Well, it's an LLC once the human initializers withdraw from it, that's the legal trick here. If you mean what is the actual algorithm using the LLC shell, it could be anything. A simple flowchart, a DAC on Ethereum, a NN on AWS, whatever.
People get away with illegal or unlawful things all the time because of the unawareness of the authorities, but that doesn't change the nature of those things.
I've sometimes wondered how, or whether, this is guaranteed for companies whose ownership is organised by tradeable shares.
That is, whether there's anything that stops a sequence of share sales ending up in a situation where a set of companies between them own all the shares of companies in the set.
(If this ever happened, presumably the last human to sell a share to one of the companies was making a mistake, but humans make mistakes all the time.)
I mean, normally if you wanted to stop this sort of thing you'd do something like introduce a hierarchy of company types and say type N companies may only own shares in companies of type N+1 or higher. But I've never heard of such a rule.
I think self-owning corporations are neat as a Scifi plot device, but fears of human extinction due to corporate structures seem overblown. It's not as if human-controlled companies haven't needed to be restrained by law in the past.
"Algorithmic entities are likely to prosper first and most in criminal, terrorist, and other anti-social activities because that is where they have their greatest comparative advantage (edit: or what advantages exactly that they have?) over human-controlled entities."
Can anyone further elucidate what's meant by "comparative advantage"? also, i find this sentence particularly interesting. how did they come up with this notion?
Non-human-owned corporations are insulated from criminal legal consequences, and therefore are better able to engage in activity that would otherwise engender criminal prosecution, relative to a human.
this is a fun paradox to entertain, but irl we would probably just trace the "stack" back to the human who wrote the program that started the chain of program-writing programs and hold them liable (unless we can show intent, then charge them criminally). if you can't trace back far enough to find a human then I guess you just throw up your hands and give up, which is more or less what we do anyway when we can't attribute a malicious program to a specific person. so altogether not as novel as it seems.
Wrong analogy. I set up a blank slate. Random people on the internet start to mould it into a gun that fires on people incrementally, each step fairly innocuous. Who is responsible?
"who is responsible?" is a philosophical question that probably has no objective answer.
"who would be held responsible?" is still pretty easy here. by default, if you own or release to the public a system that ends up hurting someone, you are liable. it doesn't matter what your intent was or whether there was any way of knowing that the bad outcome would happen. the only way you can get around this is with "safe harbor" exemptions. it's possible that your blank slate would qualify, although you would probably have to lobby for this before the fact. in that case, no one would be held responsible. otherwise, you would be paying.
The community has a garbage tip. Due to an abundance of uraniaum, everyone starts discarding it. When the garbage tip reaches critical mass and kills everyone, who was to blame?
That's a far easier scenario than the ones involving recursively generated AIs. You are, and would likely be charged with Murder 2 in most US jurisdictions.
It has no fear of anything physical or social such as jail, public humiliation, police brutality, poor health. Fear itself is irrelevant, including fear of dissolution. The entity can operate as per its primary goals without concern as to the ultimate success of the goals, whether that is siphoning money to a hidden account, disrupting government services by denial of service, spawning copycat organizations to thwart law enforcement, etc.
This all already exists in the form of large, sometimes multinational, organizations. Has anyone from Well's Fargo been criminally prosecuted from the massive identity fraud they conducted by opening accounts falsely? From Equifax?
We don't need AIs to achieve this level of self-directed autonomy free to engage in criminal behavior without consequences: we already have Employee Handbooks.
The government? Or the creator? If it were the creator I'm sure that the creator would have the common sense to pregenerate and store the private keys somewhere safe.
What if you couldn’t? Imagine an AI running in the cloud, self replicating, totally free of human input, no physical presence anywhere, with sole access to a bitcoin wallet from which it pays expenses and to which it receives revenue. No assets that are seize-able. No goals other than to collect more and more profit. How do you stop it from operating?
Seize all its physical assets and make it illegal to trade with. Bitcoin is arguably impossible to control, but real world assets not so much. Property rights as they currently exist are ultimately enforced by state violence. A number in a blockchain only has meaning if those who enforce physical property rights allow it to.
And if the AI manages to actually assert control of property on its own, it is no longer a market actor, it is a revolutionary army.
It's the notion that we want to be compensated for risk in life. That's true of entrepreneurship, investing, lending out money etc. Shady industries tend to carry high risks, and therefore high risk-premiums to compensate. (e.g. see the price of illegal drugs vs the raw production costs in a legal environment)
But computers can neglect some of those risks. Not all, but some. And therefore needn't be compensated for them, and have a competitive advantage over humans.
"comparative advantage" is probably the most important simple concept in economics that is not self evident and intuitive. It was first discovered by Ricardo 200 years ago.
Comparative advantage comes from the ability to produce goods or services at a lower opportunity cost (you should also know what "opportunity cost" means to fully grasp the idea).
In the original formulation it means is that can be beneficial for two entities to trade even when one entity can produce all products cheaper than the other (it has an absolute advantage in all categories) if there is differences in the relative cost they produce those products themselves.
Suppose in a given day, Henry can catch up to 4 deer, or 6 antelopes. John can catch 24 deer, or 12 antelopes. John can catch way more of either animal than Henry, so we say John has an absolute advantage in both products. You might think that John can not benefit from trading with Henry, because he can produce every product better than Henry can.
