Joho the Blog » law

May 6, 2018

[liveblog][ai] Primavera De Filippi: An autonomous flower that merges AI and Blockchain

Primavera De Filippi is an expert in blockchain-based tech. She is giving a ThursdAI talk on Plantoid, an event held by Harvard’s Berkman Klein Center for Internet & Society and the MIT Media Lab. Her talk is officially on operational autonomy vs. decisional autonomy, but it’s really about how weird things become when you build a computerized flower that merges AI and the blockchain. For me, a central question of her talk was: Can we have autonomous robots that have legal rights and can own and spend assets, without having to resort to conferring personhood on them the way we have with corporations?

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

Autonomy and liability

She begins by pointing to the 3 industrial revolutions so far: Steam led to mechanized production ; Electricity led to mass production; Electronics led to automated production. The fourth — AI — is automating knowledge production.

People are increasingly moving into the digital world, and digital systems are moving back into the physical worlds, creating cyber-physical systems. E.g., the Internet of Things senses, communicates, and acts. The Internet of Smart Things learns from the data the things collect, makes inferences, and then acts. The Internet of Autonomous Things creates new legal challenges. Various actors can be held liable: manufacturer, software developer, user, and a third party. “When do we apply legal personhood to non-humans?”

With autonomous things, the user and third parties become less liable as the software developer takes on more of the liability: There can be a bug. Someone can hack into it. The rules that make inferences are inaccurate. Or a bad moral choice has led the car into an accident.

The sw developer might have created bug-free sw but its interaction with other devices might lead to unpredictability; multiple systems operating according to different rules might be incompatible; it can be hard to identify the chain of causality. So, who will be liable? The manufacturers and owners are likely to have only limited liability.

So, maybe we’ll need generalized insurance: mandatory insurance that potentially harmful devices need to subscribe to.

Or, perhaps we will provide some form of legal personhood to machines so the manufacturers can be sued for their failings. Suing a robot would be like suing a corporation. The devices would be able to own property and assets. The EU is thinking about creating this type of agenthood for AI systems. This is obviously controversial. At least a corporation has people associated with it, while the device is just a device, Primavera points out.

So, when do we apply legal personhood to non-humans? In addition to people and corporations, some countries have assigned personhood to chimpanzees (Argentina, France) and to natural resources (NZ: Whanganui river). We do this so these entities will have rights and cannot be simply exploited.

If we give legal personhood to AI-based systems, can AI have property rights over their assets and IP? If they are legally liable, they can be held responsible for their actions, and can be sued for compensation? “Maybe they should have contractual rights so they can enter into contracts. Can they be rewarded for their work? Taxed?”Maybe they should have contractual rights so they can enter into contracts. Can they be rewarded for their work? Taxed? [All of these are going to turn out to be real questions. … Wait for it …]

Limitations: “Most of the AI-based systems deployed today are more akin to slaves than corporations.” They’re not autonomous the way people are. They are owned, controlled and maintained by people or corporations. They act as agents for their operators. They have no technical means to own or transfer assets. (Primavera recommends watching the Star Trek: The Next Generation episode “The Measure of the Man” that asks, among other things, whether Data (the android) can be dismantled and whether he can resign.)

Decisional autonomy is the capacity to make a decision on your own, but it doesn’t necessarily bring what we think of as real autonomy. E.g., an AV can decide its route. For real autonomy we need operational autonomy: no one is maintaining the thing’s operation at a technical level. To take a non-random example, a blockchain runs autonomously because there is no single operator controlling. E.g., smart contracts come with a guarantee of execution. Once a contract is registered with a blockchain, no operator can stop it. This is operational autonomy.

Blockchain meets AI. Object: Autonomy

We are getting first example of autonomous devices using blockchain. The most famous is the Samsung washing machine that can detect when the soap is empty, and makes a smart contract to order more. Autonomous cars could work with the same model; they could not be owned by anyone and collect money when someone uses them. These could be initially purchased by someone and then buy themselves off: “They’d have to be emancipated,” she says. Perhaps they and other robots can use the capital they accumulate to hire people to work for them. [Pretty interesting model for an Uber.]

