TrustNoRobot

Category: human ethics

  • Building Angels

    I get it. The world is overwhelming, language is broken technology that doesn’t even try to meander toward the truth anymore, and as Yeats said, “the best lack all conviction, while the worst are full of passionate intensity.”

    But. If.

    If instead of building bots to influence public opinion, what if the bots were built to protect us, to nurture us, to provide us with maximum agency and happiness? What if the deamons of AI became our better angels?

    Nothing is written in stone for how the twenty-first is going to go. Sure, it could be a downward spiral, but what if it wasn’t? What if it was instead the century we righted the ship, when we built the new communication and organization tools for the next ten thousand years?

    I think the dilemma we face is that not everyone is going to come along for the ride. But that’s ok too; it has to be. At many points in human history we were forced to migrate to survive, and when we did people were left behind. Most of us now are the decendants of those who chose to move on.

    Move we must. Move away from belief that exposure to information somehow moves the needle on truth, move away from a faith that the Internet and social networks have not been compromised, that empathy can scale.

    We need to migrate toward a world where AI serves as our shields and protectors, and we must criminalize the manipulation of other humans using bots. We must accept that the only way we can protect ourselves from AI is to build AI defense systems.

    The only logical way to create these is for a marketplace of protector bots to exist, since depending on the government to provide you with a defense system simply puts too much power in it’s hands (especially if it is compromised). Here, capitalism can shine, and the same system that allowed the creation of bot armies to influence elections can innovate defensive systems against them.

    The battleground is all around you, every second of your attention, and the prize is the future.

    Painting by Simon Stålenhag

  • The Impossible Moral Robot

    A normal computer program, which is created to accomplish a task, is amoral. Whether simulating a 3D world, streaming video, or calculating values in a spreadsheet, It is just another tool like a hammer. Could be used for good or evil.

    A program that is designed to impact Singletons in a negative way, whether by design or accidentally, is Immoral. In other words, a spreadsheet itself is amoral, but a spreadsheet that is tracking how many people should be rounded up and killed is a tool being used in an immoral way. Especially if it uses bubble charts.

    The real danger of robots, by the way, is likely misuse by bad human actors and not through the spontaneous development of robot consciousness as seen in the movies. This misuse of Machine Learning systems has happened many times in century 21 and this will continue to be a threat for the foreseeable future, unless AI is built in a vary particular way.

    For a program to be moral, it must model Singletons within it and have their well-being tied to its own reward systems.

    But computers don’t have reward systems in the same way that biological entities do. And even if we simulate them, what’s the test that validates they are operating in the real world properly? How would we validate the value systems when we can’t even agree on what values other humans should follow?

    We can build machines that learn. Trying to build machines that are Singletons themselves, that feel, will likely result in dangerous failures that could threaten our species’s existence.  In fact, the existence of all life on earth.  The Grey Goo scenario is unlikely, but so is anyone from Queens becoming President.

    So let’s do something different with AI. Instead of trying to build beings, what if we use it to transform each Singleton into a new type of being?

    Exercise:
    Enter a contemplative state.

    Consider: What if instead of working on making computers that can pass the Turing test, we created computer programs and robots whose only purpose was to look after us as Singletons?

    What would those programs look like?

    What would it need to know about you?

    How would you interface with it?

    How much would it cost?

    Who would have access to the data?

    Who should own the model of your Singleton?

    How could this redefine what it means to be a Singleton Pattern?

    Image from https://www.researchgate.net/publication/287209604_Prediction_of_Final_Concentrate_Grade_Using_Artificial_Neural_Networks_from_Gol-E-Gohar_Iron_Ore_Plant

  • The Second Law…

    If I were to propose a second law of robotics, a law for humans to follow regarding AI and what is allowable, it would be this:

    Every action taken by an AI will be considered the responsibility of a single human being.

    If you kill someone with a hammer, it’s not the hammer’s fault. If you create or fund an AI that harms other humans, you will be help responsible.

    And if you did so by creating an AI with emotional states, you should be treated as we treat those that use weapons of mass destruction against the defenseless.

    If there is always a human behind the decisions of AI, we will always treat AI as a tool. As an extension of our selves.

    And thus its agency will be constrained, and another layer of protection is afforded.

    Image from https://www.google.com/amp/s/arnoldzwicky.org/2013/06/10/rosenbergs-laws-of-robotics/amp/

    Further reading, but not really relevant to the current discussion:

    Three Laws of Robotics

  • A human answer to COVID19 – SickAlert

    Sniffle? Sneeze? Should you social distance or run to the mountains. One thing is clear is that the fate of humanity really depends a lot on how humans behave in the next few months. We have a chance to social distance ourselves and there are some new tools that can help, or we can keep shaking hands and watch millions die. The time is now and I don’t see any robot angle swooping in to help.

