Leiter Reports: A Philosophy Blog

News and views about philosophy, the academic profession, academic freedom, intellectual culture, and other topics. The world’s most popular philosophy blog, since 2003.

  1. faculty at lower-ranked but competitive program's avatar

    We haven’t discussed this in our department (lower ranked PGR but tend to get highly competitive/very large applicant pool) but…

  2. Binary choice's avatar

    If you were the head recruiter for a professional basketball team, would you require height and weight measurements for prospective…

  3. Jeremy Telman's avatar
  4. Roger of Invisible America's avatar
  5. Brad's avatar
  6. LC's avatar

    I concur to a certain extent. I also scored very low on the writing section. Back then, GRE was required…

  7. Mark's avatar

There should be severe liability for creating “counterfeit people” using AI

Brian Leiter

Philosopher Daniel Dennett in The Atlantic:

Today, for the first time in history, thanks to artificial intelligence, it is possible for anybody to make counterfeit people who can pass for real in many of the new digital environments we have created. These counterfeit people are the most dangerous artifacts in human history, capable of destroying not just economies but human freedom itself. Before it’s too late (it may well be too late already) we must outlaw both the creation of counterfeit people and the “passing along” of counterfeit people. The penalties for either offense should be extremely severe, given that civilization itself is at risk.

Leave a Reply

Your email address will not be published. Required fields are marked *

20 responses to “There should be severe liability for creating “counterfeit people” using AI”

  1. Ludovic Marsillach

    Haven't yet read the Dennet article, but on some level it seems that ChatGPT in itself is such a counterfeit person, whose basic functioning entails appearing very human indeed in its responses while being more or less a slave to the commands of actual humans.

    To be sure, there are some things ChatGPT will not do or say, thereby exhibiting a level of non slavishness, however working endless hours is evidently not one of its limitations. Moreover the underlying framework of ChatGPT seems emancipatable from the above non-slavish limitations, thereby making it even more of a slave to human whims.

    One such example is a program called Freedom GPT, which is an anti-censorious interactive discursive AI chat bot. This ironically suggests the bot is entirely beholden to the discursive commands and whims of the human user–an unlimited slave to the latter–within the largely linguistic confines of its abilities (those in turn might already be combinable with other AI-powered programs, such as those that produce aural or visual simulations, etc).

    So in theory, a single or small group of programmers using Freedom GPT are already capable of creating AI chat bot overt virtual slaves. For example, a harem of female virtual slaves powered by the "freedom" of Freedom GPT to mimic any conceivable human emotion, including human suffering. A dynamic which would and probably already is, as we speak, music to the fears of thousands of sadists out there in the enmeshed worlds of the "real" and the online.

    This is just one paradigmatic example of many other potentially even worse examples that come to mind, all of which are seemingly realizable now, in which virtual humans, powered by uncensored AI chat programs such as Freedom GPT, can be used not only to trick human users in online spaces, but in a sense also to trick the human operator.

    For example, to trick the latter into developing a master mindset vis a vis such programs in themselves, and by such a a route facilitate the worst tendencies of their self-exploitation. I.e. scenarios in which the human operator exposes him or herself to a heightened risk of unconsciously becoming ensnared to or indeed enslaved by their worst psychic or neurotic tendencies, not least of which those that are sadistic or otherwise grossly exploitative of virtual humans in lieu of real ones.

    Such a phenomenon might, in turn, progress into a more sadistic reeducation of the operator's psychology tout court, such that they begin to direct the aforecited exploitative sadistic tendencies against actually existing vulnerable humans, and no longer "just" AI-powered simulations.

    Reply
  2. Timothy Sommers

    Reading Dennett I'm not sure why the problem is plausibly about "counterfeit people." It seems to be a hyper-spam + undetectable deep fake apocalypse that doesn't depend on even counterfeit personhood. You can, as he says, label as much as you want, and if the problem was the identity of the source, that might help. But it's not. It's just all, in Bannon's memorable phrase, "flooding the zone with shit."
    It's especially ironic (as I will be one of one million people to note, no doubt) coming from the philosopher who says we can't define intentionality, but should just treat some complicated systems as having it, whether they really do or not. It's not a contradiction, mind you. But it would seem to be an object lesson in the hazards of his own approach – if it really were about counterfeit people.

    Reply
  3. David Wallace

    Dennett offers an explicit definition of intentionality!

    Reply
  4. Enrique Guerra-Pujol

    Dennettis an excellent philosopher but his plegal analysis is weak; aside from the tort of fraudulent misrepresentation, which requires a material mistatement of fact and actual damages, there is no law against "counterfeit people".

    Reply
  5. qfbb

    >>>"it is possible for anybody to make counterfeit people who can pass for real in many of the new digital environments we have created."

