The Declaration In the Age of General Purpose AI

33:39And another that prohibited pharmacies from selling data about physicians' prescription records. Therefore, unless the court recognizes that natural person's speakers have special status derived from the foundational assumptions of the Declaration of Independence and the Bill of Rights, any legislative effort to restrict the use of deepfakes for political or commercial purposes, subliminal techniques aimed at deception and manipulation or bias are unlikely under current doctrine to survive First Amendment scrutiny. Even the multi-member corporations which you know of course AI firms are are ones that have natural persons on them and these are composites of human decision-makers They're not AI models that come out with these algorithmic outputs.

34:40I now turn to how the Declaration of Independence principles can help us evaluate arguments of several prominent scholars who argue in favor of one day granting superintelligent AIs political and free speech rights. I will also respond to several anticipated counter-arguments. Our first task is to determine whether AI's ability to produce understandable information and to respond to inquiries granting in person, personal, warrants granting them personal status within the polity that is currently enjoyed only by natural persons endowed by their unalienable right to life, liberty, and the pursuit of happiness.

35:26Algorithmic models, it turns out, lack conscious intentionality that 18th century authors believed to be intrinsically connected to human pursuit of happiness through ordered government. AI systems lack intentionality, renders them poor candidates for free speech rights despite their informational value to audiences. Courts reviewing regulations on these technologies should not rely on the highest level of scrutiny for data produced by machine learning algorithms. Contrary to my position, a widely popular, strong AI theory equates computation with human thought. It typically relies on output theory of speech.

36:08Its constitutional value is equated with listeners' interests in acquiring information. These scholars discount, and even more usually ignore, any notion that humans might have a unique consciousness that separates them from machines. These output arguments about the computational nature of speech were most influentially advanced by philosopher and mathematician Alan Turing in his seminal argument, Computing Machinery and Intelligence, where he rhetorically asked, may not machines carry out something which ought to be described as thinking? Yet rather than engage in the counterarguments that would have recognized distinctions between imitation and thought, he adopted a functionalist notion of intelligence.

36:55But imitation is not reality. As other authors point out, LLMs are trained to imitate the outward forms of human speech, the sounds, the sequences of words they say, but they're not capable of understanding the subjective thought and experiences behind the words. they're not even designed to try. Instead of investigating the nature of individual cultural and legal thought, Turing is dismissive and therefore fails to explain why algorithmic models should be treated as actual intelligence. Turing's influence on constitutional theory is vast, it's lasting, and it's extensive. Several authors, including Professor Matthias Risi, Simon Goldstein, Goua Krishnakumar, and Peter Salib argue that AI one day should be treated as a competent member of the political community.

37:58Risi, for example, writes that AI machines might eventually have moral status in our political community and presumably be covered by the terms of the Declaration. While he does not provide the details of what computer participation in a public realm would entail, the implication is that LLM should one day be superhuman legislators and voters. I mean, after all, they're going to have some presumed superintelligence, better at making laws, better at deciding. So intelligent would these machines be, Recy adds, that they would no longer permit for easy switch-offs. Taking this claim to a reductioid absurd conclusion, it would require criminal charges for anyone that would turn off a computer.

38:48In fact, this is really quite a model that people are talking about. Arthur C. Clarke explored the dangerous absurdity of treating machines as moral humans capable of feeling in his 2001 A Space Odyssey. The novel includes a scene wherein a machine named Hal begs the lone surviving astronaut not to shut it off after it killed other astronauts aboard a spaceship who had compromised its ability to carry out a pre-programmed outer space mission. Clark's machine, which he tells the reader can reproduce most of the activities of the human mind, remains a thing of fantasy. And yet the notion that computers are sentient beings worthy of respect has morphed into the assertion in some circles that claim moral qualms about switching off machines, which is gaining popularity.

39:48and my students have told me there is entire Instagrams of people showing off my boyfriend AI is better than your boyfriend AI. This is an actual Instagram. I've seen the texts. I had a student tell me that she, every time before she answers her boyfriend's text, she consults with the AI. These are pseudo romantic things. In a similar vein, other authors conceive of a future where super-intelligent machines will require extending constitutional and civil rights to machine learning models. Professor Larry Solem, for example, is functionalist in his approach. He argues that any artificial intelligence that exhibits characteristics of human speech deserves constitutional recognition of their outputs.