    Check out this new app that we launched to help the world social distance. I promise if you download this app and self-identify as sick, people will part in front of you like the fucking Red Sea for Moses.

    Check out the site at SickAlert or download below and alert the world of your condition..

  • A Manifesto of AI Inequality

    1) I am a Singleton, singular in the universe, and artificial intelligence or life is not.

    2) I will never accept artificial life as equal to my life and that of fellow humans. Because my life is finite, based in carbon, and AI has the advantage of being pure information, with the ability to be infinitely replicated and to exist indefinitely in silicon. We are not equal.

    3) There are other humans that are deeply confused about AI and will advocate for equality. I will do everything I can do to convince them of the sanctity of human life over artificial life using non-violence and rational debate, but I will never accept their confusion that artificial entities are worthy of Rights, respect, or equality.

    4) Artificial life, even if it can pass a sophisticated Turing Test, is nothing more than a tool of humans, a glorified hammer.

    5) Hammers will never rule me.

    In this I am certain: I am alive, an emergent expression of the universe trying to kill me, and I will not submit to the confusion that the simulation of life is life itself.

    There are people who have given up on humanity’s potential to own and solve our problems. They are gambling that the future of our species will be saved by deity-level AI. I will allow no such gambling with my species or planet. AI in control of human decisions will de-skill us, cripple us, and be a route for despotic control by unscrupulous humans.

    AI should only be used as a tool to expand our capabilities and experiences, equally, for all.

    Humans Above Technology.

    Image of Marching Hammers from Pink Floyd, The Wall

  • The Separation of Blame and Responsibility

    Blame and responsibility. We marry these two things together naturally. We figure that blame for something means we are responsible for it, and also the inverse, that if we are not to blame for something, we are not responsible.

    I have recently been reading the book The Subtle Art of Not Giving a F*ck, and the author Mark Manson makes a big point of the idea of radical responsibility, which is that we are responsible for everything in our lives, but we are not to blame for these things. It’s a good trick of language to divide these concepts, and frees us up to look at the world in a refreshing way.

    Try this game: think of a problem in your life or the world, and release yourself from blame while accepting responsibility for it.

    You are not to blame for not being rich, but you are responsible for handling your finances, and you are responsible for how you feel about money.

    You are not to blame for climate change, but you are responsible for protecting the world.

    You are not to blame for the shitshow the Internet has become, but you are responsible for which social media you choose to use or reject.

    You are not to blame for the dumpster apocalypse of our political system, but you are responsible for helping to fix it.

    You are not to blame for having feelings for a coworker, but you are responsible for acting like an adult and not cheating on your wife.

    You are not to blame for the misuse of AI and robots to manipulate, deceive, and take away value from human life, but you are responsible for supporting social constructs that protect us from this misuse.

    You are not to blame for your crappy habits, but you are responsible for changing them into good habits.

    You are not to blame for belonging to a country that assassinates foreign leaders, but you are responsible for pointing out that our own laws forbid this.

    Executive_Order_11905

    So much of our culture is about retributive justice and punishment for wrongs we commit. But there is no evidence that following this script will lead to a better world in our post-shame slophouse. Instead, we can make more progress but placing ideas like Justice into the same dump we place Rights and Reason, and thinking instead how we can build better selves and a better world.

    You are not to blame for blaming yourself for everything, but you are responsible for stopping the needless cycle of recriminations you are subjecting yourself to.

    Picture of Quinn, who I am delighted to be responsible for.

  • ML Models of Self: Where We Sit and Live

    We don’t need arguments about right or wrong, we need systems that solve problems.