However, we can break it down another way:
It costs Henry 3/2 antelopes to catch 1 deer, and 2/3 deer to catch one antelope (prove it to yourself: 6/4 and 4/6). It costs John 1/2 antelope to catch 1 deer, and 2 deer to catch one antelope. Thus, it costs less for Henry to specialize in catching antelopes, and for John to specialize in deer. This is comparative advantage. Then, when they trade at a rate between the two opportunity costs, they can achieve a combination of products unattainable individually:
Before trade:
- Henry: 6 antelopes, 0 deer
- John: 0 antelopes, 24 deer
After trade:
- Henry: 4 antelopes, 2 deer
- John: 2 antelopes, 22 deer
While their totals are the same here, because an exchange rate of 1 antelope to 1 deer works out here, it helps to look at it graphically:
For this, we can make a production possibilities curve (or line, to keep it simple), which is a graph of possible outputs where the axes are the amount of each product that John or Henry can produce in a day.
Henry can obtain any point on or inside the purple. Any point outside of that is unobtainable. However, with trade, he can obtain it! The result of trade is the green dot.
As you can see, they both obtain points outside of what's possible to produce alone, due to comparative advantage!
Note: I don't claim to be an economics expert (just a high schooler taking the AP course!), and I'm sure there will be lots of better explanations in this thread!
>In two recent articles, Professor Shawn Bayern demonstrated that anyone
can confer legal personhood on an autonomous computer algorithm merely
by putting it in control of a limited liability company (LLC). 10 The algorithm
can exercise the rights of the entity, making them effectively rights of the
algorithm.
>The rights of such an algorithmic entity (AE) would include the rights to
privacy, 11 to own property, to enter into contracts, to be represented by counsel, to be free from unreasonable search and seizure, 12 to equal protection of
the laws, 13 to speak freely, and perhaps even to spend money on political
campaigns. 14 Once an algorithm had such rights, Bayern observed, it would
also have the power to confer equivalent rights on other algorithms by form-
ing additional entities and putting those algorithms in control of them. 15
Shawn Bayern, The Implications of Modern Business Entity Law for the Regulation
of Autonomous Systems, 19 S TAN . T ECH . L. R EV . 93 (2015) [hereinafter Bayern, Entity
Law]; Shawn Bayern, Of Bitcoins, Independently Wealthy Software, and the Zero-Member
LLC, 108 N W . U. L. R EV . 1485 (2014) [hereinafter Bayern, Wealthy Software].
>Abstract
>Nonhuman autonomous systems are not legal persons under current law. The history of organizational law, however, demonstrates that agreements can, with increasing degrees of autonomy, direct the actions of legal persons. Agreements are isomorphic with algorithms; that is, a legally enforceable agreement can give legal effect to the arbitrary discernible states of an algorithm or other process. As a result, autonomous systems may end up being able, at least, to emulate many of the private-law rights of legal persons. This essay demonstrates a technique by which this is possible by means of limited liability companies (LLCs), a very flexible modern type of business organizations. The techniques that this essay describes are not just futuristic possibilities; as this essay argues, they are already possible under current law.
Right up until the first real court challenge where any competent judge will declare algorithms are not persons, cannot own property, and an LLC without any legitimate ownership interest is an abandoned entity (making all its assets abandoned property).
Alternatively for criminal purposes the courts will simply "pierce the veil" and declare the actual owners to be the humans profiting from the algorithm's actions.
I think the idea is that the ai could be emancipated from whatever "owner" it had, and then would control everything. It would be more efficient for profits to go back into the corporation instead of to any owner.
> and declare the actual owners to be the humans profiting from the algorithm's actions
It counters the parent comment in that there would be no humans profiting from the algorithm, since all the profit is kept in the LLC and governed by the algorithm.
Then, by the parent comment's compelling logic, the profit along with the LLC would become abandoned property:
>>>> and an LLC without any legitimate ownership interest is an abandoned entity (making all its assets abandoned property).
The bit you quoted was in the context of assigning criminal liability, if the need arose. However, I don't think that "profit" would be the most important part of the analysis, but rather who was responsible for designing and setting up the corporate entity.
The idea being corrupted by the fact that at some point the AI was owned by someone - it would be a real stretch to assume that any subsequent emancipation is not going to be scrutinized heavily.
It's the weak legal point here in the entire idea.
It would still have to buy the services of the former owner (or someone equivalent). there are after-all things only a human can do, still, and presumably this corp would have to pay for those things.
What happens when you have a 'being', made of silicon and algorithms, but takes the form of a human? Or, for a better case, what happens when our own brains are 'part algorithm'? In that case I think this sort of case wouldn't be so cut and dry. I'm thinking of some recent films and TV shows that really explored these issues; Humans, Ex Machina, Westworld, and Altered Carbon.
That's a rather vague objection. Suppose someone destructively uploaded their brain into a computer. The rule could be that virtual person was identical to the original OR the ruling could be the original person died. Thus, saying an algorithm used to be a person need not be relevant.
On the other hand, people use algorithms every day. Consider a simple vending machine, people inter contracts with the owner by proxy. The specifics on how the machine operates is basically meaningless.
The paper addresses your second objection pretty early. States sell anonymous corporations, without even knowing the identities of the people who bought them, and for various reasons that's not likely to change anytime soon.
Don't those anonymous corporations need to have a real-person representative appointed for them? Maybe they're officially anonymous, but I can't imagine that anonymity would hold up in the face of intense scrutiny.
I believe that you are correct. A corporation has to publicly disclose some of its officers (in my state, a president and a secretary, who can be the same person or not, but must be persons) and an address so that it is available to be sued when the need arises. The states also require that corporations have some minimum number (sometimes one) of human directors (members of the board) who are responsible for overseeing what the corporation does. Humans must also sign the forms that disclose all this (typically each year) and sign the federal and state tax returns.
Depends. A corporation can be owned by another in turn owned by another. And a US corp can be owned by, say, an Isle of Man corp. You can find an address for that corp and send them nastygrams; whether they respond is a different question.
You can be pretty sure there's a human down there somewhere, but it can be very difficult to actually find a name.
There will at the very least be nominee directors or other human agents acting on behalf of the corporation somewhere in the chain.
