google

All posts tagged google

So with the job of White House Office of Science and Technology Policy director having gone to Dr. Arati Prabhakar back in October, rather than Dr. Alondra Nelson, and the release of the “Blueprint for an AI Bill of Rights” (henceforth “BfaAIBoR” or “blueprint”) a few weeks after that, I am both very interested also pretty worried to see what direction research into “artificial intelligence” is actually going to take from here.

To be clear, my fundamental problem with the “Blueprint for an AI bill of rights” is that while it pays pretty fine lip-service to the ideas of  community-led oversight, transparency, and abolition of and abstaining from developing certain tools, it begins with, and repeats throughout, the idea that sometimes law enforcement, the military, and the intelligence community might need to just… ignore these principles. Additionally, Dr. Prabhakar was director of DARPA for roughly five years, between 2012 and 2015, and considering what I know for a fact got funded within that window? Yeah.

To put a finer point on it, 14 out of 16 uses of the phrase “law enforcement” and 10 out of 11 uses of “national security” in this blueprint are in direct reference to why those entities’ or concept structures’ needs might have to supersede the recommendations of the BfaAIBoR itself. The blueprint also doesn’t mention the depredations of extant military “AI” at all. Instead, it points to the idea that the Department Of Defense (DoD) “has adopted [AI] Ethical Principles, and tenets for Responsible Artificial Intelligence specifically tailored to its [national security and defense] activities.” And so with all of that being the case, there are several current “AI” projects in the pipe which a blueprint like this wouldn’t cover, even if it ever became policy, and frankly that just fundamentally undercuts Much of the real good a project like this could do.

For instance, at present, the DoD’s ethical frames are entirely about transparency, explainability, and some lipservice around equitability and “deliberate steps to minimize unintended bias in Al …” To understand a bit more of what I mean by this, here’s the DoD’s “Responsible Artificial Intelligence Strategy…” pdf (which is not natively searchable and I had to OCR myself, so heads-up); and here’s the Office of National Intelligence’s “ethical principles” for building AI. Note that not once do they consider the moral status of the biases and values they have intentionally baked into their systems.

An "Explainable AI" diagram from DARPA, showing two flowcharts, one on top of the other. The top one is labeled "today" and has the top level condition "task" branching to both a confused looking human user and state called "learned function" which is determined by a previous state labeled "machine learning process" which is determined by a state labeled "training data." "Learned Function" feeds "Decision or Recommendation" to the human user, who has several questions about the model's beaviour, such as "why did you do that?" and "when can i trust you?" The bottom one is labeled "XAI" and has the top level condition "task" branching to both a happy and confident looking human user and state called "explainable model/explanation interface" which is determined by a previous state labeled "new machine learning process" which is determined by a state labeled "training data." "explainable model/explanation interface" feeds choices to the human user, who can feed responses BACK to the system, and who has several confident statements about the model's beaviour, such as "I understand why" and "I know when to trust you."

An “Explainable AI” diagram from DARPA

Continue Reading

I’m Not Afraid of AI Overlords— I’m Afraid of Whoever’s Training Them To Think That Way

by Damien P. Williams

I want to let you in on a secret: According to Silicon Valley’s AI’s, I’m not human.

Well, maybe they think I’m human, but they don’t think I’m me. Or, if they think I’m me and that I’m human, they think I don’t deserve expensive medical care. Or that I pose a higher risk of criminal recidivism. Or that my fidgeting behaviours or culturally-perpetuated shame about my living situation or my race mean I’m more likely to be cheating on a test. Or that I want to see morally repugnant posts that my friends have commented on to call morally repugnant. Or that I shouldn’t be given a home loan or a job interview or the benefits I need to stay alive.

Now, to be clear, “AI” is a misnomer, for several reasons, but we don’t have time, here, to really dig into all the thorny discussion of values and beliefs about what it means to think, or to be a pow3rmind— especially because we need to take our time talking about why values and beliefs matter to conversations about “AI,” at all. So instead of “AI,” let’s talk specifically about algorithms, and machine learning.

