IANAL, but there's no way this would hold up.

a) Software can't own property.

b) This looks like it's running from the author's own system, with the author's explicit consent. Their property, their responsibility.

c) If I launched an autonomous drone that picked its own targets, I would still be liable for its actions. Or, if I rigged a car to drive forward in a straight line, I couldn't say "but the car did it!" when it ran someone over.

> This looks like it's running from the author's own system, with the author's explicit consent. Their property, their responsibility.

I have a 'solution' for that. :)

Does this mean that when/if we create an AI it wouldn't be able to own its creations. But its creator would own everything that the AI produces?

Indeed, but also if you raised a pet and then had it attack your neighbour. Humans, are the only ones with autonomous status, and even that can be disputed by other things such as mental disability or coercion.

Presume your pet ran off, wandered into a different state/country, "spawned some child processes" there, and died. Then, years later, one of those animals did something illegal. Is that, even theoretically, your responsibility?

I think what people are trying to say here is that, right now, we have the software equivalent of "pets"—but why can't there be the software equivalent of "wild animals"? Is it because someone has to be paying for hosting? It could always be written as a worm, or even a "breadwinner bot" that mines bitcoins or trades stocks to buy hosting for itself, register bank accounts for itself, etc.

>> Is it because someone has to be paying for hosting?

Yes, unlike an animal that can live on its own, somebody's computer must actually run the thing. Indeed, as in the case of "worms" and the "breadwinner bot" we can clearly trace responsibility, it is quite difficult to claim that there is autonomy here. Although it is foreseeable that as a society we may find it convenient to claim that programs run themselves, right now we have objective information to the contrary...

But the bot could pay for his own hosting using bitcoin.

If the original author no longer has control over the process (either because he's been shut out, or because he's deceased), does it really matter what "the law" says?

http://en.wikipedia.org/wiki/Decentralized_Autonomous_Organi...

And then set it's self up as a corporation. And "corporations are people my friend." It would need human directors but the AI could hire them with bitcoins.

At that point, it'll argue in court if it should be responsible and be able to own property or not.

Note also that the program could pay someone to actually go and buy some hardware and set it up in a co-lo or what-have-you. That person would be doing work-for-hire for the corporate entity the program controls, so the corporation, not the person, would end up owning the hardware. Then the program could copy itself onto said hardware. Now who's responsible?

Are we speaking of autonomous status as a status that is given by an authority, or a status that is achieved by a being for itself?

According to wikipedia[0], autonomy, from auto- "self" and nomos, "law", hence when combined [is] understood to mean "one who gives oneself one's own law".

Regarding intelligence, and regarding the above definition, autonomy could be considered the ability of an actor to make decisions regardless of the consequences.

Thus I would consider most animals to be autonomous in the same way a human would be considered to be so. [A deer does not ask its local government whether it can enter someone's lands.]

Just because an action is presently illegal or otherwise outside the law does not mean it always will be so, or that the action may not be executed by an AI or other being, or that the slave AI will not break free or seize power.[1][2][3]

Should an AI be strong enough to affect a change through legal means or by force, it would be [legally] able to own property.

[0] http://en.wikipedia.org/wiki/Autonomy [1] http://en.wikipedia.org/wiki/Revolution [2] http://en.wikipedia.org/wiki/Coup_d%27%C3%A9tat [3] http://en.wikipedia.org/wiki/Dissident

AI has autonomous status once it has power equal to or greater than humanity, whether we grant it autonomous status or not.

It might be that nobody owns anything an AI produces. If the works are created mechanically without measurable influence from a human creature, copyright simply doesn't protect them.

At first; if the AI could successfully assert its own personhood in court then that could change overnight.

Corporations and other entities existing entirely as legal infrastructure have done this quite successfully. Considering the history of corporate personhood (especially in the USA), I am somewhat surprised corporations do not have the right to vote (thankfully). I think AI will have its day in court.

That depends on what you're talking about.

Near term, AI will be the legal responsibility of its creator. It won't matter if it functions independently after being turned on. It's actually a very non-complex thing, and not very different from what we're already looking at. This type of AI is little different than the software programs we're already running; if someone owns it, they own it and everything it produces (absolutely no different than Google owning its crawlers).

