or need to maintain an existing solution and encounter a new bug, the "can an AI do it" rapidly fails at those tasks as it lacks the domain knowledge necessary.
-----
Btw, I'm going to suggest giving Accelerando a read ( https://www.antipope.org/charlie/blog-static/fiction/acceler... ). It has an interesting take on it...
> ... Anyway, Matrioshka brains, the end product of a technological singularity – they're bandwidth-limited. Sooner or later the posthuman descendants evolve Economics 2.0, or 3.0, or something else and it, uh, eats the original conscious instigators. Or uses them as currency or something. The end result we found is a howling wilderness of degenerate data, fractally compressed, postconscious processes running slower and slower as they trade storage space for processing power. ...
(the part that reminded me of it is the GP post's comment... and the intro:
> Manfred pauses in mid stride, narrowly avoids being mown down by a GPS-guided roller blader. This is getting weird enough to trip his weird-out meter, and that takes some doing. Manfred's whole life is lived on the bleeding edge of strangeness, fifteen minutes into everyone else's future, and he's normally in complete control – but at times like this he gets a frisson of fear, a sense that he might just have missed the correct turn on reality's approach road. "Uh, I'm not sure I got that. Let me get this straight, you claim to be some kind of AI, working for KGB dot RU, and you're afraid of a copyright infringement lawsuit over your translator semiotics?"
)
It takes an interesting path of the acceleration (note that I don't agree with the essence of the slope but instead believe we are seeing the product of a lot of sigmoid curves with the slope near a maximum - but as the easy stuff gets done, the (heh) second derivative gets negative as harder and harder problems need to be solved https://towardsdatascience.com/derivative-of-the-sigmoid-fun... ). We are seeing the seeds of ideas sown decades ago sprouting in rapid succession - and get to the maximum where things are neat and everyone is going "whee!" but the next step of maintaining that growth will take time and much more effort than sowing the ideas.
Lots of people are working on the first half of the sigmoid curve... but people lose interest or resources when it comes to the second half and so you get a lot of things that are abandoned at or near where things started getting hard.