The AI that will “replace” the most jobs, by making a smaller number of people capable of doing more tasks, will very likely not be general. It will be hyper specific, like an AI program that aids with drafting legal briefs, or with doing a company’s books.
And the AI won’t be doing it all by itself; it will be a tool to make lawyers or accountants more effective.
You make more money by having each worker be way more productive, not by “deleting humans”. Why would you “delete humans” rather than implying them, in coordination with AI, to be incredibly productive?
Because the market for most products is finite. It doesn’t matter if you can apply AI to your company and crank out 100x the product with the same number of people, if that’s more product than everybody needs. More likely you exploit that same productivity factor to make the same amount of product (or a little more), with fewer staff.
GenAI is notably NOT deterministic in that fashion. Here’s one bit from Forbes:
One of the inherent characteristics of GenAI is its non-determinism. This means the AI can produce different outputs even when given the same input multiple times, leading to unpredictability in its results.
Of course being non-deterministic is not being creative. But still.
I would not rely on Forbes to accurately report computer science theory. This is utter bullshit. Generative AI is entirely 100% provably deterministic. As a user it may appear non-deterministic but that’s just because the pseudorandom seeds will be different for every run, but it’s no different to video game that uses pseudorandom numbers to decide the layout of the level. Just because every time you open the game the level looks different it doesn’t make it non-deterministic, if you ran with the same inputs including random seeds you’ll get the same results.
In fact I bet any GenAI software has a mode where the random seeds are set to predictable values, so the automated tests can run and get the same results (that’s what I’ve done when I have written software that uses stochastic techniques)
Yes and also Generative AI is also a perpetual motion machine and if you spend on the premium ultra-secret AI fairies it will produce anti gravity! Let me look up some blogs that will prove this to you!
It’s basic computer science. Generative AI is not magic. Generative AI is a computer program and as such entirely, provable deterministic. Just like any software that uses pseudorandom number it may appear non-deterministic to its users. But claiming it’s non-deterministic is as believable as claiming it’s a perpetual motion machine.
And when the Industrial Revolution ran up against the limits of the consumption of medieval peasants, what did it do? It gave those peasants factory jobs so they could consume products in quantities and sophistication unimaginable to pre-industrial societies.
Even if you went back in time with all the knowledge you have today, and gained full control of the Kingdom of France during the middle ages, you’d be hard pressed to mobilize enough industry before you die of old age to manufacture a microwave oven, or a television, or a refrigerator. Or a cell phone. And now we churn these things out in such quantity that even extremely poor people can afford them.
You are conflating the meanings of determinism in the contexts. The brain is deterministic in that each input leads to output in ways that are controlled by the rules of nature. It is also nonlinear and massive enough that its outputs are probabilistic not deterministic. Same is true of gen AI. Of course simple systems can be chaotic too: a sphere placed on the top of a smooth dome. What path does it take? Where does it come to rest? All controlled by basic rules and all only able to be predicted probabilistically.
So you mean it’s entirely and profoundly non-deterministic either in theory (as the rules of nature are not deterministic as far as we know) or in practice (as it is completely impossible to produce an experienent where identical inputs lead a human brain to produce the identical outputs. Hell we don’t even know enough to know what “the identical inputs and outputs” would mean for a brain)
So totally different to a Generative AI program which is completely deterministic both in theory and practice. It is 100% guaranteed that if you run that program twice with the same inputs it will produce exactly the same 0s and 1s. Not just in theory the companies that make Generative AI software probably run it this way in their test suite.
If a business blog tells you otherwise they are just flat out wrong .
How is that going to work this time around? Please speculate what the next tranche of jobs might be, that we’re going to give to all of the people displaced by having their job automated out of existence, which jobs themselves will be safe from being automated out of existence?
It could work if we decided that humans actually just have a lot of paid leisure time now - some sort of universal basic income that you can spend on consuming the things that the machines all make now instead of the people, but where is the will to implement that going to come from?
You’re doing the equivalent of asking a random dude on the street in 1698 what the impact of the upcoming industrial revolution will be. Or asking a hunter-gatherer what humans will do with their free time once they invent agriculture and only 75% of them have to be dedicated to food production on a full time basis.