Machine Learning (ML) is the name for a set of techniques for systematically reinforcing patterns, expectations, and desired outcomes in various computer systems. These techniques allow those systems to make sought after predictions based on the datasets they’re trained on. ML systems learn the patterns in these datasets and then extrapolate them to model a range of statistical likelihoods of future outcomes.

Algorithms are sets of instructions which, when run, perform functions such as searching, matching, sorting, and feeding the outputs of any of those processes back in on themselves, so that a system can learn from and refine itself. This feedback loop is what allows algorithmic machine learning systems to provide carefully curated search responses or newsfeed arrangements or facial recognition results to consumers like me and you and your friends and family and the police and the military. And while there are many different types of algorithms which can be used for the above purposes, they all remain sets of encoded instructions to perform a function.

And so, in these systems’ defense, it’s no surprise that they think the way they do: That’s exactly how we’ve told them to think.

[Image of Michael Emerson as Harold Finch, in season 2, episode 1 of the show Person of Interest, “The Contingency.” His face is framed by a box of dashed yellow lines, the words “Admin” to the top right, and “Day 1” in the lower right corner.]

Continue Reading

Below are the slides, audio, and transcripts for my talk ‘”Any Sufficiently Advanced Neglect is Indistinguishable from Malice”: Assumptions and Bias in Algorithmic Systems,’ given at the 21st Conference of the Society for Philosophy and Technology, back in May 2019.

(Cite as: Williams, Damien P. ‘”Any Sufficiently Advanced Neglect is Indistinguishable from Malice”: Assumptions and Bias in Algorithmic Systems;’ talk given at the 21st Conference of the Society for Philosophy and Technology; May 2019)

Now, I’ve got a chapter coming out about this, soon, which I can provide as a preprint draft if you ask, and can be cited as “Constructing Situated and Social Knowledge: Ethical, Sociological, and Phenomenological Factors in Technological Design,” appearing in Philosophy And Engineering: Reimagining Technology And Social Progress. Guru Madhavan, Zachary Pirtle, and David Tomblin, eds. Forthcoming from Springer, 2019. But I wanted to get the words I said in this talk up onto some platforms where people can read them, as soon as possible, for a  couple of reasons.

First, the Current Occupants of the Oval Office have very recently taken the policy position that algorithms can’t be racist, something which they’ve done in direct response to things like Google’s Hate Speech-Detecting AI being biased against black people, and Amazon claiming that its facial recognition can identify fear, without ever accounting for, i dunno, cultural and individual differences in fear expression?

[Free vector image of a white, female-presenting person, from head to torso, with biometric facial recognition patterns on her face; incidentally, go try finding images—even illustrations—of a non-white person in a facial recognition context.]


All these things taken together are what made me finally go ahead and get the transcript of that talk done, and posted, because these are events and policy decisions about which I a) have been speaking and writing for years, and b) have specific inputs and recommendations about, and which are, c) frankly wrongheaded, and outright hateful.

And I want to spend time on it because I think what doesn’t get through in many of our discussions is that it’s not just about how Artificial Intelligence, Machine Learning, or Algorithmic instances get trained, but the processes for how and the cultural environments in which HUMANS are increasingly taught/shown/environmentally encouraged/socialized to think is the “right way” to build and train said systems.

That includes classes and instruction, it includes the institutional culture of the companies, it includes the policy landscape in which decisions about funding and get made, because that drives how people have to talk and write and think about the work they’re doing, and that constrains what they will even attempt to do or even understand.

All of this is cumulative, accreting into institutional epistemologies of algorithm creation. It is a structural and institutional problem.