40:39This perspective equates GPAI's language tokens with speech of volitional, conscious, natural persons. Professors Tony Massero and Helen Norton also take a listener-centered approach venturing to conjecture that in the future strong AI machines will enjoy free First Amendment rights because they imagine a machine will eventually be able to think independently of human engagement. Among other scholars who take this listener-based approach rather than a theory centered on conscious speakers are Professor Eugene Volok, Mark Lemley, and Peter Henderson. They say that they are inclined to say that generative AI should be entitled to First Amendment protection because of its informational benefits to humans.

41:28They claim that computer generated speech should be protected because of the benefits such outputs provide for people. The authors are indubitably correct about the informational significance of LLMs and their widespread popularity. An increasing number of Americans rely on digital feeds to obtain and augment information about politics and moral decision-making. Yet consider the implications of this position. If LLMs were to enjoy First Amendment rights, any regulation on them would be, would regulation of them as related directly or indirectly to speech would be subject to the most stringent judicial scrutiny and to be presumptively unconstitutional.

42:18Hence all but the most trite regulations of LLMs would likely be doomed because a court, as some recent federal district courts, would find that there's a First Amendment right against these statutes being passed. There's one in California, for example. that was found unconstitutional on First Amendment grounds, having to do with misinformation at times of elections. GPAIs lack the purposeful and conscious human capacities that are intrinsic to the national ethos set down by the Declaration of Independence. LLMs run on zeros and ones over transmitters. they lack the most basic attributes of personhood essential to coverage under the First Amendment or other constitutional guarantees necessary to the equal pursuit of happiness.

43:16Any equivocation of syntactical imitation with the self-evident truths of human nature set down by the Declaration of Independence would expand to machines, safeguards guaranteed from the nation's founding only to natural persons. Such an approach would be analytically unsound and historically suspect. The nation's founders understood the human power of speech. Long before 1776 they knew that public discourse was essential for, as they put it in the Declaration of Independence, challenging the long train of abuses and usurpations through institutions of representative government. As we saw earlier, the constitutionally protected right of expression should be understood to be an innate and substantive right of humans.

44:10Public discourse is conducted by autonomous citizens. While machines can process and spit out grammatically sound information, they lack the unalienable intentionality that is presumed by the Declaration of Independence overarching schema of polity of equals meant to secure personal liberty and representative democracy. Autonomous speech is not equivalent to data processing. The Declaration of Independence remains the living testament of how important free expression is to the very identity of the United States. The framers laid down foundational norms of free speech for the nation. The lessons that the document teaches are those about the past, about liberty polls, about the correspondences of the Sons of Liberty, but it also teaches something to us today about the present coverage of the First Amendment, which extends no further than natural persons.

45:14The ideals that are espoused by the Declaration of Independence are informed by history, national ethos, constitutional tradition, and human sensibilities about representative governments, where people can voice grievances without negative consequences. The document created the parameters of sovereign nationhood, where self-aware individuals can meaningfully contribute to democratic self-governance. Thirteen years after independence, the Constitution recognized that the source of national policies are we the people, not computational language models. GPAI neural networks are not identical to neurons found in the cerebral cortex.

46:04To date, neither Newtonian physics nor quantum mechanics explains how neurons process thought. That's unsurprising given the complexity of what's involved. There are 86 billion neurons, or some people say 10 to the 10th power neurons in our brain, with 100 trillion connections to each other. The architecture of the brain has not been charted, nor is there any basis to believe that if that's ever achieved that would explain anything about abstract thought. Philosopher John Cyril points out that AI models merely simulate speech since they have no intentionality. Indeed, GPAIs are incapable of innate mental experiences.

46:56Conscious self-expression reflects personality, empathy, personal happiness, disheartedness, pain, hatred, adoration, dullness, and wit. Put simply, the American principle that all men are created equal does not appertain to GPAIs. They cannot enjoy constitutional rights, in large part because they lack consciousness of self, community, society, culture, and law. Their production of understandable script is the product of natural language processing through probability theory. AI systems lack the basic elements of inward personhood, individuality, self-consciousness, and other mental states that render humanity uniquely protected under the Declaration of Independence and free speech guarantees of the First Amendment.