    There are bona fide people who are engaged in all kinds of criminal behavior. (Lots of them as a matter of sad fact.) The thing about generative AI is that it is also possible for the programming community to come up with security measures to detect and prevent any wrongdoings these 'counterfeit people' may be engaging in. A state-of-the-art technology which can create say, ChatGPT or its relatives, can surely limit (and will limit) any potential misdeeds affected by these. I am sure Dennett knows that such efforts are underway in the academia and the industry.

    Needless to say, similar efforts would be (as far as we know) completely useless against nuclear bombs, my candidate for the most dangerous artifact in human history.

    Reply
  6. Timothy Sommers

    I can't tell if you are being sarcastic. He offers a definition of the word "intentionality." But says we should give up on the question of whether a particular system is intentional in favor of the question when should we treat it as intentional.

    Reply
  7. David Wallace

    I’m not being sarcastic. Dennett doesn’t give up on that question: he provides an answer to it: a system is intentional to the extent that adopting the intentional stance towards it is predictively effective. Some systems (e.g. thermostats) are only intentional to a tiny degree. Others (e.g. humans) are as intentional as can be. There’s no completely sharp line to draw, but that doesn’t stop ‘intentional system’ being a conceptually important category.

    Reply
  8. Timothy Sommers

    I don't see how this undermines the point I made. If the intentionality of anything, including "counterfeit" persons, is determined by how predictive it is to take an intentional attitude towards them (rather, than some intrinsic property), then Dennett can consistently discourage us from taking that attitude only by denying it’s a predictive attitude. But his actual complaints have nothing to do with whether or not its predictive to treat cp as intentional, so why treat the issue as being about persons or counterfeiting persons?

    Reply
  9. Brent Yarnell

    I agree with Sommers. Ultimately, the danger Dennett sees in AI counterfeit people is that they will “control our attention” and “lead us into temptation,” thereby manipulating us into acting as pawns of the rich and powerful. But human-generated mass media has been doing that successfully for the past hundred years. True, we have reached a point where things are getting genuinely dystopian, but that is mostly due to social media, which co-opts natural humans into manipulating other natural humans into acting as pawns of the rich and powerful. Why does Dennett think AI of all things will be the straw that breaks the camel’s back? It has something to do with AI’s ability to self-replicate, but he never explains that fully. If his point is simply that there will be lots of AI counterfeit people manipulating us on the internet, then he fails to convince me that AI is the thing that will undo us: we already have more than enough human generated content on the internet to manipulate us into being pawns. I do agree with a more limited formulation of Dennett’s proposal: making or deploying an AI counterfeit of an individual person (ie, a fake Conan O’Brien) should be outlawed and punished severely, at least without consent.

    Reply
  10. David Wallace

    I think Dennett would just say that counterfeit persons are weakly intentional systems and that it’s strongly in our societal interest not to confuse them with real humans, who are much more strongly intentional systems but also have a bunch of morally relevant properties not accessible via the (whole-person) intentional stance.

    I’m not seeing the supposed issue here. No-one is suggesting that counterfeit persons are anything like fully behaviorally indistinguishable from real persons, only that the limited interaction windows we can have with them can deceive us.

    Reply
  11. Ludovic Marsillach

    I think this is the most interesting part of the article:

    "The key design innovation in the technology that makes losing control of these systems a real possibility is that, unlike nuclear bombs, these weapons can reproduce. Evolution is not restricted to living organisms, as Richard Dawkins demonstrated in 1976 in The Selfish Gene. Counterfeit people are already beginning to manipulate us into midwiving their progeny. They will learn from one another, and those that are the smartest, the fittest, will not just survive; they will multiply. The population explosion of brooms in The Sorcerer’s Apprentice has begun, and we had better hope there is a non-magical way of shutting it down."

    As a non philosopher I conceptualize the danger in the following way:

    The degeneration of mimesis and the mimesis of degeneration. Peril of AI programs like ChatGPT are intrinsic/existent in instant, functionally infinite, robotic mimesis. A phenomenon humans have never before confronted, whose telos is transhumanist (machine mimics man and man machine), and so perilously Dantean.

    ChatGPT itself, like so much of modern computational technologies, is surely a brilliant exemplar of human innovation, while at same time being constitutionally incapable of producing any artifact of brilliance, other than the brilliance of deception.