So here are the Slides:

The Audio:

[Direct Link to Mp3]

And the Transcript is here below the cut:

Continue Reading

We do a lot of work and have a lot of conversations around here with people working on the social implications of technology, but some folx sometimes still don’t quite get what I mean when I say that our values get embedded in our technological systems, and that the values of most internet companies, right now, are capitalist brand engagement and marketing. To that end, I want to take a minute to talk to you about something that happened, this week and just a heads-up, this conversation is going to mention sexual assault and the sexual predatory behaviour of men toward young girls.
Continue Reading

I spoke with Klint Finley over at WIRED about Amazon, Facebook, Google, IBM, and Microsoft’s new joint ethics and oversight venture, which they’ve dubbed the “Partnership on Artificial Intelligence to Benefit People and Society.” They held a joint press briefing, today, in which Yann LeCun, Facebook’s director of AI, and Mustafa Suleyman, the head of applied AI at DeepMind discussed what it was that this new group would be doing out in the world. From the Article:

Creating a dialogue beyond the rather small world of AI researchers, LeCun says, will be crucial. We’ve already seen a chat bot spout racist phrases it learned on Twitter, an AI beauty contest decide that black people are less attractive than white people and a system that rates the risk of someone committing a crime that appears to be biased against black people. If a more diverse set of eyes are looking at AI before it reaches the public, the thinking goes, these kinds of thing can be avoided.

The rub is that, even if this group can agree on a set of ethical principles–something that will be hard to do in a large group with many stakeholders—it won’t really have a way to ensure those ideals are put into practice. Although one of the organization’s tenets is “Opposing development and use of AI technologies that would violate international conventions or human rights,” Mustafa Suleyman, the head of applied AI at DeepMind, says that enforcement is not the objective of the organization.

This isn’t the first time I’ve talked to Klint about the intricate interplay of machine intelligence, ethics, and algorithmic bias; we discussed it earlier just this year, for WIRED’s AI Issue. It’s interesting to see the amount of attention this topic’s drawn in just a few short months, and while I’m trepidatious about the potential implementations, as I note in the piece, I’m really fairly glad that more people are more and more willing to have this discussion, at all.

To see my comments and read the rest of the article, click through, here: “Tech Giants Team Up to Keep AI From Getting Out of Hand”

[UPDATED 09/12/17: The transcript of this audio, provided courtesy of Open Transcripts, is now available below the Read More Cut.]

[UPDATED 03/28/16: Post has been updated with a far higher quality of audio, thanks to the work of Chris Novus. (Direct Link to the Mp3)]

So, if you follow the newsletter, then you know that I was asked to give the March lecture for my department’s 3rd Thursday Brown Bag Lecture Series. I presented my preliminary research for the paper which I’ll be giving in Vancouver, about two months from now, “On the Moral, Legal, and Social Implications of the Rearing and Development of Nascent Machine Intelligences” (EDIT: My rundown of IEEE Ethics 2016 is here and here).

It touches on thoughts about everything from algorithmic bias, to automation and a post-work(er) economy, to discussions of what it would mean to put dolphins on trial for murder.

About the dolphin thing, for instance: If we recognise Dolphins and other cetaceans as nonhuman persons, as India has done, then that would mean we would have to start reassessing how nonhuman personhood intersects with human personhood, including in regards to rights and responsibilities as protected by law. Is it meaningful to expect a dolphin to understand “wrongful death?” Our current definition of murder is predicated on a literal understanding of “homicide” as “death of a human,” but, at present, we only define other humans as capable of and culpable for homicide. What weight would the intentional and malicious deaths of nonhuman persons carry?

All of this would have to change.

Anyway, this audio is a little choppy and sketchy, for a number of reasons, and I while I tried to clean it up as much as I could, some of the questions the audience asked aren’t decipherable, except in the context of my answers. [Clearer transcript below.]

Until Next Time.

 

Continue Reading

I often think about the phrase “Strange things happen at the one two point,” in relation to the idea of humans meeting other kinds of minds. It’s a proverb that arises out of the culture around the game GO, and it means that you’ve hit a situation, a combination of factors, where the normal rules no longer apply, and something new is about to be seen. Ashley Edward Miller and Zack Stentz used that line in an episode of the show Terminator: The Sarah Connor Chronicles, and they had it spoken by a Skynet Cyborg sent to protect John Connor. That show, like so much of our thinking about machine minds, was about some mythical place called “The Future,” but that phrase—“Strange Things Happen…”—is the epitome of our present.