The problem as I see it is that you can probably "bootstrap" some ownership by using two or more corporations owning shares in each other (you'd have problems with laws relating to controlling stakes, but they probably are not unsurmountable) and make it work in the normal case with enough automation and nominee directors with strict, limited instructions on how to act and how to seek approvals - insist the owner is a recluse who will only interact via a strictly limited app or something.
The problem is likely to be what happens if anyone starts suspecting "nobody is home". There are too many opportunities for people then to take advantage, or take actions that an algorithm will have problems countering. Something as simple as filing a basic lawsuit against the company would likely make it all come tumbling down, because even if the algorithm contains the intelligence to make the right decisions, humans have this annoying tendency to still want things like signatures or want people to appear in person barring legal rights to equal treatment for algorithms.
So such companies may be possible, whether or not technically legal, but may not end up surviving very long. For now.
To repeat myself: Even if the algorithm contains the intelligence to make the right decisions, humans have this annoying tendency to still want things like signatures or want people to appear in person barring legal rights to equal treatment for algorithms.
The algorithms might know to hire a lawfirm, but that just shifts the problem around a bit. What happens when the law firm wants a signature on engagement documents? Or needs a deposition? Trying to avoid even providing signatures will set off red flags from the outset. Banks deal with "know your customer" laws that makes them liable if they don't look into who they're dealing with. Lawyers and accounts and many others can get struck off or face legal sanctions as well.
As an example, my accountants does everything to do with my finances, no problems. I could easily script my interactions with them. Apart from one little detail: They regularly need me to sign stuff. Like my tax returns and those of my company. At which point said algorithm would need to commit fraud, or use the signature of a responsible human.
You might be able to eliminate some of these e.g. by hiring a board consisting of humans acting on behalf of a "reclusive" shareholder, but the problem is that you're then handing these people extensive power over the company with no real way of reigning them in if they uncover the truth.
Presumably, this same algorithmic actor would bring to bear the power available to it to render this 'so-called judges'' ruling void. Money buys many things invisible to the untrained eye.
Don't take this wrong, I want to be wrong, but justice better learn to swing the sword hard. Money shouldn't buy the public's opinion, nor their representatives.
That is, of course, provided that the judge will find out that the company is being autonomous entity.
I envision that some crazy billionaire will create an autonomous hedge fund, which will automatically off-hand manage a group of ordinary looking funds. People managing the children funds will not know that the parent fund is an AI entity and will not know about the existence of other funds. The AI entity overseeing the main fund will compare profits and other behavior of each children fund and any significant deviation will be automatically killed off and replaced.
This is a key plot point in Person of Interest. I won't say more because it'd be a shame to spoil such a fantastic series, but I'll go ahead and recommend it to anyone amused by the idea :)
Normally I agree about judges' disdain for cute tricks. However, judges already accept corps having legal entity status, which is the hard part of this idea.
If this was used to try and shelter a human identity then a judge might pierce that, but if there is just a company that has bought up all of its shares (or controlling shares), then there is no particular shielding involved.
We already see corps today preventing individual liability...many corps committed fraud (?) in their risk assessments (at a minimum) leading to the Great Recession, but I'm unaware of any human going to jail, and if any human was convicted of anything it was relatively trivial. Instead what happened was some corporate ownership traded hands.
As soon as/if remote-work-via-contract becomes a thing at the higher levels, it is easy to think of a company having human workers, but lacking actual humans at the top.
That doesn't mean skynet...but it does mean no human for the judge to unveil. Just a corporation represented by human lawyers. We probably wouldn't even notice in most cases.
But the basis of decisions like Citizens United is that corporations have rights (free speech and so on) because they're just collections of people, who have those rights as individuals. The key idea behind CU is that people don't give up those rights just because they formed a corporation.
In other words, the rights of corporations are really just the rights of the people in them.
If there are no people in them, I could very easily see a judge deciding the corporation itself has no rights.
> If there are no people in them, I could very easily see a judge deciding the corporation itself has no rights
I might quibble with "very easily", but I don't argue your core point. I can, however, ALSO see a judge ruling that any corporation that employs humans retains rights in the name of the interests of those humans. Not a guarantee, possibly even a stretch, but believable (And CU did NOT say "the humans of a corp are not limited in their free speech outside of the corp, therefore the actions of the corp can be limited without similarly restricting the rights of those humans", so I'll stick by my 'corps have legal entity status beyond their humans' suspicions, particularly given how long corps have had legal status in multiple countries before CU was even a concept.
Regardless of right/wrong, if a reasonable case can be made either way, then there's a good reason to not have full confidence that self-owning corporations would be tossed out of court, which was my point.
> However, judges already accept corps having legal entity status
To be fair, that isn't a trick. That's been the established legal norm since 1886. Corporations are people for a variety of reasons, including (as Gargoyle mentions) because they're just collections of people, but also because without at least some forms of personhood, we wouldn't be able to sue corporations who deliberately or negligently violate the law and harm their customers (or neighbors, et al.)
Because of personhood, a corporation has _liability_, beyond just the usual chicanery.
> Because of personhood, a corporation has _liability_, beyond just the usual chicanery.
I wasn't denying that a self-owned corporation could be held liable, I was arguing that we can't count on a judge tossing out the concept of a self-owned corporation. (It's possible, maybe even probable, though I doubt it, but not basically guaranteed).
Understood. My point wasn't that a self-owned corporation would have liability, I was arguing that corporate personhood isn't a good example of legal trickery, since not only is that opinion pretty exclusive to non-law people, but it's totally been that way for the entire lives of every sitting judge.
A lot of this does kinda sound like the stuff "sovereign citizens" try to pull as "gotchas" over the courts, with stuff like "maritime fringe" and the "can't impede traveling" stuff.