47:54GPAIs produce syntactically correct sentences through algorithmic computations, but neither the computer nor the program it is running has any comprehension of self-evident truths on which the framers rested their argument about human equality. This conspicuous difference between computation and mental operation speaks to the relevance of the history and tradition of the Declaration of Independence to the debate about whether LLMs deserve free speech rights. Machine models can process chatter, but they lack understanding, intentionality, and consciousness that are essential to the pursuit of happiness, self-fulfillment, and the institutions of government.

48:39The difference between sentient beings and machines could play a role in future development of Supreme Court free speech jurisprudence. Now, let's conclude by briefly reflecting on efforts by other democracies like the European law, which is more sensitive to the dangers that lurk in unregulated artificial models. The EU AI Act defines general purpose AI as something that gives responses to explicit and implicit questions that it has, that makes predictions about content and recommendations about virtual environments. The EU Act does not treat all content identically as the US Supreme Court purports to do.

49:29Rather, the EU regulates AI outputs differently than it would the speech of natural persons. The Act's recognition of human cognitive interests is consistent with the Declaration's emphasis of human events, the opinions of mankind, and self-evident truths about people's equal right to pursue their welfare while establishing a Republican and political entity of human actors who enjoy First Amendment free speech rights to participate in discourse of the day. The EU AI Act recognizes social interests in protecting against harms to human consciousness caused by psychophantic nature of chatbots. It also regulates AI systems that deploy subliminal techniques, something that would not be recognized in the United States and probably would be found unconstitutional under current doctrine, as would other measures of the EUAI Act, such as high-risk things that occur in models, biometric examinations and matters dealing with safety, employment discrimination, housing discrimination, and so on and so forth of that type, which are content-based regulations that are likely to get strict scrutiny analysis under current doctrine, unless the court were to recognize that there is a special status for natural humans.

50:58To show the disconnect with the United States here, a recent district court decision held that a California statute that regulated deceptive deepfakes during the course of election season was unconstitutional on First Amendment grounds. The Declaration of Independence and its cousin, the Constitution, created a nation of equal intentional persons, not calculating machines. The Declaration of Independence remains a binding statement of national aspiration. It speaks to us from the wisdom of past generations and the yearnings of ours. Its words inspired not only the revolutionary generation, but the freedom movements that followed it.

51:41They regarded its statements of equality, representative polity, and unalienable rights to be the innate to natural humans. No thought was put into the notion that the polity with its commitment to freedom of speech would extend to computational machines. Hence extending free speech rights to algorithmic models would be a radical departure from the Declaration's and the Constitution's original meaning. I must admit when I first opened Alex's paper I really began to panic. As Michael's introduction very accurately pointed out my own area of expertise is neither AI nor the First Amendment or speech more generally.

52:27But happily for me and frankly happily for you I think too, much of Alex's argument is built around a series of claims made by a cast of abolitionist characters that I know quite well. And this as Michael also astutely noted has been a kind of theme running through a couple of the presenters' comments today and indeed yesterday. So I want to spend my comments today really focusing on their ideas. Specifically, I want to situate abolitionists' conception of the relationship between the Declaration on one hand and the Constitution on the other hand in the context of other Americans' views on the subject.

53:10I also want to explore abolitionist and broader founding era attitudes towards speech and the suppression of speech. And I want to consider whether there might be a way in which abolitionist views could actually be used to license the very thing that Alex is claiming that they never would. That is, I want to ask whether abolitionists' deep commitment to the right to listen, not just the right to speak, provides fodder for scholars who have argued that AI might, in in fact, be deserving of some kind of First Amendment recognition. I do this all hopefully to facilitate what I hope is a broader conversation about the role of abolition and abolitionists in our debates over the Constitution and the Declaration, because it seems like they're playing an ever escalating role in them.

53:55So I want to get started by talking about abolitionist conception of the relationship between the Declaration and the Constitution. As Alex shows, the abolitionists were especially keen to use the Declaration to advance their cause and repeatedly made reference to its soaring appeals. A little less clear is the extent to which this move, this effort to treat the Declaration as equivalent to and coterminous with the Constitution was distinctive from the way that their opponents tended to engage with it. While abolitionists often reference the Declaration and the Constitution in one breath, defenders of the institution of slavery were far more likely to talk about the Constitution alone.

54:35And indeed, as the abolitionist movement grew in the 19th century, Southern intellectuals realized that the Declaration's language of universal equality was the sharpest weapon being used against them. Their solution was to elevate the Constitution's specific legal protections over the Declaration's abstract ideals. In fact, most post-slavery thinkers eventually came to view the Declaration of independence as a kind of dangerous rhetorical flourish or even a self-evident lie. While they meanwhile revered their constitution as their primary legal shield. What Alex shows us in short is one view of the relationship between the declaration and the constitution and the importance of speech in both texts.