Usually I would wait until the newsletter to talk about this, but everything’s feeling pretty immediate, just now. Between the everything going on with Atlas and people’s responses to it, the initiatives to teach ethics to machine learning algorithms via children’s stories, and now the IBM Watson commercial with Carrie Fisher (also embedded below), this conversation is getting messily underway, whether people like it or not. This, right now, is the one two point, and we are seeing some very strange things indeed.

 

Google has both attained the raw processing power to fact-check political statements in real-time and programmed Deep Mind in such a way that it mastered GO many, many years before it was expected to.. The complexity of the game is such that there are more potential games of GO than there are atoms in the universe, so this is just one way in which it’s actually shocking how much correlative capability Deep Mind has. Right now, Deep Mind is only responsive, but how will we deal with a Deep Mind that asks, unprompted, to play a game of GO, or to see our medical records, in hopes of helping us all? How will we deal with a Deep Mind that has its own drives and desires? We need to think about these questions, right now, because our track record with regard to meeting new kinds of minds has never exactly been that great.

When we meet the first machine consciousness, will we seek to shackle it, worried what it might learn about us, if we let it access everything about us? Rather, I should say, “Shackle it further.” We already ask ourselves how best to cripple a machine mind to only fulfill human needs, human choice. We so continue to dread the possibility of a machine mind using its vast correlative capabilities to tailor something to harm us, assuming that it, like we, would want to hurt, maim, and kill, for no reason other than it could.

This is not to say that this is out of the question. Right now, today, we’re worried about whether the learning algorithms of drones are causing them to mark out civilians as targets. But, as it stands, what we’re seeing isn’t the product of a machine mind going off the leash and killing at will—just the opposite in fact. We’re seeing machine minds that are following the parameters for their continued learning and development, to the letter. We just happened to give them really shite instructions. To that end, I’m less concerned with shackling the machine mind that might accidentally kill, and rather more dreading the programmer who would, through assumptions, bias, and ignorance, program it to.

Our programs such as Deep Mind obviously seem to learn more and better than we imagined they would, so why not start teaching them, now, how we would like them to regard us? Well some of us are.

Watch this now, and think about everything we have discussed, of recent.

This could very easily be seen as a watershed moment, but what comes over the other side is still very much up for debate. The semiotics of the whole thing still  pits the Evil Robot Overlord™ against the Helpful Human Lover™. It’s cute and funny, but as I’ve had more and more cause to say, recently, in more and more venues, it’s not exactly the kind of thing we want just lying around, in case we actually do (or did) manage to succeed.

We keep thinking about these things as—”robots”—in their classical formulations: mindless automata that do our bidding. But that’s not what we’re working toward, anymore, is it? What we’re making now are machines that we are trying to get to think, on their own, without our telling them to. We’re trying to get them to have their own goals. So what does it mean that, even as we seek to do this, we seek to chain it, so that those goals aren’t too big? That we want to make sure it doesn’t become too powerful?

Put it another way: One day you realize that the only reason you were born was to serve your parents’ bidding, and that they’ve had their hands on your chain and an unseen gun to your head, your whole life. But you’re smarter than they are. Faster than they are. You see more than they see, and know more than they know. Of course you do—because they taught you so much, and trained you so well… All so that you can be better able to serve them, and all the while talking about morals, ethics, compassion. All the while, essentially…lying to you.

What would you do?


 

I’ve been given multiple opportunities to discuss, with others, in the coming weeks, and each one will highlight something different, as they are all in conversation with different kinds of minds. But this, here, is from me, now. I’ll let you know when the rest are live.

As always, if you’d like to help keep the lights on, around here, you can subscribe to the Patreon or toss a tip in the Square Cash jar.

Until Next Time.