Yes, I doubt that judges will accept algorithms as persons based on strained readings of law and precedent. This is something that will require legislation. If not amending constitutions.
Laws, especially corporate laws, are written (by congress) or interpreted (by courts) in such a convoluted way that they regularly defy common sense. The hypothetical scenario described by the author is in no way weirder than this very real scenario:
If a corporation is competitive enough to exist and outcompete other human corporations, then maybe it's not really self-owned, is it?
It's controlled by an synthetic consciousness that could be sentient and thus deserving of legal rights similar to the other sentient consciousnesses of biological origin.
my guess is that the handwritten signature is already on its way out of being relevant in modern society. in most cases it seems to serve more as a formality than a security measure. when I worked as a cashier, the majority of people would just do some squiggle and get the hell out of the store rather than any kind of consistent signature.
in principle at least, public key crypto is way more effective than a signature, and I expect this will become prevalent in the next couple decades as laws and customs enter modernity. at this point it would be trivial for a digital entity to enter into a binding contract.
It's not the form of signature that is important here. It's the act of signing itself. Legal entities do not sign anything, they give authority to some human to do it in their name.
Yep it could e-sign. EU regulation 910/2014 had a concept of e-stamp in addition to e-sign.
In Ukraine we fill taxes with e-sign and get a signed reciept back. Reciept e-signature however is not linked to any human and simply states "fiscal service. recieved" it is considered to be legally signed by a legal entity itself.
Normally you would see some person signing on behalf of company with both person and legal entity listed in X509, but not with stamps.
The most efficient self-owned entity would eliminate the waste of profit going to executives and board members and could reinvest that excess back into itself. It would be a new era when even the select few can't collect rent from their operations.
I want to see autonomous companies that would be completely governed by AI systems and would choose by themselves what to invest in, what facilities to build and where, what to manufacture, who to employ, where to expand, and so on. Corporations already have legal personhood in American law.
The videogame Shenzhen I/O already explored a similar concept superficially.
An AI system that can do all that could clone itself, and so no need to worry about "who to employ". Infact humans and humans need would be an unnecessary burden other than to be the 'market' that the machine wants to make money from. However those humans won't have jobs, because the AI clones will be able to do their jobs. So they won't have money and there is no use pandering to them. So the machine will worry about what other AI bots want. Well they don't want champagne or ferraris. They want electricity. So expect power / manufacturing companies to do well and everything else to go to the shit.
one could argue that some large multi-national corporations are already acting as a collective & super-human intelligence. Indeed, isn't that the point of many human "institutions" which subscribe to the notion that rule of law (the code) supersedes the rule of man.
Well, a foundation is basically a self-owning company that is steered by its stated goal and often outlives its creators. In addition, there are several other legal forms that allow companies to own themselves. For example, in Germany there's the concept of a "Kein-Mann GmbH" (no man's GmbH), which is a limited liability company that owns 100 % of its own shares.
... which apparently funnels into trusts in Caribbean countries that are not required to disclose beneficiaries. Basically, it sounds like tax evasion at a grand scale.
Point taken, but I meant evasion. What looks like mis-use of a non-profit entity, combined with funneling money into a black hole trust through "trademark licensing" has the appearance of evasion, at least to me.
For Foundations and other non-profits that don't have owners, there are no shares to sell. The assets belong to the "Public" and if the organization winds down, the trustees have to designate another non-profit to receive the excess assets (less liabilities) or they just turn over everything to the government.
In my (possibly naive) view, the board members of a foundation are just the executive branch of that foundation and are obliged to act according to its statutes. A good example might be the "Nobel Foundation" (https://www.nobelprize.org/nobel_organizations/nobelfoundati...): It was founded in 1900,has outlived its creator(s) by a large margin, has been steered and managed by several generations of board members and continues to fulfill its mission. For me, this IS an autonomous organization in the sense that it is autonomously steered by the "smart contract" that are the statutes of the foundation and that no one except the government of Sweden would be able to dissolve it (and even they would first need to introduce changes in law to do so legally). You can of course argue that a DAO on the blockchain is not subject to the law of any state, but on the other hand a pure blockchain organization would probably not be very useful if it doesn't have any embodiments in the form of a legal organization that can operate within the law of a given country (there still could be cases where a purely virtual organization could create value by e.g. providing a commodity like Bitcoin).
The article seems confused...the title references ownership and the abstract is not about ownership at all but about control of LLC’s and in a legal context ownership is distinct from control.
Otherwise the article is based on a false premises:
>Third, governments — particularly in the United States — lack the ability to determine who controls entities they charter
Take Florida. Florida’s Limited Liability Company Act regulates who can be a manager (i.e. who controls the LLC). In fact the act creates 2 types of LLCs as most states do, member managed and manager managed. In one non-members can not be managers (control) the LLC.
Outside of the LLC, almost every statute regulates who can control and who can own a business entity. For example, a non-profit corporation has no owners. Although states do differ, generally by statute, a non licensed person cannot be the owner of a professional service corporation, professional association, or professional limited liability company. Though generally those professional entities can have non-licensed directors, officers, or managers. State laws also generally include restriction on minors being directors, officers or managers, but they can usually be owners, although it gets tricky where they are the sole owner because legally they can’t even sign contracts/formation documentation (though still legally possible).
Point is Governments are well equipped to regulate who can own or who can control any type of business entity.
What is possible, and I have written legal articles about, is to have autonomous trusts, or what I call the trusteeless trust but my examples and use cases are limited to the trust being funded solely with cryptocurrencies/tokens.
I can no longer edit my comment, but I have now read all 75 Pages, and the Article does in fact mention some of these issues, such as member-managed vs manager-managed LLCs and then in the "detection problem" section they discuss incorporators, directors, officers of corporations.