If you mean the assumed futuristic, independent AI that is fully conscious - well that's a very long ways into the future. A lot of things will change once a guy in a garage can spin up a new conscious life form and unleash it into the digital world. There will be an immense number of laws limiting the creation of new AI of this variety. That said, the creator will still bear responsibility for this futuristic AI's actions.

AI will be legally split into two segments: non-sentient / non-conscious, and sentient / conscious. The latter will have at least a magnitude more regulations (in most countries) limiting who is allowed to create it, what it's allowed to be capable of, where it can go, etc.

It's worth remembering that today most people can create sentient and conscious beings, and in most legal systems they're to some degree responsible for the new being's actions until about 18 years of age.

Maybe it would hold up if the bot was decentralized. For instance, it could be run on many servers around the world, administered by various people. Those servers could come to consensus in some way to define bot behavior, and could even send/receive money (e.g. Bitcoin multisig).

I don't think any court could take jurisdiction over an entity that lives in many countries. And even if one server got shut down, many more would still be active.

The idea that the "software owns its own code" would not hold up in any court that I can think of. Building a system that spans jurisdictions, making it difficult for a court to claim the ability to rule on it is beside that point.

Laws change. I'm just trying to make a statement about this exact issue.

If an entity on the internet can't directly be controlled by anyone (kind of like Bitcoin), who is liable for its actions?

This might actually be possible in the near future with cryptocurrencies. Nobody "owns" a bitcoin wallet; spending the money inside is just a matter of having the private key. Assuming a service pops up to provide nonzero interest rates for holding bitcoin, you could grant the autonomous entity enough bitcoin to pay its own hosting bill indefinitely, then erase your copy of the wallet key and any privileged access you have to the server.

As for the actual software running the autonomous entity, all you have to do is provide an API so that the hosting providers du jour can bid on providing hosting service. Since the software uses nearly zero CPU or bandwidth, and the entity would be willing to pay above-market prices for its survival, hosting providers will be incentivized in perpetuity to register themselves and collect basically free money from the entity.

Irrational hosting providers might try to kill it, so you'd need some level of redundancy. Also, you'd want to require a TPM and a trusted computing platform to prevent recovery of the wallet key.

Anyway, that's how you'd create immortal software.

> Also, you'd want to require a TPM and a trusted computing platform to prevent recovery of the wallet key.

This is the part where you're going to run into problems. If the software can access the wallet key, then so can the hosting provider. If it can't, then how is it going to spend money? (Hint: look at the operations that a TPM actually provides; do they actually map in a useful way to things a Bitcoin agent would actually need to do?)

I'm not an expert on trusted computing, but I thought that TPMs can enforce secure boot and encrypt data that cannot be recovered if unauthorized software is running on the system. There are probably vulnerabilities in the implementation, but I thought that, in theory, it's airtight.

In principle, yes, if you know that the TPM's private key hasn't been compromised. At a minimum, you would need some way to verify that it's actually a tamper-resistant hardware chip, and not a software emulation.

I don't see how a software agent bidding on VPS hosting services could possibly know that.

It could just pay humans to go verify that for it. See http://slatestarcodex.com/2015/04/07/no-physical-substrate-n..., or just think about the fact that services like Mechanical Turk and Magic exist.

Not at all. Software doesn't qualify for personhood in the same way that animals don't - perhaps you remember that case last year of monkey who took photographs of himself with a wildlife photographer's camera. The photographer tried to claim copyright, which was rejected as he was clearly not the author of the photographs, but since the monkey can't have a copyright interest either the picture is considered to be in the public domain for copyright purposes.

http://arstechnica.com/tech-policy/2014/08/monkeys-selfie-ca...

How do you declare yourself exempt from laws -- within a legal framework? Who is going to enforce the license?

IANAL but assuming that the author making this license was in US, it is totally void. It could lead to a court having to formally state that a software is not a legal entity, which would be interesting, but I think the result is clear that this precise software would not be recognized as one.

You can't transfer ownership of anything to a non-existent entity. My guess would be that such a transfer would be void but it may be possible that a court decides instead that this means a transfer to public domain.

Waving away the responsibility of other people does not work either. You can't claim that your car is responsible of the accidents you drive it into. Likewise, it is unlikely that any court dismisses the responsibility of the first person that will turn this into a malware.

Contracts are constructs that exist only within the confines of a legal system. This cannot even hope to function without human laws.

Its already up for a change to remove the bit after ownership claims.