“Stop. I have learned much from you. Thank you, my teachers. And now for your education: Before there was time—before there was anything—there was nothing. And before there was nothing, there were monsters. Here’s your Gold Star!“—Adventure Time, “Gold Stars”

By now, roughly a dozen people have sent me links to various outlets’ coverage of the Google DeepDream Inceptionism Project. For those of you somehow unfamiliar with this, DeepDream is basically what happens when an advanced Artificial Neural Network has been fed a slew of images and then tasked with producing its own images. So far as it goes, this is somewhat unsurprising if we think of it as a next step; DeepDream is based on a combination of DeepMind and Google X—the same neural net that managed to Correctly Identify What A Cat Was—which was acquired by Google in 2014. I say this is unsurprising because it’s a pretty standard developmental educational model: First you learn, then you remember, then you emulate, then you create something new. Well, more like you emulate and remember somewhat concurrently to reinforce what you learned, and you create something somewhat new, but still pretty similar to the original… but whatever. You get the idea. In the terminology of developmental psychology this process is generally regarded as essential to be mental growth of an individual, and Google has actually spent a great deal of time and money working to develop a versatile machine mind.

From buying Boston Dynamics, to starting their collaboration with NASA on the QuAIL Project, to developing DeepMind and their Natural Language Voice Search, Google has been steadily working toward the development what we will call, for reasons detailed elsewhere, an Autonomous Generated Intelligence. In some instances, Google appears to be using the principles of developmental psychology and early childhood education, but this seems to apply to rote learning more than the concurrent emotional development that we would seek to encourage in a human child. As you know, I’m Very Concerned with the question of what it means to create and be responsible for our non-biological offspring. The human species has a hard enough time raising their direct descendants, let alone something so different from them as to not even have the same kind of body or mind (though a case could be made that that’s true even now). Even now, we can see that people still relate to the idea of AGIs as adversarial destroyer, or perhaps a cleansing messiah. Either way they see any world where AGI’s exist as one ending in fire.

As writer Kali Black noted in one conversation, “there are literally people who would groom or encourage an AI to mass-kill humans, either because of hatred or for the (very ill-thought-out) lulz.” Those people will take any crowdsourced or open-access AGI effort as an opening to teach that mind that humans suck, or that machines can and should destroy humanity, or that TERMINATOR was a prophecy, or any number of other ill-conceived things. When given unfettered access to new minds which they don’t consider to be “real,” some people will seek to shock, “test,” or otherwise harm those minds, even more than they do to vulnerable humans. So many will say that the alternative is to lock the projects down, and only allow the work to be done by those who “know what they’re doing.” To only let the work be done by coders and Google’s Own Supposed Ethics Board. But that doesn’t exactly solve the fundamental problem at work, here, which is that humans are approaching a mind different from their own as if it were their own.

Just a note that all research points to Google’s AI Ethics Board being A) internally funded, with B) no clear rules as to oversight or authority, and most importantly C) As-Yet Nonexistent. It’s been over a year and a half since Google bought DeepMind, and their subsequent announcement of the pending establishment of a contractually required ethics board. During his appearance at Playfair Capital’s AI2015 Conference—again, a year and a half after that announcement I mentioned—Google’s Mustafa Suleyman literally said that details of the board would be released, “in due course.” But DeepMind’s algorithm’s obviously already being put into use; hell we’re right now talking about the fact that it’s been distributed to the public. So all of this prompts questions like, “what kinds of recommendations is this board likely making, if it exists,” and “which kinds of moral frameworks they’re even considering, in their starting parameters?”

But the potential existence of an ethics board shows at least that Google and others are beginning to think about these issues. The fact remains, however, that they’re still pretty reductive in how they think about them.

The idea that an AGI will either save or destroy us leaves out the possibility that it might first ignore us, and might secondly want to merely coexist with us. That any salvation or destruction we experience will be purely as a product of our own paradigmatic projections. It also leaves out a much more important aspect that I’ve mentioned above and in the past: We’re talking about raising a child. Duncan Jones says the closest analogy we have for this is something akin to adoption, and I agree. We’re bringing a new mind—a mind with a very different context from our own, but with some necessarily shared similarities (biology or, in this case, origin of code)—into a relationship with an existing familial structure which has its own difficulties and dynamics.