She introduces Plantoid, a blockchain-based life form. “Plantoid is autonomous, self-sufficient, and can reproduce.”It’s autonomous, self-sufficient, and can reproduce. Real flowers use bees to reproduce. Plantoids use humans to collect capital for their reproduction. Their bodies are mechanical. Their spirit is an Ethereum smart contract. It collects cryptocurrency. When you feed it currency it says thank you; the Plantoid Primavera has brought, nods its flower. When it gets enough funds to reproduce itself, it triggers a smart contract that activates a call for bids to create the next version of the Plantoid. In the “mating phase” it looks for a human to create the new version. People vote with micro-donations. Then it identifies a winner and hires that human to create the new one.

There are many Plantoids in the world. Each has its own “DNA”. New artists can add to it. E.g., each artist has to decide on its governance, such as whether it will donate some funds to charity. The aim is to make it more attractive to be contributed to. The most fit get the most money and reproduces themselves. BurningMan this summer is going to feature this.

Every time one reproduces, a small cut is given to the pattern that generated it, and some to the new designer. This flips copyright on its head: the artist has an incentive to make her design more visible and accessible and attractive.

So, why provide legal personhood to autonomous devices? We want them to be able to own their own assets, to assume contractual rights, and legal capacity so they can sue and be sued, and limit their liability. “ Blockchain lets us do that without having to declare the robot to be a legal person.” Blockchain lets us do that without having to declare the robot to be a legal person.

The plant effectively owns the cryptofunds. The law cannot affect this. Smart contracts are enforced by code

Who are the parties to the contract? The original author and new artist? The master agreement? Who can sue who in case of a breach? We don’t know how to answer these questions yet.

Can a plantoid sure for breach of contract? Not if the legal system doesn’t recognize them as legal persons. So who is liable if the plant hurts someone? Can we provide a mechanism for this without conferring personhood? “How do you enforce the law against autonomous agents that cannot be stopped and whose property cannot be seized?”


Could you do this with live plants? People would bioengineer them…

A: Yes. Plantoid has already been forked this way. There’s an idea for a forest offering trees to be cut down, with the compensation going to the forest which might eventually buy more land to expand itself.

My interest in this grew out of my interest in decentralized organizations. This enables a project to be an entity that assumes liability for its actions, and to reproduce itself.

Q: [me] Do you own this plantoid?

A: Hmm. I own the physical instantiation but not the code or the smart contract. If this one broke, I could make a new one that connects to the same smart contract. If someone gets hurt because it falls on the, I’m probably liable. If the smart contract is funding terrorism, I’m not the owner of that contract. The physical object is doing nothing but reacting to donations.

Q: But the aim of its reactions is to attract more money…

A: It will be up to the judge.

Q: What are the most likely senarios for the development of these weird objects?

A: A blockchain can provide the interface for humans interacting with each other without needing a legal entity, such as Uber, to centralize control. But you need people to decide to do this. The question is how these entities change the structure of the organization.

Comments Off on [liveblog][ai] Primavera De Filippi: An autonomous flower that merges AI and Blockchain

January 15, 2013

Tim Wu on prosecuting Aaron

… Swartz must be compared to two other eccentric geniuses, Steve Jobs and Steve Wozniak, who, in the nineteen-seventies, committed crimes similar to, but more economically damaging than, Swartz’s. Those two men hacked A.T. & T.’s telephone system to make free long-distance calls, and actually sold the illegal devices (blue boxes) to make cash. Their mentor, John Draper, did go to jail for a few months (where he wrote one of the world’s first word processors), but Jobs and Wozniak were never prosecuted. Instead, they got bored of phreaking and built a computer. The great ones almost always operate at the edge.

That was then. In our age, armed with laws passed in the nineteen-eighties and meant for serious criminals, the federal prosecutor Carmen Ortiz approved a felony indictment that originally demanded up to thirty-five years in prison. Worse still, her legal authority to take down Swartz was shaky. Just last year, the Ninth Circuit Court of Appeals threw out a similar prosecution. Chief Judge Alex Kozinski, a prominent conservative, refused to read the law in a way that would make a criminal of “everyone who uses a computer in violation of computer use restrictions—which may well include everyone who uses a computer.” Ortiz and her lawyers relied on that reading to target one of our best and brightest.

It’s one thing to stretch the law to stop a criminal syndicate or terrorist organization. It’s quite another when prosecuting a reckless young man. The prosecutors forgot that, as public officials, their job isn’t to try and win at all costs but to use the awesome power of criminal law to protect the public from actual harm. Ortiz has not commented on the case. But, had she been in charge when Jobs and Wozniak were breaking the laws, we might never have had Apple computers. It was at this moment that our legal system and our society utterly failed.

Tim Wu

Full article in the New Yorker.


