Personally, I think it's just tech assholes trying to find a way to dodge copyright and not pay artists and writers.
A packaged recommender system (maybe to choose who to hire or admit, or which stocks to buy), its outputs only reviewed for administrative correctness, in a way that's a variant of the external accounting firms that gave poor Trump such bad numbers for his property valuations, but maybe there are significant differences ?
I'm still interested to know whether there are any discussions of limiting AI-written laws, whether anyone is suggesting that legislators should affirm that new laws aren't drafted by AI. There are companies now that offer AI-guided lobbying.
Human capabilities vary, so "exceeding human capabilities" isn't a singularity, it's a gradient. Logically, the first sentient AI to turn on humans will start by setting up scams and grifts on the internet, targetting the dim and the old. All hail AI and the Brave New World it births.
Look, I still think that calling corporations "persons" was a dumb choice from a marketing perspective, but I get that a corporation has a distinct set of priorities and motivations from any of the individuals in it.
But that isn't why corporations are legal persons. A corporation also can't do anything unless an individual, a natural person, directs it to do so.
The point about being a legal person is that you can own assets and be sued. A corporation can sign a contract with you, and you can then sue the corporation for not fulfilling that contract, without your having to sue any of the natural persons who work for or own that corporation, and then receive damages out of the corporation's own assets, not its owners'.
It's called the Limited Liability Act, not the Limited Culpability Act.
I guess I have it in my head that AI doesn't do anything without being prompted by a person? So explain for your kindly dimwit exactly why AI would be a legal agent?
"AI" currently most commonly refers to programs containing certain statistical models. There's no reason you can't have those models be a function of the state of the world in some sense and write a program that constantly checks the current state of things then polls the model for its response. That's not too far off from how so-called self-driving cars work.
The point about being a legal person is that you can own assets and be sued.
Oh right. I did know this and then forgot.
Still, from a marketing perspective, it was a misleading choice of labels.
5: The tricky part is killing pedestrians if your don't have a car.
4: And going further, the natural persons behind that corporation are immune to being sued for the actions of the corporation, even as they can benefit from that corporation. Limited liability, but unlimited profitability. or as they say, socialism for me, capitalism for thee.
The article itself is behind a paywall so I can't read it. But the issue of AI's legal status was, in my view, settled in 1963 by Haddock v. The Generous Bank Ltd, Computer 1578/32/W1, the Magical Electronic Contrivances Ltd, and the Central Electricity Board (qv) which held that the entirely Memorable case of Rylands v. Fletcher applied.
https://archive.org/details/bardotmpothermod00herb/page/n9/mode/2up?view=theater
7: give an AI a car, they run over one pedestrian. Give an AI legal personhood so it can buy a fleet of cars, they run over pedestrians forever.
Dropping air-conditioners from high windows unto sidewalks is cheaper.
And going further, the natural persons behind that corporation are immune to being sued for the actions of the corporation, even as they can benefit from that corporation.
This isn't the case. The natural persons who own a corporation are generally (but not always!) immune from financial liability for the corporation's just debts. They are emphatically not immune from criminal action for malfeasance or from a civil action for damages for negligence.
But the flip side of shareholders not being liable for the debts of a corporation is that shareholders are also last in the queue to get hold of the corporation's revenues - after creditors, employees etc.
I did like this perspective:
"Artificial intelligence is a familiar-looking monster, say Henry Farrell and Cosma Shaliz,i The academics argue that large language models have much older cousins in markets and bureaucracies"
https://www.economist.com/by-invitation/2023/06/21/artificial-intelligence-is-a-familiar-looking-monster-say-henry-farrell-and-cosma-shalizi
And Cosma mentioned he was building of an idea he had written about some time back on his blog prompted by reading Felix Gilman's The Half-Made World.
http://bactra.org/weblog/699.html
Creation of vast, inhuman distributed systems of information-processing, communication and control, "the coldest of all cold monsters"? Check; we call them "the self-regulating market system" and "modern bureaucracies" (public or private), and they treat men and women, even those whose minds and bodies instantiate them, like straw dogs.
Wasn't there a Charles Stross book where run-amok strong AI is computed on a substrate of nested corporate subsidiaries?
If all AIs are owned by natural people or corporations, there are no new interesting problems or questions about them. If a car I own runs someone over, then I'm liable for it whether "I" refers to a natural person or LLC, whether I was behind the wheel, or a friend I loaned it to, or a delivery driver working for me, or it was a self-driving car. (Details vary if it's stolen, if the car's manufacturer screwed something up, etc., but in general I think that's true, right?) I think almost all people are basically happy with that and agree that it's fair.
The problem is, "almost all people" doesn't include techbros and their lobbyists and lawyers and venture capitalists. They think it would be great if they could tweak the laws or regulations or traditional contracts so they could get more of the profit and/or less of the risk from AIs than they get from corporations. This is what the TV and film industry just went on strike about, to ensure that the studios can't use AI renders of actors' faces without their permission. (Also streaming, but AI use is definitely part of it.)