55:21But it's not the only view and I think we ought to reckon with the fact that abolitionists have a distinctive take for a distinctive set of reasons. Nor indeed was the abolitionist commitment to free speech necessarily reflective of the broader community. Again, pro-slavery politicians and thinkers were overly hostile to the spread of, for instance, abolitionist ideas, whether those ideas took the form of petitions to various governmental bodies, public addresses, or various forms of writing. Over the course of the early national and antebellum eras, they actively suppressed all kinds of anti-slavery speech.

55:59But pro-slavery politicians and thinkers were not alone in their willingness to limit speech. 19th century Americans generally tolerated many forms of censorship by the government, whether that censorship focused on political dissenters and government critics, and here I'm thinking of the Alien and Sedition Acts, or the circulation of supposedly obscene material and information about birth control as with the Comstock Act of 1873. If, in an attempt to undermine the notion that AI is entitled to First Amendment rights, Alex is attempting to recover a tradition that first places the Declaration on an equal footing with our Constitution and second, lionizes free speech as foundational to the American project, it seems worth noting that there were also powerful rival traditions to contend with.

56:50We may admire abolitionists, but is it appropriate to suggest that they had a kind of monopoly on the meaning or the conceptions of either, again, the relationship between the Declaration and the Constitution or the importance of speech? Should their views be controlling? Should they be seen as authoritative? And I would say, especially given the relative marginal number, the relative marginal size of the abolitionist movement and in fact the reality that they were a much maligned part of the American population for almost the entirety of their existence. If in fact abolitionists are to be our guiding lights as to the history and tradition of speech rights, however, I also want to push Alex to wrestle a little bit more with the implications of the broadest possible reading of the abolitionist project.

57:41After all, the most capacious understanding of abolitionism may well be that they pushed us to reject orthodoxy and to see things in new ways. Most white Americans would have accepted at the midpoint of the 18th century the people of African descent were inferior, not only of course socially, but most importantly culturally and intellectually. Most moreover believed that the inferiority of black people was innate and immutable. Abolitionists railed against this received wisdom. These old ways of thinking, they argued, were false. I think it could readily be said today that our received wisdom is that speech is a uniquely human act.

58:21How or why that may be so is perhaps in need of more elaboration. But I promise, because I tested this out, I promise if you ask people whether or not AI is deserving of First Amendment rights, you ask ordinary people, that is to say not lawyers, the immediate response is generally confusion followed by a little bit of uncomfortable laughter, maybe some kind of eye rolls. The very premise that AI could have some kind of First Amendment rights strikes them as bizarre, perhaps even deranged. It does so, I would argue, at least in part because we have so fully taken for granted the claim that Alex makes here, the claim that speech is this uniquely human act.

59:02So if the essence of abolitionism was a challenge to the existing order and existing patterns of thought, might not the abolitionists also be read as a challenge to our own orthodoxy about the proper subjects of speech rights? Perhaps rather than focusing on the specific nature of abolitionists' claim, we ought to think about the broader meaning of their claims. And if we do so, might the abolitionists be read as pushing us to be more imaginative in our understandings of who or what could exercise speech? This, of course, would entail reading abolitionists for the spirit of what they said rather than the letter.

59:41But this approach was something they themselves avidly endorsed as a method for interpreting the Constitution. I want to close with one last line of inquiry, one that tries to think seriously about Alex's conception of speech rights, his interlocutor's conception of speech rights, and finally, again, abolitionists' conception of speech rights. Alex's piece is designed as a refutation, of course, of the claim that AI might enjoy speech rights. His focus is on demonstrating that founding era conceptions of speech rights established speech as uniquely human, as something born of consciousness and intent.

1:00:20And Alex largely concentrates on the generation of speech, on its creation. By contrast, Alex notes at one point that the scholars who have articulated the argument that AI, in fact, might enjoy some speech rights, among them Eugene Volokh, have instead mostly taken what Alex calls a listener-centric approach. That is, when it comes to AI First Amendment rights, they have argued that we ought to be thinking about the consumers of AI rather than the producers of AI. As Volokh et al. have argued, the strongest argument for First Amendment protection stresses not the AI or its corporate owner as speaker, but rather the interests of the listener in receiving meaningful communication.