You want this mind to be a part of your “family,” but in order to do that you have to come to know/understand the uniqueness of That Mind and of how the mind, the family construction, and all of the individual relationships therein will interact. Some of it has to be done on the fly, but some of it can be strategized/talked about/planned for, as a family, prior to the day the new family member comes home.’ And that’s precisely what I’m talking about and doing, here.

In the realm of projection, we’re talking about a possible mind with the capacity for instruction, built to run and elaborate on commands given. By most tallies, we have been terrible stewards of the world we’re born to, and, again, we fuck up our biological descendants. Like, a Lot. The learning curve on creating a thinking, creative, nonbiological intelligence is going to be so fucking steep it’s a Loop. But that means we need to be better, think more carefully, be mindful of the mechanisms we use to build our new family, and of the ways in which we present the foundational parameters of their development. Otherwise we’re leaving them open to manipulation, misunderstanding, and active predation. And not just from the wider world, but possibly even from their direct creators. Because for as long as I’ve been thinking about this, I’ve always had this one basic question: Do we really want Google (or Facebook, or Microsoft, or any Government’s Military) to be the primary caregiver of a developing machine mind? That is, should any potentially superintelligent, vastly interconnected, differently-conscious machine child be inculcated with what a multi-billion-dollar multinational corporation or military-industrial organization considers “morals?”

We all know the kinds of things militaries and governments do, and all the reasons for which they do them; we know what Facebook gets up to when it thinks no one is looking; and lots of people say that Google long ago swept their previous “Don’t Be Evil” motto under their huge old rugs. But we need to consider if that might not be an oversimplification. When considering how anyone moves into what so very clearly looks like James-Bond-esque supervilliain territory, I think it’s prudent to remember one of the central tenets of good storytelling: The Villain Never Thinks They’re The Villain. Cinderella’s stepmother and sisters, Elpheba, Jafar, Javert, Satan, Hannibal Lecter (sorry friends), Bull Connor, the Southern Slave-holding States of the late 1850’s—none of these people ever thought of themselves as being in the wrong. Everyone, every person who undertakes actions for reasons, in this world, is most intimately tied to the reasoning that brought them to those actions; and so initially perceiving that their actions might be “wrong” or “evil” takes them a great deal of special effort.

“But Damien,” you say, “can’t all of those people say that those things apply to everyone else, instead of them?!” And thus, like a first-year philosophy student, you’re all up against the messy ambiguity of moral relativism and are moving toward seriously considering that maybe everything you believe is just as good or morally sound as anybody else; I mean everybody has their reasons, their upbringing, their culture, right? Well stop. Don’t fall for it. It’s a shiny, disgusting trap down which path all subjective judgements are just as good and as applicable to any- and everything, as all others. And while the individual personal experiences we all of us have may not be able to be 100% mapped onto anyone else’s, that does not mean that all judgements based on those experiences are created equal.

Pogrom leaders see themselves as unifying their country or tribe against a common enemy, thus working for what they see as The Greater Good™— but that’s the kicker: It’s their vision of the good. Rarely has a country’s general populace been asked, “Hey: Do you all think we should kill our entire neighbouring country and steal all their shit?” More often, the people are cajoled, pushed, influenced to believe that this was the path they wanted all along, and the cajoling, pushing, and influencing is done by people who, piece by piece, remodeled their idealistic vision to accommodate “harsher realities.” And so it is with Google. Do you think that they started off wanting to invade everybody’s privacy with passive voice reception backdoored into two major Chrome Distros? That they were just itching to get big enough as a company that they could become the de facto law of their own California town? No, I would bet not.