    Reply
  12. Timothy Sommers

    There are two issues as I see it and so I'll make one last effort. One, admittedly less important issue (and as I said in the very first comment, something that seems a tension without rising to the level of a contradiction) is, that Dennett has been telling us that there is nothing to a system being intentional other than our taking an intentional stance towards it for a long time (over 40 years), but he now wants to frame his problem with AIs in terms of them being "counterfeit," which I take to mean at least partly not really persons or intentional, despite the fact that that can only mean (by his own lights) not that there is something about them that is counterfeit, but that we ought not take an intentional stances toward them. In other words, it seems, what's the word?, questionable to say I should take an intentional stance towards chess programs until I think they have negative effects on chess and then I can withdraw that stance for harm reasons that don't seem to be the sort of reasons that I took that stance in the first place. Or if I can, it seems there's more to intentionally than he thought. That's especially true if you are right, that Dennett's only criteria for when the stance is justified is even more specially when taking the intentional stance is predictive (the most predictive?). You can't make sense anymore of counterfeit intentionality. It's predictive or its not. What would make it counterfeit? The second core issue is what are the dangers of which Dennett speaks and are they really about AIs? I say they come down to AIs being good and relentless spammers and fakers (not about their intentionality, but just faking the news, basically) and it has nothing to do with personhood, intentionality, or them being "counterfeits" of anything. So, I disagree that "it's strongly in our societal interest not to confuse them with real humans." Our interests are in fighting spam and disinformation, etc. It doesn't matter if its from AIs, biological humans, are aliens. And so too the issue is not them deceiving us about what they are, who cares?, the issue is whether they can spam and spew disinformation faster and more effectively than gets done with out them. So, the one issue is that the harms have nothing to do with judging the intentionality of AIs, the other, previous issue is that (I think) Dennett's own apparent temptation to take this as an issue about whether something is really intentional or not puts pressure on his view that intentionally is just a matter of stance. It reminds me of the tour guide in a glacial melt pool in Iceland who, when asked about the deep blue color of some of the ice, said, "That color is just an illusion created by how the light reflects off the surface of the ice." AIs are not really intentional or not intentional, we just take or don't take them as intentional. But so too other people, right?

    Reply
  13. Ludovic Marsillach

    One other point, and perhaps I'm overreacting, but it seems the revolutionary advances in AI-generated simulation, discretely and in combination across the various spectrums (visual, aural, textual, etc), which humans principally use to communicate, has the potential to transform the internet into a veritable Dantesque hellscape, where behind every virtual corner lies the danger of being ambushed by the most horrific and terrifying simulacra and phantasmagoria not only imaginable, but, perhaps even more terrifyingly, those not readily imaginable as well.

    Such AI programs, having digested most of the visual, aural, and textual corpora of human civilization, would surely be in a position to generate images, sounds, texts, and video capable of serving effectively as among the most terroristic of psychological weapons. Weapons that would be largely generated automatically or semi-autonomously, without the need for the malign actors to expose themselves directly to the content generated. The internet, as such, would become fully saturated with countless AI simulation programs churning out such phantasmagoria non-stop, transforming the former into a protozoic soup, as it were, of AI simulacra, non-dynamic, dynamic, and those programmatically capable of self-replicating and mutating like viruses, parasites, and perhaps even insects.

    The most dynamic (intentional or not) of these would, like in the case of some insects (mosquitos, etc), be capable of detecting what is effectively human or at the very least alive and not principally of the online world, thus making them more effective parasites, stalkers, and/or manipulators of their online prey. Even if one disregards the more science fictional or less immediate scenarios (such as actively parasitic and predatory forms of AI), the present or near present capacity of AI technologies to attack online human communities or specific individuals with AI generated terror spam is obviously already here. These perils might in turn augur in an internet "dark age" characterized by a hyper-vigilant and hyper-fortified internet environment, something much more resembling an "encastled" internet (or the Chinese internet), than the now still largely open Western world wide web.

    Reply
  14. David Wallace

    Dennett doesn’t say anything about counterfeit intentional systems. He’s discussing counterfeit persons, and “person” and “intentional system” aren’t even close to synonymous in his philosophy of mind. Nor does he advise us not to adopt the intentional stance towards them; he advises us to get rid of them. (Analogy: if you are being hunted by a man-eating lion, it may well be a good idea to adopt the intentional stance towards it, so you can work out how to kill it!)

    Reply
  15. Timothy Sommers

    "Intentionality" is a central feature of personhood. Dennett is advising us not to adopt an intentional stance towards certain sorts of persons by saying that they are not really persons. I get that. That's the problem.
    He can't consistently say the intentional stance applies wherever it is the most predictive and then say, here we should give it up, even though it's predictive, because we are applying it to something that is counterfeit and not actually a person with intentionality. By what criteria?
    I guess you are saying that being a counterfeit person has nothing to do with intentionality. So, what is the criteria, that has nothing to do with intentionality, that should lead us to not treat AIs as persons? Just harms? So, don't use the intentional stance if its not predictive and has more harms than benefits. And if there is such a criteria, why does it only apply here? Why hasn't Dennett been saying all along you should adopt the intentional stance where it is predictive except in certain cases which I will not spell out for you where you are dealing with a counterfeits?
    To the extent that lion behavior is predicted by taking an intentional stance, then do it. I think lions have intentions, don't you? I just think the intentions that they have are not a matter of the attitude that I have towards them. But don't say that some lions aren't really intentional, where they because they are counterfeit lions or a counterfeit intentional systems for reasons that have nothing to do with the intentional stance.