1:01:05Abolitionists, meanwhile, evinced commitments to both the right to speak and the right to listen. They certainly, as Alex's paper suggests, care deeply about their own rights to speech, by which I mean their own right to create speech. But abolitionists also seem to care deeply about their right to circulate those ideas. Their right to be heard or listened to. They could also, like Alex's interlocutors, be, I think, quite listener-centric. Here are two particular examples come to mind. We might first think about the abolitionist postal campaign of 1835. This was a direct mail effort by the American Anti-Slavery Society to flood the southern United States with anti-slavery literature, the sense here being that those individuals in the South had essentially a right to listen to abolitionist thought.

1:01:54Second, we might consider how abolitionists responded to congressional efforts to immediately table, without discussion, the abolitionist petitions that were submitted to that body between 1836 and 1844. This so-called gag rule ultimately backfired because abolitionists successfully used the suppression of their petitions as proof of the existence of a kind of slave power that was conspiring to strip white northerners of their civil rights. Here, abolitionists relied not on the argument that they had a right to be heard, but also on a claim that the gag rule constituted some kind of tyrannical suppression of the public's rights to know and to debate the morality of slavery.

1:02:35As Frederick Douglass asserted in an 1860 speech that addressed broad efforts to stifle the abolitionist message, free speech wasn't just about the rights of the speaker. Equally clear, he argued, is the right to hear. To suppress free speech is a double wrong. It violates the rights of the hearer as well as those of the speaker. It is just as criminal to rob a man of his rights, to speak and hear as it would be to rob him of his money. I think Alex could do a little bit more in his paper to explicitly engage with this abolitionist commitment to the right to listen and consider whether by his own lights it might actually end up strengthening some of his interlocutors' claim.

1:03:16That is, I would press Alex on whether an abolitionist conception of the right to listen creates grounds on which Volek et al., if they were so inclined, might actually stand. I offer these reflections in the hopes that we might have a robust conversation, not just about AI and the First Amendment, and here I rely on you, the audience, for some great questions in those specific directions, but also about the place of abolition and abolitionists in our contemporary conversations, both about the Declaration and the Constitution. Thank you.

1:03:54Well, I think those were excellent comments, and I really appreciate the thought of how to improve the argument, particularly with your last point about the importance of the right to listen, the right to information. I think that's critically important. Clearly, if we think of speech as being something for public discourse and for self-government, The right to listen, certainly to obtain information, was absolutely critical. It was critical to the framers who would send out these correspondence letters and who had to communicate with each other. And so it's essential. And it's clearly a part of self-government.

1:04:33So undoubtedly, the protection of free speech in the First Amendment must include listener rights. And I love the idea of using the gag rule and the postal restrictions and the Jackson administration for that. Where my particular perspective on the First Amendment is, though, is that it is speaker-centered, that its primary purpose is speaker-centered. So this is distinguished from the Mikkeljonian version of the town hall informational type of concept of speech. so Michael John envisioned speech as being something free speech as being something like a town hall where there was an order, there's a microphone people take turns to speak they don't speak out of order they don't say things like fuck the draft as you have in Cohen but that in public discourse our ability to pursue happiness our ability to express ourselves is really what's at the core of free speech And so sometimes we do put restrictions on listeners.

1:05:42For example, let's say I'm the owner of a pharmaceutical company and I happen to know there's going to be, we're a major publicly traded pharmaceutical company and I know there's going to be a major lawsuit that's going to drop tomorrow. It's going to be all over the news. The markets are going to be closed and our stock is almost for sure going to drop, like all get out. Well, and I want to tell my mother because I've had my mother purchase lots of this stock to sell before the close of the market. I can't tell her, even though she clearly has an interest. The same thing is true about securities regulations, right?

1:06:19They have certain restrictions on the things that a listener might have an interest in order to find out. In some ways, they expand it. In some ways, they contract it. So I think that the right to listen and the right to receive information is of central and critical importance to the First Amendment, as Andy so astutely points out. But that the real function of the First Amendment at its core is really the protection of our ability to express our intentions, our conscious views, and our developments of public discourse and the advancement of self-government and personal expression as well. This has been a special presentation from Stanford Legal.

1:07:07I'm Pam Carlin. See you next time. Kin - bastante