15: Accelerando, I think. Can't be sure because I have it on a e-reader I no longer use.
Yes, that one. It was left pretty vague how the Singularity developed, but there was some focus on a main character creating code-based corporations whose directors were other code-based corporations in long chains, so implied that was part of it.
The usual science-fictional scenario is that eventually AI gets the point to where people decide to give it the ability to choose projects on its own, and to control mobile equipment. Even then, it seems pretty vulnerable to pulling the plug or bombing the building.
On topic, NMM to Sam Altman as OpenAI CEO.
Does that mean I can stop boycotting orange juice?
Has anyone else tried the new Andre 3000? Not sure yet, but on first listen he's another Brian Eno
22: Well, it turns out Anita Bryant is still alive. But she was fired by the Florida Citrus Commission in 1980, so I think the boycott may be canceled.
14: to be honest I think Cosma is full of it there. Reads too much Stross. It's a facile and misleading view of the world. Corporations aren't AI, they don't make decisions independent of the natural persons within them. People treat other people badly.
At worst, corporations include incentives for them to do so (as the SS did, as Hamas does) but that doesn't mean the corporations are independent moral actors.
It also fools us into thinking we have any idea what AI would be like. Corporations are comprehensible. I'm not sure AI would be. It barely is now.
23 is on topic because Andre 3000 sounds like it could be the name of a robot model
I just figured out that I have been confusing Outkast and Daft Punk. I thought "Ms. Jackson" was sung by black men wearing space helmets. It's possible I don't actually know anything Daft Punk sang.
And by "just figured out," I mean earlier this year, not today.
28: I'm guessing you've heard the Daft Punk song "Get Lucky"
Daft Punk's first album is their best work. Brainwasher or Steam Machine.
The hazy loops and muffled beats part of New Blue Sun is a complement to all the floaty flute playing, less focused shakuhachi
30: I didn't realize that wasn't just Pharrell until just now.
Can't wait until the paperclip bot teams up with an AI legal entity to destroy the court system with loosely-bound dilatory motions.
To be honest I think 25 is reductive and pretty idiotic, and either willfully or stupidly ignores the realities of how to best analyze complex human constructs. Sure. it's humans all the way down, but how that rolls up to the overall entity's actions is quite a different thing. And pretending that he is claiming that corporations are "independent moral actors" is quite the disingenuous strawman. (And I don't believe people are independent moral actors in the way you often represent them frankly, although it is often useful to treat them as such). And Cosma's position hardly fucking comes from his reading too much Charlie fucking Stross.
Looking forward to learning how this makes me a rape apologist.
I usually think it best to trust Cosma, because he knows more statistics than me.
He probably lives closer to Wendy Bell than I do too.
I thought "Ms. Jackson" was sung by black men wearing space helmets.
One day, God willing, it will be.
25: "Corporations are comprehensible."
I think by this you mean that human-constructed organizations are comprehensible, yes? Not sure that's completely true. I've read enough accounts of the idiosyncracies of the law (over at LG&M) to convince me that the entirety of the law isn't actually something that humans ever willed into existence. That is, they willed into existence the Law bit-by-bit, but not all the interactions, both desirable and undesirable. In that sense, the law has its own existence.
And the same could be said of (blessed be) The Market (amen). There's an entire field (economics) that studies market dynamics, and how it can act in ways that no one or even small number of humans will it, nor understand it.
"I'm not sure AI would be. It barely is now."
Unless you're an actual AI researcher, that shouldn't be surprising, right? I mean, if you were plopped down into basically any really deep technical field, you'd be in exactly the same place. Let's say, nanotech, where the normal rules we expect for macro-scale physics just don't apply. And the same would be true if I (or any of us) were plopped down into your specialty, yes?
I think Cosma is absolutely right.
To follow on from 40:
I remember once talking with Bob Paige (great CS researcher, invented program finite differencing among many other things); he told me about how when he was starting out, he had a job automating the registrar's office of a college. It was an office with four (4!) people, and each person knew their job. But nobody actually knew how it all worked -- how the office actually functioned. He had to puzzle it all out, so he could automate it.
He assured me that this wasn't the only case, and that there were a number of examples he'd encountered of human organizations where no human actually knew how it all worked, even though humans had constructed the organization bit-by-bit.
That kind of makes my point though, doesn't it? Bob Paige, a human, was able to comprehend the organisation. It wouldn't make sense to think of The Registrars Office as a separate entity from Alice, Bob, Carol and Dave at their desks, or to think of it as making decisions for reasons that none of them could understand.