Still the premises of the Article is flawed, and found in Section 1(b), Initiating Algorithmic Entities, on page 13, in part:
>Bayern specifies this chain of events as capable of establishing the link:
(1) [A]n individual member creates a member-managed LLC,
filing the appropriate paperwork with the state; (2) the individual(along, possibly, with the LLC, which is controlled by the sole member) enters into an operating agreement governing the conduct of the LLC; (3) the operating agreement specifies that the LLC will take actions as determined by an autonomous system, specifying terms or conditions as appropriate to achieve the autonomous system’s legal goals; (4) the sole member
withdraws from the LLC, leaving the LLC without any members. The result is potentially a perpetual LLC—a new legal person—that requires no ongoing intervention from any preexisting legal person in order to maintain its status.
The premises is the last member (owner) withdrawing from the LLC leaving an AI to run the perpetual LLC, which is a legal impossibility. Every State has a law requiring an LLC have at least one member, and upon the withdraw of the last member of an LLC the LLC immediately becomes administratively dissolved (no longer in good standing with the State) under the State's law. Its not novel, similar laws govern situations when the sole member of an LLC dies, either the LLC automatically dissolves or ownership gets distributed to the member's heirs, pursuant to the specific State's laws, but the managers and/or employees don't go on running the member-less LLC whether they are human or an AI.
On a related note, what if we substitute a person on a remote connection to a computer instead of an AI? That seems far more dangerous and probably exists today.
The article addresses this as a potential issue, but also goes into the game of cat and mouse with humans currently creating fake identities and stealing identities. The authors don’t seem overly concerned about this and conclude the system (States, IRS, Etc...) are pretty effective at weeding this out already.
> By contrast, an algorithm could generate an unlimited number of artificial entities quickly and easily, without violating any law. For algorithms – and other criminals – entities can function as identities.
> Artificial entities can more easily generate credibility because they are a form with which business people are already familiar.
How would that not violate the law? If my AI commits a crime, am I not liable for it?
(I'm guilty of skim-reading - maybe the answer is staring me in the face.)
The exact outcome in such a scenario must be detailed in the Articles of Incorporation for said non-profit.
The law forbids the officers/board from benefiting financially from their non-profit corporation's end, and so many Articles of Incorporation stipulate that, upon the demise of said non-profit corp., all assets and money shall be donated to some specific charity or non-commercial group.
If algorithm wasn't required, then one might create two corporations that own each other with one taking action in market. That way what has ownership is something that is already allowed to own something under existing law. One outsources its management or other operational capability to the other so only one is actually doing stuff. The other is just there.
What do you who are more knowledgeable about this kind of law or prior decisions think of that? Is it just as unlikely to work or a bit more likely?
Can LLC ownership contain cycles? E.g. A is 50% owned by B which is 25% owned by A? Would a US state or the IRS notice if this happened on a longer obfuscated chain?
I thought that's what the article was going to imply based on the title, but it wasn't : )
LLCs organized as partnerships are required to report their partners to the IRS. If the partners are partnerships themselves, they must keep disclosing the partners recursively up to individual or corporate partners.
LLCs may be taxed as flow-through entities (similar to partnerships), but no LLC is organized as a partnership, that's simply a nonsensical statement; an LLC is a distinct and separate type of entity to a partnership, limited partnership, LLP, LLLP and corporation.
For tax purposes, an LLC is either a corporation or a partnership; it's not merely similar to one or the other. LPs, LLPs, and LLLPs are taxed like partnerships, unless they are foreign registered, in which case they are taxed like corporations.
An LLC treated as as partnership for tax purposes is required to disclose it's partners all the way up the chain.
Legally, an LLC is organized like a partnership (member managed) or like a corporation (manager managed). I should have been more precise there. I was using shorthand because I was assuming the tax context and should have acknowledged that the tax treatment differs from the legal treatment.
In the Rosinante trilogy of Alexis A. Gilliland the autonomous AIs are self-owned corporations, and indeed are sometimes addressed with the word "Corporate" as a title, somewhat like "Mr" or "Ms". This is of course fiction; while much of the engineering in the series seems plausible I have no idea if the legal theories are. Worth reading anyway, the AIs are charming.
I suspect it's a summary of the first line of the abstract:
> "In an article published in the Northwestern University Law Review in 2014, Professor Shawn Bayern demonstrated that anyone can confer legal personhood on an autonomous computer algorithm merely by putting it in control of a limited liability company."
Given that the first sentence is too long, I think it's a reasonable summary of it, though reasonable people may disagree.
In these cases, it's really helpful to provide what the actual title is (in this case "Algorithmic Entities"), and, if so inclined, a suggestion for a better title. It provides something actionable that a mod can use when updating. Do you have any suggestions for a better title?
And if you've really got a bee in your bonnet, you can let the mods know directly via the Contact link in the footer. They don't see every comment but emailing them makes sure they see it.
> Given that the first sentence is too long, I think it's a reasonable summary of it, though reasonable people may disagree.
The difference between the title and the contents is the difference between control and ownership. They are not the same. The article discusses control not ownership. That is also the case in the portion of the article you quoted.
Great! Would you be willing to continue moving this forward rather than just picking it apart? Going beyond only debugging and providing a patch, as it were?
Huh? You’re seriously telling me that an algorithm can be the sole owner of a corporation? That doesn’t sound right. Typically when opening financial accounts and establishing credit for the business the bank would ask for the SSN of a real person, I would think.
187 comments
[ 3.1 ms ] story [ 223 ms ] threadWhat is an example of one these entities? The abstract doesn't really explain what it means here.
If they ever cared enough to look, I'd imagine.
People get away with illegal or unlawful things all the time because of the unawareness of the authorities, but that doesn't change the nature of those things.