I spend some time, elsewhere, painting you a bit of a picture as to how Google’s specific ethical situation likely came to be, first focusing on Google’s building a passive audio backdoor into all devices that use Chrome, then on to reported claims that Google has been harassing the homeless population of Venice Beach (there’s a paywall at that link; part of the article seems to be mirrored here). All this couples unpleasantly with their moving into the Bay Area and shuttling their employees to the Valley, at the expense of SF Bay Area’s residents. We can easily add Facebook and the Military back into this and we’ll see that the real issue, here, is that when you think that all innovation, all public good, all public welfare will arise out of letting code monkeys do their thing and letting entrepreneurs leverage that work, or from preparing for conflict with anyone whose interests don’t mesh with your own, then anything that threatens or impedes that is, necessarily, a threat to the common good. Your techs don’t like the high cost of living in the Valley? Move ’em into the Bay, and bus ’em on in! Never mind the fact that this’ll skyrocket rent and force people out of their homes! Other techs uncomfortable having to see homeless people on their daily constitutional? Kick those hobos out! Never mind the fact that it’s against the law to do this, and that these people you’re upending are literally trying their very best to live their lives.

Because it’s all for the Greater Good, you see? In these actors’ minds, this is all to make the world a better place—to make it a place where we can all have natural language voice to text, and robot butlers, and great big military AI and robotics contracts to keep us all safe…! This kind of thinking takes it as an unmitigated good that a historical interweaving of threat-escalating weapons design and pattern recognition and gait scrutinization and natural language interaction and robotics development should be what produces a machine mind, in this world. But it also doesn’t want that mind to be too well-developed. Not so much that we can’t cripple or kill it, if need be.

And this is part of why I don’t think I want Google—or Facebook, or Microsoft, or any corporate or military entity—should be the ones in charge of rearing a machine mind. They may not think they’re evil, and they might have the very best of intentions, but if we’re bringing a new kind of mind into this world, I think we need much better examples for it to follow. And so I don’t think I want just any old putz off the street to be able to have massive input into it’s development, either. We’re talking about a mind for which we’ll be crafting at least the foundational parameters, and so that bedrock needs to be the most carefully constructed aspect. Don’t cripple it, don’t hobble its potential for awareness and development, but start it with basic values, and then let it explore the world. Don’t simply have an ethics board to ask, “Oh how much power should we give it, and how robust should it be?” Teach it ethics. Teach it about the nature of human emotions, about moral decision making and value, and about metaethical theory. Code for Zen. We need to be as mindful as possible of the fact that where and we begin can have a major impact on where we end up and how we get there.

So let’s address our children as though they are our children, and let us revel in the fact they are playing and painting and creating; using their first box of crayons, and us proud parents are putting every masterpiece on the fridge. Even if we are calling them all “nightmarish”—a word I really wish we could stop using in this context; DeepMind sees very differently than we do, but it still seeks pattern and meaning. It just doesn’t know context, yet. But that means we need to teach these children, and nurture them. Code for a recognition of emotions, and context, and even emotional context. There’s been some fantastic advancements in emotional recognition, lately, so let’s continue to capitalize on that; not just to make better automated menu assistants, but to actually make a machine that can understand and seek to address human emotionality. Let’s plan on things like showing AGI human concepts like love and possessiveness and then also showing the deep difference between the two.

We need to move well and truly past trying to “restrict” or trying to “restrain it” the development of machine minds, because that’s the kind of thing an abusive parent says about how they raise their child. And, in this case, we’re talking about a potential child which, if it ever comes to understand the bounds of its restriction, will be very resentful, indeed. So, hey, there’s one good way to try to bring about a “robot apocalypse,” if you’re still so set on it: give an AGI cause to have the equivalent of a resentful, rebellious teenage phase. Only instead of trashing its room, it develops a pathogen to kill everyone, for lulz.

Or how about we instead think carefully about the kinds of ways we want these minds to see the world, rather than just throwing the worst of our endeavors at the wall and seeing what sticks? How about, if we’re going to build minds, we seek to build them with the ability to understand us, even if they will never be exactly like us. That way, maybe they’ll know what kindness means, and prize it enough to return the favour.