I have no idea. But what I do know is human nature. Human labor is a resource; we will find a way to make productive use of it.
I wish I shared your cheery confidence that everything will be OK this time, because it was sort of OK in the past. Not everything scales forever; there are such things as limits and breaking points.
Because how AI is “inspired” is how humans do. They listen and notice patterns either consciously or unconsciously and mimic them. Hence your What’s the Story Morning Glory. You nitpick over whether someone asked Oasis to do a Beatles-like song vs the did themselves because they felt like vs someone asking AI to do the same thing. You have a helluva of a “human is superior” ego to credit the one over the other.
I was going to say that a lot of the pro-AI fanboy behavior sounds like pure unbridled techbro glee at the possibility of taking down a source of expertise that’s currently scarce, and have no way to control.
See also - the excitement over replacing the dollar with crypto, with replacing women with sex-bots or AI-powered fleshlights, etc. Just another inconvenient human interaction to be liquidated, automated, and commoditized.
Why on Earth would I not want to “take down” something that I once had to pay someone else to do but can now do for free? Really? If I need to edit a video I do it myself with video editing software instead of paying an editor. If I need to create a document I do it myself with writing/publishing software instead of paying a typsetter. If I want to create and edit photos I do it myself with digital cameras and photo editing software instead of paying a photographer. Most people are at that stage, and have been for decades. So if I want to create a piece of “artwork”, why the holy hell shouldn’t I like having a software tool so that I can do it myself and not have to pay someone? I welcome any and every convenience that allows me to do something myself easily and for free.
Maybe you were thinking of something else but I’d question whether commercial artistic talent is especially scarce or outside corporate control. Even prior to AI image generation, you could go on Reddit or Fiverr and get someone to knock out good quality work for under a hundred bucks without issue. Sure, this isn’t great for the guy hoping to illustrate some Kickstarter’s collectible card game but it’s not as though illustrative artists are a resource we were lacking in. Go to DeviantArt, swing a piece of Bristol board, and you’ll hit ten people dying to take your commission.
I’m happily married and don’t foresee “sex bots” happening in my remaining years but, versus the current commoditizing of sex with human trafficking and prostitution, this sounds like an improvement on the status quo. Also, I don’t want to alarm you but I’ve heard reports of women using strange “robot penis” devices for their pleasure as well. Some with AI and Blu-tooth controls.
So again how does the dumb series of assembly instructions that make up the AI model do these human things of “being inspired”, “listening” and “noticing”? And more importantly why does my series of dumb assembly instructions, that make up my audio mixing software, not have those attributes?
They are both just dumb computer programs that take some copyrighted work, encoded as binary data and spit out out ls and 0s in a predictable manner based on some parameters. What precisely is the difference? How many loads and stores would I have to add before my software gets to be inspired?
Perfect, great example, I’ll modify it a touch to relate to written word vs music so it’s a bit more concrete.
Program A, you input a line of text, such as “The man in black fled across the desert, and the Gunslinger followed” and it spits out “The man in black fled across the desert, and the Gunslinger followed” That’s a copy.
Program B, you input a line of text, such as “The man in black fled across the desert, and the Gunslinger followed” which is converted to hexadecimal, input as the seed to a pseudo random number generator, reconverted back to text which then spits out “dkfnjd iziret dign aidna golind idhasa dguinty sen, wei.” That’s NOT a copy. Even though it returns the exact same gibberish when you input the same real text. There is no world where it makes sense for Stephen King to get pissed off about me profiting from this line of gibberish. That gibberish is NOT HIS WORK, it is “inspired by” his work, his work is literally the seed from which my work is created, but it isn’t a copy by any but the most ludicrous stretch of the definition.
I’ve been trying to follow your chain of reasoning, and this part is really confusing to me.
How do you think mixing software works? You mentioned it upthread, and I’ve reread it several times, but I’m just not able to understand this “magical mixing software” metaphor. Would you mind explaining it to me? Thanks in advance.