That is, whether there's anything that stops a sequence of share sales ending up in a situation where a set of companies between them own all the shares of companies in the set.
(If this ever happened, presumably the last human to sell a share to one of the companies was making a mistake, but humans make mistakes all the time.)
I mean, normally if you wanted to stop this sort of thing you'd do something like introduce a hierarchy of company types and say type N companies may only own shares in companies of type N+1 or higher. But I've never heard of such a rule.
https://en.wikipedia.org/wiki/Endgame:_Singularity
[1] http://www.antipope.org/charlie/blog-static/2018/01/dude-you...
Can anyone further elucidate what's meant by "comparative advantage"? also, i find this sentence particularly interesting. how did they come up with this notion?
But for the person who created the entity, losing the company is a better outcome than going to jail.
"who would be held responsible?" is still pretty easy here. by default, if you own or release to the public a system that ends up hurting someone, you are liable. it doesn't matter what your intent was or whether there was any way of knowing that the bad outcome would happen. the only way you can get around this is with "safe harbor" exemptions. it's possible that your blank slate would qualify, although you would probably have to lobby for this before the fact. in that case, no one would be held responsible. otherwise, you would be paying.
We don't need AIs to achieve this level of self-directed autonomy free to engage in criminal behavior without consequences: we already have Employee Handbooks.
Is that illegal though? I mean, I can create an account for you on my web platform, and everyone else.
And that's what WF did: https://en.m.wikipedia.org/wiki/Wells_Fargo_account_fraud_sc...
And if the AI manages to actually assert control of property on its own, it is no longer a market actor, it is a revolutionary army.
But computers can neglect some of those risks. Not all, but some. And therefore needn't be compensated for them, and have a competitive advantage over humans.
Comparative advantage comes from the ability to produce goods or services at a lower opportunity cost (you should also know what "opportunity cost" means to fully grasp the idea).
In the original formulation it means is that can be beneficial for two entities to trade even when one entity can produce all products cheaper than the other (it has an absolute advantage in all categories) if there is differences in the relative cost they produce those products themselves.
https://en.wikipedia.org/wiki/Comparative_advantage
Suppose in a given day, Henry can catch up to 4 deer, or 6 antelopes. John can catch 24 deer, or 12 antelopes. John can catch way more of either animal than Henry, so we say John has an absolute advantage in both products. You might think that John can not benefit from trading with Henry, because he can produce every product better than Henry can.
However, we can break it down another way:
It costs Henry 3/2 antelopes to catch 1 deer, and 2/3 deer to catch one antelope (prove it to yourself: 6/4 and 4/6). It costs John 1/2 antelope to catch 1 deer, and 2 deer to catch one antelope. Thus, it costs less for Henry to specialize in catching antelopes, and for John to specialize in deer. This is comparative advantage. Then, when they trade at a rate between the two opportunity costs, they can achieve a combination of products unattainable individually:
Before trade:
- Henry: 6 antelopes, 0 deer
- John: 0 antelopes, 24 deer
After trade:
- Henry: 4 antelopes, 2 deer
- John: 2 antelopes, 22 deer
While their totals are the same here, because an exchange rate of 1 antelope to 1 deer works out here, it helps to look at it graphically:
For this, we can make a production possibilities curve (or line, to keep it simple), which is a graph of possible outputs where the axes are the amount of each product that John or Henry can produce in a day.
Here's Henry's: https://i.imgur.com/k9xS7Sj.png (vertical axis is antelopes, horizontal is deer)
Henry can obtain any point on or inside the purple. Any point outside of that is unobtainable. However, with trade, he can obtain it! The result of trade is the green dot.
Here's John's: https://i.imgur.com/V6aFZkm.png
As you can see, they both obtain points outside of what's possible to produce alone, due to comparative advantage!
Note: I don't claim to be an economics expert (just a high schooler taking the AP course!), and I'm sure there will be lots of better explanations in this thread!
>The rights of such an algorithmic entity (AE) would include the rights to privacy, 11 to own property, to enter into contracts, to be represented by counsel, to be free from unreasonable search and seizure, 12 to equal protection of the laws, 13 to speak freely, and perhaps even to spend money on political campaigns. 14 Once an algorithm had such rights, Bayern observed, it would also have the power to confer equivalent rights on other algorithms by form- ing additional entities and putting those algorithms in control of them. 15
---
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2758222
Shawn Bayern, The Implications of Modern Business Entity Law for the Regulation of Autonomous Systems, 19 S TAN . T ECH . L. R EV . 93 (2015) [hereinafter Bayern, Entity Law]; Shawn Bayern, Of Bitcoins, Independently Wealthy Software, and the Zero-Member LLC, 108 N W . U. L. R EV . 1485 (2014) [hereinafter Bayern, Wealthy Software].
>Abstract
>Nonhuman autonomous systems are not legal persons under current law. The history of organizational law, however, demonstrates that agreements can, with increasing degrees of autonomy, direct the actions of legal persons. Agreements are isomorphic with algorithms; that is, a legally enforceable agreement can give legal effect to the arbitrary discernible states of an algorithm or other process. As a result, autonomous systems may end up being able, at least, to emulate many of the private-law rights of legal persons. This essay demonstrates a technique by which this is possible by means of limited liability companies (LLCs), a very flexible modern type of business organizations. The techniques that this essay describes are not just futuristic possibilities; as this essay argues, they are already possible under current law.
Alternatively for criminal purposes the courts will simply "pierce the veil" and declare the actual owners to be the humans profiting from the algorithm's actions.
Judges don't, as a rule, enjoy "cute tricks".
If the judge thinks the "AI" is not a person, it cannot be emancipated by some trick with an LLC.
It counters the parent comment in that there would be no humans profiting from the algorithm, since all the profit is kept in the LLC and governed by the algorithm.