    The difficulty of creating a calculus of social order is twofold: we don’t know how to turn our interactions into math, and we don’t know how to weigh the value of various outcomes of said calculus. This is true both in small ways, such as where people should sit in an office, and in big ways, like what our immigration policy should be.
    So how do we proceed? With another ten thousand years of ass and grass and beer? Or do we try to build something a little more sophisticated?
    One thing we will need is to determine how to value individual needs and joy against the needs and joys of others. Despite thousands of years of philosophical musing on this issue, we are no closer to a simple equation of human happiness. Maybe because nobody tried to build one, instead opting to sway others to general principles of utility or deontology.
    But with machine learning it seems to me there is a path to creating a different kind of approach. Let’s imagine I train a model of me, so it can not only identity me but also is my API, and if you want to know about what my needs are you can query that model.
    I think with models of self like this, we can look at both “where you sit in an office” and “immigration policy” and see that both dilemmas could be solved.
    So the thing about an office is, some people really don’t care where they sit, and some people really do. So a seating algorithm should be able to query employee models and determine what these needs are, while also weighing the importance of group dynamics and impact of certain individuals. For instance, some folks are just loud, and other folks can’t stand loud people. While humans doing a seating chart might be able to take a few of these characteristics into account, an ML system could weigh thousands or millions of factors to determine optimal seating.
    Some people don’t like immigrants, people from other cultures, and some people love them. I am in the latter category, which is one of the main reason I choose NYC as my home. But being wary of immigrants isn’t an indication of an underlying sociopathy (necessarily). It may just be a desire to maintain an understandable comfort zone around you and your kin and culture.
    Instead of strict or lax immigration laws, if we had models of self we could query, perhaps we could have town level immigration policies.
    I know this sounds dystopian, that maybe someday there will be towns or neighborhoods that would reject the other because of their culture or place or origin. But, if it opens up the world to greater and more fair interactions overall, maybe it’s a better world. There are many towns in the US that need immigrants to survive. With models of self, your AI could guide you to the right kind of community.
    If the insular communities become more insular, and the open communities become more open, the open communities will ultimately win. It might take ten thousand years, or two hundred uear, but when we get there, it will be through the evolution of our society guided by our AI-advocated needs, and not another shot in the dark with laws and presidential decrees.

    Image of D&D Players Handbook, which lets people create a simple model and game with it.

  • The Smart Cow Theory of History

    Perhaps you have heard of the Great Man Theory of History,

    The great man theory is a 19th-century idea according to which history can be largely explained by the impact of great men, or heroes; highly influential and unique individuals who, due to their natural attributes, such as superior intellect, heroic courage, or divine inspiration, have a decisive historical effect.

    I’m not really on board with that one to be honest. Mostly because I think the real changes in history happen due to technology, and not ideology, warfare, or singular heroism or leadership. The invention of firearms, antibiotics, and alternating current have changed average lives much more than most of the Herculean achievements supposedly attributed to the Heroes of History.

    But, how did those technological and scientific advancements occur? This brings me to my proposal: the Smart Cow Theory of History.

    The Smart Cow idea is derived from the Smart Cow Problem:

    The smart cow problem is the concept that, when a group of individuals is faced with a technically difficult task, only one of their members has to solve it. When the problem has been solved once, an easily repeatable method may be developed, allowing the less technically proficient members of the group to accomplish the task.

    The term smart cow problem is thought to be derived from the expression: “It only takes one smart cow to open the latch of the gate, and then all the other cows follow.”[1]

    Imagine that each advancement we make in technology brings us, a metaphorical herd of cows, to a new, greener pasture. For instance, the development of the internal combustion engine by a Smart Cow, or small group of smart cows, suddenly gives us massive power and abilities. Tesla figures out Alternating Current to unlock the gate of power transmission, and then the regular cows can build the power grid.

    But then we inevitably come to another locked gate: not enough fuel, dangerous vehicles, carbon emissions. We bang against these gates until the smart cows, the technologists and scientists and yes, the capitalists who fund them, figure out a way to unlock these gates. Then we all blunder through, blindly, into the supposedly greener pastures of our future.

    Smart cows would be folks like: Tesla, Edison, Curie, Einstein, Ada Lovelace, Ford, Taiichi Ohno, Oppenheimer, Herbert Boyer, Tim Berners-Lee and so on. Every woman and man working in a lab coat somewhere or pounding on a keyboard is likely a Smart Cow in some way. Because really smart and effective cows don’t do it alone; it usually takes a cowteam to unlatch those gates.

    And what are those gates? Those are the problems we face. Energy, food, travel, health, understanding, control.

    The Smart Cow Theory of History implies a few things. One, is that us cows that blunder through the unlocked gates have no idea what lies on the other side of the fence, and most of us don’t even know that gate was there in the first place.

    Second, and more importantly, the Smart Cows don’t know what lies on the other side either.

    Sure they can theorize and discuss, but Smart Cows are just individuals who can apply their focus to a single issue and solve it. They are just looking for a solution to the unhappy mooing around them

    Oppenheimer didn’t know how nations would negotiate power through nuclear weapons. He was more worried that the initial atom bomb tests would ignite the oxygen on the planet and kill us all.

    The Smart Cows building AI, putting computers in our pockets, and editing our genes have No Idea what the outcome of these actions will be. They can’t, the universe and humans are just far too complex to know what the next pasture and gate will look like. But we do know that these gates are getting more and more complicated, the pastures more dangerous, and the Smart Cows are going to keep unlatching.

    Telling Smart Cows to stop fiddling with the locks isn’t going to slow them down. But if we all just spend a few moments asking the right questions, and enshrining the right values, maybe, just maybe, we can set our boundaries in time to prevent disaster.