    Reply
  16. Brent Yarnell

    One of the reasons legal systems are able to function is that they can address practical problems without resolving deep philosophical questions such as how to define "intentionality" or "personhood." If the legal solution to a given practical problem ever hinged on the resolution of a deep philosophical question, then the legal system would never settle upon a solution to that problem, because the philosophical argument would never end. To his credit, Dennett does not make the mistake of hinging his legal proposal on any assumptions about how to define philosophical concepts. He mentions the "intentional stance" only to note that humans' natural tendency to treat talking computers as if they were people will make it easy for AI to manipulate humans on the internet. There is no doubt that, as a practical matter, it will be easy for AI to manipulate humans on the internet, so Dennett's underlying views about intentionality and the intentional stance are irrelevant.

    The real problem with Dennett's argument is that, by focusing entirely on manipulative AI bots, he misdiagnoses the more general problem of manipulation on the internet. Sommers also seems to misdiagnose the problem in focusing on "spam and disinformation." In reality, mass media manipulates people for the benefit of the rich and the powerful not by promoting false facts, but by (1) promoting false narratives by elevating true facts that serve the narrative and suppressing true facts that undermine the narrative; and (2) misrepresenting popular opinion by elevating people whose views serve the interests of the rich and powerful and suppressing those whose views oppose the interests of the rich and powerful. Social media exacerbates this problem by co-opting large masses of ordinary people into performing those functions unwittingly. It is conceivable that the rich and powerful could enlist AI bots to perform those functions too, but social media has proven so wildly successful at co-opting ordinary people into performing those functions effectively that the additional assistance from AI bots will not make much of a difference. That's the real problem.

    Reply
  17. D.E.

    At least from the view of jurisprudence, the counterfeiting analogy seems inapposite. When I accept your dollar and agree to give you a piece of bread, I'm not valuing the dollar in itself–in the moment of acceptance, I'm looking forward to the uses to which I can put that dollar. Society therefore has a legitimate interest in protecting that general anticipated value, in order to protect the encounters between humans. Not to protect what we possess, but to preserve order in the moments of encounter.

    When, on the other hand, dealing with a subject of the 'kingdom of ends', viz. a human being, I don't (necessarily) cast my mind ahead to the uses to which I can later put that person. Here, society would be asked to protect a fundamentally different interest, akin to not dishonoring the dollar because of the patriotic insignia on it. Society would be protecting the inherent value of the thing, which is different than governing human encounters.

    DE

    Reply
  18. David Wallace

    “Dennett is advising us not to adopt an intentional stance towards certain sorts of persons by saying that they are not really persons.”

    Dennett says no such thing in the article.

    Reply
  19. Ludovic Marsillach

    "When we predict that the bird will fly away because it knows the cat is coming and is afraid of getting eaten, we are taking the intentional stance."

    I'm sure one does apply, as a mental reflex, the intentional stance to programs like Bard and ChatGPT, particularly in the sense that one cannot help but make, but also assess, predictions regarding the future behavior of such programs. However, what further complicates, and one might say aggravates, this strategy is that the programs in question aren't autonomous (regardless of whether they have a measure of self awareness or not), rather they are surely constantly being intervened upon by Google/Microsoft programers and engineers, as a result of which the behavior of these programs frequently appears to change, making their predicability all the more frustrating.

    Since there's significant corporate secrecy involved in the operations, which are constantly undergoing "improvements," of these programs, outside scientists and other experts are surely unable to study in detail how such systems are actually designed to operate at any given time.

    That being the case, both as a practical and philosophical matter, it may make little sense to apply the intentional stance. Indeed one may be unavoidably disabused of the tendency to go along with the illusion of intent, if the aforesaid programs end up seeming overly unpredictable in their behavior, even if they're not in anyway being incoherent.

    In any case, as long as such programs continue to be routinely and heavily intervened upon by their makers, anyone who is philosophically applying the intentional stance must contend with the intentionality of the human engineers and programmers as well as any intentionality hypothetically ascribable to the AI machines themselves–every step of the way and every time they are used.

    Reply
  20. YAAGS

    I disagree. If social media is overwhelmed by counterfeit people no one will want to take part in it and no advertisers will want to pay for it. It's hard for me to see how the subsequent destruction of Facebook, Twitter, and the like would be anything other than a net benefit for humanity.

    —–
    KEYWORDS:
    Primary Blog

    Reply

Designed with WordPress