>>>> and an LLC without any legitimate ownership interest is an abandoned entity (making all its assets abandoned property).
The bit you quoted was in the context of assigning criminal liability, if the need arose. However, I don't think that "profit" would be the most important part of the analysis, but rather who was responsible for designing and setting up the corporate entity.
It's the weak legal point here in the entire idea.
On the other hand, people use algorithms every day. Consider a simple vending machine, people inter contracts with the owner by proxy. The specifics on how the machine operates is basically meaningless.
You can be pretty sure there's a human down there somewhere, but it can be very difficult to actually find a name.
The problem as I see it is that you can probably "bootstrap" some ownership by using two or more corporations owning shares in each other (you'd have problems with laws relating to controlling stakes, but they probably are not unsurmountable) and make it work in the normal case with enough automation and nominee directors with strict, limited instructions on how to act and how to seek approvals - insist the owner is a recluse who will only interact via a strictly limited app or something.
The problem is likely to be what happens if anyone starts suspecting "nobody is home". There are too many opportunities for people then to take advantage, or take actions that an algorithm will have problems countering. Something as simple as filing a basic lawsuit against the company would likely make it all come tumbling down, because even if the algorithm contains the intelligence to make the right decisions, humans have this annoying tendency to still want things like signatures or want people to appear in person barring legal rights to equal treatment for algorithms.
So such companies may be possible, whether or not technically legal, but may not end up surviving very long. For now.
The algorithms might know to hire a lawfirm, but that just shifts the problem around a bit. What happens when the law firm wants a signature on engagement documents? Or needs a deposition? Trying to avoid even providing signatures will set off red flags from the outset. Banks deal with "know your customer" laws that makes them liable if they don't look into who they're dealing with. Lawyers and accounts and many others can get struck off or face legal sanctions as well.
As an example, my accountants does everything to do with my finances, no problems. I could easily script my interactions with them. Apart from one little detail: They regularly need me to sign stuff. Like my tax returns and those of my company. At which point said algorithm would need to commit fraud, or use the signature of a responsible human.
You might be able to eliminate some of these e.g. by hiring a board consisting of humans acting on behalf of a "reclusive" shareholder, but the problem is that you're then handing these people extensive power over the company with no real way of reigning them in if they uncover the truth.
Don't take this wrong, I want to be wrong, but justice better learn to swing the sword hard. Money shouldn't buy the public's opinion, nor their representatives.
I envision that some crazy billionaire will create an autonomous hedge fund, which will automatically off-hand manage a group of ordinary looking funds. People managing the children funds will not know that the parent fund is an AI entity and will not know about the existence of other funds. The AI entity overseeing the main fund will compare profits and other behavior of each children fund and any significant deviation will be automatically killed off and replaced.
Some people decide to take their property and destroy it. Others decide to put it in a trust, run on certain rules.
A trust is effectively the same thing as this.
If this was used to try and shelter a human identity then a judge might pierce that, but if there is just a company that has bought up all of its shares (or controlling shares), then there is no particular shielding involved.
We already see corps today preventing individual liability...many corps committed fraud (?) in their risk assessments (at a minimum) leading to the Great Recession, but I'm unaware of any human going to jail, and if any human was convicted of anything it was relatively trivial. Instead what happened was some corporate ownership traded hands.
As soon as/if remote-work-via-contract becomes a thing at the higher levels, it is easy to think of a company having human workers, but lacking actual humans at the top.
That doesn't mean skynet...but it does mean no human for the judge to unveil. Just a corporation represented by human lawyers. We probably wouldn't even notice in most cases.
In other words, the rights of corporations are really just the rights of the people in them.
If there are no people in them, I could very easily see a judge deciding the corporation itself has no rights.
I might quibble with "very easily", but I don't argue your core point. I can, however, ALSO see a judge ruling that any corporation that employs humans retains rights in the name of the interests of those humans. Not a guarantee, possibly even a stretch, but believable (And CU did NOT say "the humans of a corp are not limited in their free speech outside of the corp, therefore the actions of the corp can be limited without similarly restricting the rights of those humans", so I'll stick by my 'corps have legal entity status beyond their humans' suspicions, particularly given how long corps have had legal status in multiple countries before CU was even a concept.
Regardless of right/wrong, if a reasonable case can be made either way, then there's a good reason to not have full confidence that self-owning corporations would be tossed out of court, which was my point.
To be fair, that isn't a trick. That's been the established legal norm since 1886. Corporations are people for a variety of reasons, including (as Gargoyle mentions) because they're just collections of people, but also because without at least some forms of personhood, we wouldn't be able to sue corporations who deliberately or negligently violate the law and harm their customers (or neighbors, et al.)
Because of personhood, a corporation has _liability_, beyond just the usual chicanery.
I wasn't denying that a self-owned corporation could be held liable, I was arguing that we can't count on a judge tossing out the concept of a self-owned corporation. (It's possible, maybe even probable, though I doubt it, but not basically guaranteed).
https://www.bloomberg.com/view/articles/2015-07-14/banks-for...
Considering the court's decision in the latter, I wouldn't be the least surprised if they act as the author imagines in the former.
It's controlled by an synthetic consciousness that could be sentient and thus deserving of legal rights similar to the other sentient consciousnesses of biological origin.
You could throw together a "self-governing" online store in a few days. Drop-ship all products, make sure it stays up to date on its bills.
It wouldn't be very resilient, but it could feasibly just sit there accumulating money, no intelligence required.
I guess it could e-sign it?
give someone power of attorney to sign papers on it's behalf?
in principle at least, public key crypto is way more effective than a signature, and I expect this will become prevalent in the next couple decades as laws and customs enter modernity. at this point it would be trivial for a digital entity to enter into a binding contract.