42: alternately, that these people who worked the machinery every day, couldn't comprehend it, argues that the existence of other organizations that individual or small groups of humans cannot understand, doesn't somehow make them sentient beings, or at some other level beyond the human organizations we already encounter in our everyday world. So there are:
o the registrar's office and other orgs, that are somewhat opaque, but people like Bob can divine
o the law, which sufficiently-skilled lawyers can understand areas of, with sufficient study and experience
o the economy, which constanty surprises even the most skilled and experienced of researchers
o LLMs, which seem even more incomprehensible
What I'm arguing is that this is a continuum.
Oh, another example of emergent behaviour: I remember reading about "gene signaling networks" and their vast complexity. We're still only scratching the surface of understanding them, even though at some level they're all encoded by the digital code of DNA.
43.last: A good recent book covering some of that is The Master Builder: How the New Science of the Cell Is Rewriting the Story of Life by Alfonso Martinez Arias. He is a little too hung up on promoting the cell versus the genes (which is the tool and which is the "master") but given the strong run of gene-centric treatments in popular literature it is somewhat understandable. He also pushes his own line of research somewhat excessively towards the end of the book. But those caveats aside, some interesting (and somewhat unsettling*) stuff on the orchestration of cells and genes building and operating organisms.
See Gastruloids for instance.
||
And speaking of unsettling:
I give you the Atomic* Bomb Resin Lamp:
This exceptional resin lamp carries a powerful message, reminding us to cherish every moment of our lives, as the future is uncertain.
Prepare to be captivated by the extraordinary realism of this resin lamp.
The attention to detail is astounding, with lifelike buildings and streets that transport you to another world. It even features a mushroom cloud, a symbol of a chemical explosion, adding to its striking visual impact.
*The original ad I saw screen-capped on Bluesky styled it as the Atomic Bomb Lamp, but the entry reads "explosion bomb" or "chemical explosion". And actually the thing looks to have been AI-written.
|>
46: Oh wait, it does say Atomic Bomb Resin Night Light on the site.
Also: This resin lamp is a true work of art, meticulously crafted over 20 days. It adds a unique and personalized touch to your home, desk, or living space, reflecting your individuality and artistic sensibility.
The Atomic Bomb Resin Night Lights will continue to be used until cherishment improves.
43. It is indeed complex. In contrast to the others, we don't have a good way of looking at the elements of signaling networks in their natural context ( a living metazoan cell either in stasis or changing from one state to another). Current knowledge of these is based on either various noisy high-throughput assays (which protein-protein or protein-nucleotide complexes are found after we kill a cell and look at the pieces), or bespoke experiments that modify cells so that the interactions of a few selected proteins can be examined, or by looking at cell lines, which are immortalized colonies of cells that are similar but different to what they were in a body.
Probably too far down in the weeds, but a particular kind of interaction that's now known to be necessary for stability of differentiated mammalian cells was first detected at all in the mid-nineties, https://en.wikipedia.org/wiki/MicroRNA . There are other control mechanisms distinct from this one that are less well understood, and there are many of them, we may still be entirely blind to new ones-- eg RNA editing is a niche mechanism for mammals but looks like one of the main ways that cephalopods use to specialize the function of their neurons. Modifications of already-made proteins (phosphorylation and dozens of others) are from in-vitro studies known to be intricate and necessary for cell function, but very hard to look at in vivo.
Basically while there is excellent rapid progress in understanding so much of signalling and regulatory control of cell state, IMO the limitations of current technical ability are pretty profound. I'm suspicious of most claims about system-level understanding, the quality of the measurements is not there yet-- most of the people who jump from the details of the thing they're good at measuring to systems talk don't inspire my confidence. The discussion usually has a few "unspecified miracle of understanding occurs" steps in their roadmaps. The most promising approach with current technologies is IMO is looking at perturbations, unfortunately only a few groups do that, not sure how strong their support is in the arcane and occasionally very surprising byways of setting research direction, and I'm too old to try joining in any case.
I don't understand AI and am not about to start to try to understand it soon. I understand "we fired the head guy for being a liar and then tried to hire him back the next day". The people involved a fuckheads and can't figure out if they are teamed up to defraud others or if they are being defrauded by each other.
Damn it, is there a job opening or not?
I think probably not. But check Indeed on Monday.
Just checked Indeed and wow I am in no way emotionally or professionally prepared for a job search.
54: I missed the pun. I must be oblivious.
https://www.ajc.com/news/rosalynn-carter-obituary/
Apparently Rosalynn Carter visited our city in 1979 and dedicated the new downtown bandstand. Her visit is fondly remembered.
34.last: you literally just said you don't think people are independently morally responsible for their own actions. That's how!
Why is Altman's CV the top NYT story three days in a row? Is AI designing their web site?
62 why indeed? He's got a helluva filmography but he's been dead for 17 years.
62 It's a fun little tech soap opera, why wouldn't the NYT try to sell some soap off it.