My friend David Isenberg cautions us not to think of this as Aaron encountering one bad apple in the system. Rather, says David, “The legal system was working just like it always works…The case of US v Swartz was business as usual.”

Comments Off on Tim Wu on prosecuting Aaron

January 17, 2011

Stupid lawyer tricks

I’m greatly enjoying not only Cory Doctorow’s bang-on response to a truly dumb, threatening legal letter, but the BoingBiong reader’s comment thread about it. [SPOILER ALERT: Is Academic Advantage a scam or is not Academic Advantage a scam? You can decide for yourselves whether those three words — Academic Advantage scam — belong together.]


August 18, 2009 has a Firefox extension that both gives access to public docket records and makes them actually publicly accessible. The courts charge for access to these dockets, including every time you search and for every page of search results. The system is called PACER. RECAP gives you access to PACER (and is PACER spelled backwards). When you use RECAP to view a docket through PACER, RECAP uploads it into the Internet Archive, since the docket info is in the public domain even though the courts charge you for accessing it. The next time someone goes through RECAP to find that docket, she’ll get it for free from the Internet Archive. RECAP also adds helpful headers and other metadata.

RecapTheLaw comes out of the Princeton Center for Information Technology Policy. Well done!

[Tags: ]


July 31, 2009

Tenenbaum trial bloggage

Marc Bourgeois is doing some excellent blogging of the RIAA v. Tenenbaum trial. Fascinating.


Comments Off on Tenenbaum trial bloggage

July 28, 2009

Annals of openness in peril

1. The court has rejected Charlie Nesson’s basic defense of Joel Tenenbaum’s sharing of music files. The case is going to jury which may levy the same sort of insanely excessive fines as in the Jammie Thomas-Rassert trial. I hope Charlie’s team can convince the jury that the fines and the entire process are so onerous and disproportionate that the RIAA has been abusing the court system. Of course, IANAL, and IANAOTJ (I am not on the jury).

2. Barnes and Noble has launched its e-book software. It runs on iPhones as well as on PC’s and Mac’s. I’m having trouble finding which formats it supports, but judging from its Open dialogue, not PDF, .doc, .html, .mobi, or text. It does support .PBD books.

After a very very quick session playing with it, it seems quite competitive with the Kindle, and because I’m running it on my Mac and not on the little piece of crippled hardware I bought from Amazon — the Kindle is just barely adequate as a reader, and is still overpriced by more than 100% in terms of its value, imo — having the use of a keyboard and a mouse is a big step up. And, unlike the Kindle, you can use whatever fonts you have on your machine. Still, it’s only incrementally better than the Kindle’s software (again, on a quick look), not a great leap forward for readers.

One of B&N’s big advantages is that it’s hooked into Google Books, enabling you to download public domain books that Google has scanned in. You do this by searching for a book on the B&N site and noticing the “free from Google Books” label. Be sure to sort by price; otherwise B&N lists the for-pay versions first. If B&N wants to be aggressive in this space (= succeed), it should create an easy-to-find section that lets you browse Google’s free books. Get us using the ereader and then sell us the copyrighted books. (If B&N has such a section, I couldn’t find it quickly enough.)

BTW, I presume (and thus may be wrong) that Google did a special deal with B&N to enable this. If so, I find it worrisome. If Google is going to be granted a special right to scan in books without fear of copyright reprisals, it will be the de facto national e-library, discouraging others from undertaking similarly scaled scanning projects, and thus should be making its public domain books equally and maximally freely available. IMO.

2a. [Later that evening:] B&N stores are now providing free Wifi. Yay!

3. Apple is not permitting the Google telephone service into the Apple App store, thus simultaneously and inadvertently making the case for Zittrainian generativity.

4. [Later that day]: On the happy front, Google has open-sourced an implementation of Wave.

[Tags: ]


April 10, 2009

Pam Samuelson on excessive copyright infringement awards

The abstract of a new paper by the pioneering Pam Samuelson and Tara Wheatland:

U.S. copyright law gives successful plaintiffs who promptly registered their works the ability to elect to receive an award of statutory damages, which can be granted in any amount between $750 and $150,000 per infringed work. This provision gives scant guidance about where in that range awards should be made, other than to say that the award should be in amount the court “considers just,” and that the upper end of the spectrum, from $30,000 to $150,000 per infringed work, is reserved for awards against “willful” infringers. Courts have largely failed to develop a jurisprudence to guide decision-making about compensatory statutory damage awards in ordinary infringement cases or about strong deterrent or punitive damage awards in willful infringement cases. As a result, awards of statutory damages are frequently arbitrary, inconsistent, unprincipled, and sometimes grossly excessive.