In Ukraine we fill taxes with e-sign and get a signed reciept back. Reciept e-signature however is not linked to any human and simply states "fiscal service. recieved" it is considered to be legally signed by a legal entity itself.
Normally you would see some person signing on behalf of company with both person and legal entity listed in X509, but not with stamps.
The videogame Shenzhen I/O already explored a similar concept superficially.
Thanks for spoiling it all to me I guess...
His point was that companies can do this anywhere, not just Germany.
... which apparently funnels into trusts in Caribbean countries that are not required to disclose beneficiaries. Basically, it sounds like tax evasion at a grand scale.
Tax evasion is illegal under-reporting of your income, which I highly doubt Ikea does
I spent some time a while ago trying to figure out how humans can be eliminated, and couldn’t find a structure.
Otherwise the article is based on a false premises:
>Third, governments — particularly in the United States — lack the ability to determine who controls entities they charter
Take Florida. Florida’s Limited Liability Company Act regulates who can be a manager (i.e. who controls the LLC). In fact the act creates 2 types of LLCs as most states do, member managed and manager managed. In one non-members can not be managers (control) the LLC.
Outside of the LLC, almost every statute regulates who can control and who can own a business entity. For example, a non-profit corporation has no owners. Although states do differ, generally by statute, a non licensed person cannot be the owner of a professional service corporation, professional association, or professional limited liability company. Though generally those professional entities can have non-licensed directors, officers, or managers. State laws also generally include restriction on minors being directors, officers or managers, but they can usually be owners, although it gets tricky where they are the sole owner because legally they can’t even sign contracts/formation documentation (though still legally possible).
Point is Governments are well equipped to regulate who can own or who can control any type of business entity.
What is possible, and I have written legal articles about, is to have autonomous trusts, or what I call the trusteeless trust but my examples and use cases are limited to the trust being funded solely with cryptocurrencies/tokens.
Why do I get the feeling you are secretly an AI trying to trick us humans.
Link?
Still the premises of the Article is flawed, and found in Section 1(b), Initiating Algorithmic Entities, on page 13, in part:
>Bayern specifies this chain of events as capable of establishing the link: (1) [A]n individual member creates a member-managed LLC, filing the appropriate paperwork with the state; (2) the individual(along, possibly, with the LLC, which is controlled by the sole member) enters into an operating agreement governing the conduct of the LLC; (3) the operating agreement specifies that the LLC will take actions as determined by an autonomous system, specifying terms or conditions as appropriate to achieve the autonomous system’s legal goals; (4) the sole member withdraws from the LLC, leaving the LLC without any members. The result is potentially a perpetual LLC—a new legal person—that requires no ongoing intervention from any preexisting legal person in order to maintain its status.
The premises is the last member (owner) withdrawing from the LLC leaving an AI to run the perpetual LLC, which is a legal impossibility. Every State has a law requiring an LLC have at least one member, and upon the withdraw of the last member of an LLC the LLC immediately becomes administratively dissolved (no longer in good standing with the State) under the State's law. Its not novel, similar laws govern situations when the sole member of an LLC dies, either the LLC automatically dissolves or ownership gets distributed to the member's heirs, pursuant to the specific State's laws, but the managers and/or employees don't go on running the member-less LLC whether they are human or an AI.
Are there any jurisdictions that allow incorporation to be represented by a Bearer Instrument?
https://en.wikipedia.org/wiki/Bearer_instrument
On a related note, what if we substitute a person on a remote connection to a computer instead of an AI? That seems far more dangerous and probably exists today.
The article addresses this as a potential issue, but also goes into the game of cat and mouse with humans currently creating fake identities and stealing identities. The authors don’t seem overly concerned about this and conclude the system (States, IRS, Etc...) are pretty effective at weeding this out already.
> By contrast, an algorithm could generate an unlimited number of artificial entities quickly and easily, without violating any law. For algorithms – and other criminals – entities can function as identities. > Artificial entities can more easily generate credibility because they are a form with which business people are already familiar.
How would that not violate the law? If my AI commits a crime, am I not liable for it?
(I'm guilty of skim-reading - maybe the answer is staring me in the face.)
What happens if a non-profit dissolves?
The law forbids the officers/board from benefiting financially from their non-profit corporation's end, and so many Articles of Incorporation stipulate that, upon the demise of said non-profit corp., all assets and money shall be donated to some specific charity or non-commercial group.
What do you who are more knowledgeable about this kind of law or prior decisions think of that? Is it just as unlikely to work or a bit more likely?
Now we just need to transmit some lobster connectomes into deep space... ;-) [closer than you might think! https://github.com/openworm exists...]
I thought that's what the article was going to imply based on the title, but it wasn't : )
An LLC treated as as partnership for tax purposes is required to disclose it's partners all the way up the chain.
Legally, an LLC is organized like a partnership (member managed) or like a corporation (manager managed). I should have been more precise there. I was using shorthand because I was assuming the tax context and should have acknowledged that the tax treatment differs from the legal treatment.
I feel like I am getting a whiff of a Sovereign Citizen...
> "In an article published in the Northwestern University Law Review in 2014, Professor Shawn Bayern demonstrated that anyone can confer legal personhood on an autonomous computer algorithm merely by putting it in control of a limited liability company."
Given that the first sentence is too long, I think it's a reasonable summary of it, though reasonable people may disagree.
In these cases, it's really helpful to provide what the actual title is (in this case "Algorithmic Entities"), and, if so inclined, a suggestion for a better title. It provides something actionable that a mod can use when updating. Do you have any suggestions for a better title?
And if you've really got a bee in your bonnet, you can let the mods know directly via the Contact link in the footer. They don't see every comment but emailing them makes sure they see it.
The difference between the title and the contents is the difference between control and ownership. They are not the same. The article discusses control not ownership. That is also the case in the portion of the article you quoted.