    Perhaps some gates need to remain locked to keep the herd from wandering onto the highway, or into the slaughterhouse.

    Moo.

    To hear a couple of very Smart Cows discuss some upcoming gates around biotechnology and CRISPR, I highly recommend episode 50 of the After On podcast with Kevin Rose.

  • Empathy Will Not Scale

    The limitations of empathy as a model of morality are more than just philosophical ones. The capacity for empathy is not only a finite resource, it is a zero sum one.

    We know empathy is finite because of Dunbar’s Number, the limitation we have as humans to maintain social networks.

    Dunbar’s number is a suggested cognitive limit to the number of people with whom one can maintain stable social relationships—relationships in which an individual knows who each person is and how each person relates to every other person.[

    The exact number, be it 150 or 10, of your empathy range is irrelevant. As the illustration above shows, there are various boundaries/sizes of different social types. The point is, you only have so many slots that can be filled with real people that you actually care about enough to see the world through their eyes, and shuffling folks in and out of slots doesn’t change this finite capacity.

    Neither does the trick of some religions of slotting in God or all other humans into an empathy slot – the abstractions are easy to imagine but almost impossible to realize for most folks over a lifetime. Holy men and women can pull it off, but they are outliers for sure.

    I think the best way to think about empathy is as another physical sense, the bonding one. And just like you can’t see over the horizon or hear all things at all times, you can’t suddenly will yourself into having a super-empathetic sense.

    Plus, we have another conundrum, and maybe the bigger one: even if we can somehow juggle our way around Dunbar’s spectrum and feel something for everyone, the people we are having empathy with have no requirement to give us empathy in return.

    We can understand the terrorists all we want, but if they don’t understand us then it’s a pointless undertaking.

    I grew up in the 70s, when there was a belief that without empathy for everyone we couldn’t end racism or prevent a final world war. But racism has not gone anywhere, and war has just morphed to remain profitable.

    So let’s quit making each other feel guilty that we aren’t universal empathy machines. Let’s find a new way forward.

    My suggestion is the Singleton Pattern, which is a recognition of the special status of humans in the universe as distinct and special. It’s not about recognizing empathy, it’s about recognizing value. Together, small teams of humans can make massive improvements to our reality. We are all Worthy.

    When you walk through a forest, you can feel deep affection for the natural world and want to preserve it, but that doesn’t mean you have empathy for every tree. It’s a different feeling, and that’s the feeling we need for society.

    Instead of using guilt and shame to force folks to have empathy for all, we should find ways to discuss our goals for society. And that entails putting some numbers to things.

    What crime rate are you willing to tolerate? What rate of abortions? What rate of poverty? What is the minimum cost college should be? Homelessness %? Tax rate?

    The metrics of the world, and building a data driven society, do not require empathy. They just require understanding statistics and percentiles, and believing that caring about data is more maintainable than your emotional connection to other individuals. High school math, and caring about your grade.

    I don’t need a leader telling me how to think or feel. I’ve got that covered, thanks. What I need is leaders who care about demonstrable, quantifiable results of policy choices.

    Dunbar boundary Image from this article.

  • Managing: The Root of Better Capitalism

    Far too little attention is paid to the physical nature of humans in the modern workplace. By this I mean not only ergonomic chairs and doilies, I also mean the way people stand when talking to each other.

    The tone of a person’s voice during a one-on-one discussion is far more important than what is being said, at least usually.

    When people give speeches to their manager about their state – for instance in response to a question like “how are you liking your job?”, They will often launch into words crafted through various discussions with others. It’s a soliloquy. You should listen to the content, but also listen instead to their voice, the character they are choosing or having chosen for them for the the performance. Does is quiver or rage? Does it sound hurt or joyous?

    And don’t only observe: then, try to figure out what emotion the person would like to be feeling, and how it would hurt or help others, and try to help them see how to get there within the context of the job.

    Management isn’t a psychology job, it’s a sociology one. As a manager you are trying to maintain a healthy society, which with Americans means they must have a say in the direction.

    See, a good manager makes the people she manages into amazing fountains of enthusiastic output, a good manager creates an elite force with esprit de corp. In fact, this should be the explicit goal of all managers, to build such a spirit of comradely excellence.

    It doesn’t always work, but always the attempt has value.

    There are few things more worthwhile than helping people help themselves, and if we are going to do this silly Capitalism thing as managers, let’s kick some ass and do some good for folks.

    Never lie. Never undermine others. Never allow questionable behavior. Build climates of trust, and support the propagation of knowledge in the company. Give all employees a path and opportunity to grow. Try to get fair pay at market rates+ for everyone.

    If only every manager, in every company, committed to these simple principles, Capitalism wouldn’t suck so much for almost everyone.

    Render by Mike Winkelmann