This Article argues that such awards are not only inconsistent with Congressional intent in establishing the statutory damage regime, but also with principles of due process articulated in the Supreme Court’s jurisprudence on punitive damage awards. Drawing upon some cases in which statutory damage awards have been consistent with Congressional intent and with the due process jurisprudence, this Article articulates principles upon which a sound jurisprudence for copyright statutory damage awards could be built. Nevertheless, legislative reform of the U.S. statutory damage rules may be desirable.

[Tags: ]

Comments Off on Pam Samuelson on excessive copyright infringement awards

October 28, 2008

Linking to defamation is not defamation

A Canadian court has decided that linking to a defamatory page is not itself an act of defamation. It does leave admit exceptions, such as repeating the content of the defamatory passage or linking the phrase “The truth about Wayne Crookes is found here.”

The chilling effect if the court had decided otherwise would have been positively arctic.

[Tags: ]

1 Comment »

May 6, 2008

[berkman] David Ardia: Citizen Media Law Project

David Ardia is giving a Berkman lunch talk on the Citizen Media Law Project. David begins by acknowledging his colleagues on the project, which has been student-driven to a large degree. [Caution Live-Blogging: I’m missing things, getting them wrong, etc. You will be able to see the session itself at Media Berkman. ]

David begins by looking at, a citizen journalism, the neurodiversity weblog, and wikileaks. These sites have come to the attention of CMLP because they are citizens media sites that have little or no journalism training, little or know knowledge of media law, and not a lot of money. The CMLP grew out of a desire to provide resources for groups like these. (Dan Gillmor was one of the forces behind this, says David.)

CMLP began in April 2007, got a Knight News Challenge Award in May, published its legal threats database in Nov, launched their legal guide in Jan. 2008, and in Feb. did its first amicus filing (for Wikileaks).

The legal guide site has lots and lots of material in it, covering six topics: forming a business and getting online, dealing with online legal risks, newsgathering and privacy, access to government info, intellectual property, and risks associated with publication. There are 5-10 topics under each of these. There’s a lot there.

David walks through the site. There is a rich variety of ways of finding and browsing. In David’s example, the site explains how to create a non-profit corp., and actually steps you through the process, including the specifics for the fifteen states the guide covers so far.

The legal threats database has 25 attributes by which it can be searched. Users can contribute their own entries, although most come in through email. (They also import data from the Chilling Effects site.) The database does not make judgments about the threats. There are 467 entries in the database. Over half are law suits. They include threats to bring criminal charges (16) or to bring disciplinary action (18); that last is included because the legal system backs up the contracts that permit disciplinary action. David explains that the site takes an inclusive approach since you can easily narrow your queries to the areas that interest you. [A good “miscellaneous” principle!]

Factoids: California, which has 12% of the population, is the source of 21% of the threats. 30% of the legal claims are for defamation. Copyright infringements come in second with 8%.

93 of the law suits are pending. 40 settled. The plaintiffs got an injunction in 16 of the cases and won their cases 13 times. That’s not a lot out of more than 250 cases. David says that these sorts of results are fairly normal for law suits, although (he adds) these tend to be emotion-driven litigations, not money-driven.

David gives us a tour of the iBrattleboro case entry. It’s a very well-organized, thorough research on the topic.

David ends by posing some questions for expanding the database and opening it up. [Tags: ]

1 Comment »

April 24, 2008

Citizen Media legal guide

The Berkman Center’s Citizen Media Law Project has a site that’s rich with information, written in non-legalese, about your rights and liabilities as a blogger (and general citizen media person) in the U.S. There’s lots to browse there, and it’s all quite concise and helpful.

For example, the section on whether it’s legal to record a phone call you’re having with someone else says, in part:

Federal law permits recording telephone calls and in-person conversations with the consent of at least one of the parties. See 18 U.S.C. 2511(2)(d). This is called a “one-party consent” law. Under a one-party consent law, you can record a phone call or conversation so long as you are a party to the conversation. Furthermore, if you are not a party to the conversation, a “one-party consent” law will allow you to record the conversation or phone call so long as your source consents and has full knowledge that the communication will be recorded.

In addition to federal law, thirty-eight states and the District of Columbia have adopted “one-party consent” laws…

This is an excellent resource.

[Tags: ]

Comments Off on Citizen Media legal guide