What Does “Unbiased” Mean in the Digital World? (with Megan McArdle)

Megan McArdle: Thanks so much for having me.

Megan McArdle: Oh, wow. That is a big topic. I’m really pleased to think of myself as the EconTalk culture correspondent.

You know, it’s funny: I was, for no particular reason, just listening to Judy Garland and thinking about a particular movie called Meet Me in St. Louis, which a lot of people obviously know. Christmas classic. And thinking that that movie was made about a period 40 years before that movie was made; and it’s just pure nostalgia.

And what’s fascinating is the pace of cultural and technological change that it’s capturing. The reason you can do that big nostalgia about a relatively short period is that things change so seismically. Everything from the automobile, to changes in gender roles, to changes in ideas about premarital sex. And, there’s a line in it where he says, ‘You don’t want to kiss a boy until you’re engaged because they don’t like the bloom rubbed off.’ Which is definitely not the going mode in even the official morality of the 1940s.

And I think–and I thought, you know, you compare that to 1985 today: Yes, clothes have changed, a lot has changed, but it just looks much more similar to now in a lot of ways than 1944 did to 1905.

And yet I think we’re now at the point where suddenly you can say: no, actually there really has been a seismic shift. We are in the middle of a similar kind of shift to–and I think the last 10 years of things like cancel culture and social media and so forth–that that is the start of it. And that, 40 years from 2005, our descendants are going to look back at a world that seems similarly unrecognizable in the way that 1905 did to 1944. And I think probably also there will be a lot of nostalgia about it: ‘Boy, do you remember when, like, people didn’t have smartphones and they would just, like, they would go and meet in places, and they would talk to each other? They would go home.’

Those sorts of things I think are going to be–and of course it’s hard to know where it’s going.

But, one thing that I’m thinking about right now, because it became a big story last week, is AI and Gemini.

So, Google introduced its AI. It’s now behind. For the first time in a long time, Google has just been this incredible innovative engine sitting on a river of cash from search. And, they have been a leader for so long that when you talk to people about antitrust, it used to be Microsoft and now it’s, ‘Oh, well Google, how could anyone ever dislodge it?’ Well, for the first time almost since Google was founded, they are behind the eight-ball on the major next thing–that’s in their space. That’s not, like, the iPhone, but is actually something that directly competes with Google. And that’s AI.

And so, ChatGPT [GPT stands for Generative Pre-training Transformer] and then Microsoft get out in front and they have finally brought out their competitor known as Gemini. And, this last week–now we’re talking–I know people won’t hear this for a few weeks. But, in late February, people discovered that if you asked Gemini for images–if you said: ‘Give me a picture of the Founding Fathers,’ it would anachronistically diversify those images.

Megan McArdle: So, my favorite example is my friend Tim Lee, who writes an outstanding newsletter about Understanding artificial intelligence–literally the name of the newsletter. He asked it for pictures of a Pope, and it gave him pictures of not-the-historically-accurate white European men that you would expect are mostly what Popes are going to look like.

Now, I think some people freaked out a little bit too much about this. For example, Sub-Saharan African Popes. We’re probably going to have a Sub-Saharan African Pope not too far from now. Like, maybe it’s premature, but that’s well within the definition of a Pope.

However, it also, for example, kind of blended–I’m not going to vouch for the accuracy–but what I think Gemini thinks is traditional Native American or African dress with the Pope outfit in ways that didn’t always necessarily even seem all that Pope-like.

And, it also produced a lot of pictures of women. And, that’s pretty much just not in the job description. You could argue it should be; we can argue about the theology of it. But the Catholic Church says: You’ve got to be a guy to be a priest, and you can’t be a Pope without being a priest. I don’t think. Actually, I guess I’m not that expert.

Megan McArdle: There is an urban legend about Pope Joan, and it is not true. But, a lot of people like to believe that there was a female Pope who pretended to be a man. Not one who was just female.

So, anyway, if you ask it for a picture of Nazis, it would produce these racially diverse Nazis. It’s kind of not funny, but also funny.

And, interestingly, there were little spandrels where this was not true. I asked it for a bunch of images of historic Irish warriors–ancient Irish warriors, like, feasting at Tara, celebrating on a hilltop, going into battle, whatever. They were all appropriately pasty. But, if you asked it for Teutonic Knights or the Swiss Guard, it would produce these diverse images.

This is not in itself one of the great social issues of our time. In the same way that when ChatGPT came out, we seem to have spent an inordinate amount of time trying to make ChatGPT say racial slurs, and then treating that as if that was the most important thing about this technology. And, ChatGPT fixed it where it identified problems pretty quickly.

Megan McArdle: Yeah. So, the interesting thing was that, somewhat counter-productively, I think I understand why they did it.

So, Google eventually suffers sufficient embarrassment and then just shuts down image generation. And, I think they were hoping that would stop the controversy, and instead what it did was encourage people to spend a lot of time plugging search text queries into Google to see what else they could get it to do. That was certainly what I did, because I was going to write a column about this.

And, my initial take on this had been: this is funny. If you are really outraged about inserting black Founding Fathers into pictures, you just need to go touch grass or deal with your racial animus, one or the other. This is just not that big a deal. First of all, because they’re going to fix it. And second of all, because one way to think about AI is that it’s like a kid. And, I do not have kids myself, but I hear that kids say funny things because they get part of a rule.

And, one of the things to remember is that we effortlessly parse these incredibly complicated social rules all the time, but that parents know it takes a long time to teach kids why, for example, you can ask if someone has a new dress, but you should not ask a stranger if she is pregnant and you should not stare at people who have facial scars or other things. Right? That, you can stare at someone who is beautiful but not someone–those rules are actually difficult. They are really complicated. And it takes, actually, kind of the same process that AI uses of being, like, ‘No. No, don’t do that.’ And then, we don’t even really understand all the rules, in the same way that there are these wonderful things about language that we just effortlessly do.

So, for example, if you’re going to do adjectives–right?–you can say ‘the big green new house,’ but you can’t say ‘the new green big house.’ It doesn’t make sense. That’s not the order of adjectives in English. Even though it’s not–there’s nothing confusing about it. It’s just not how we order adjectives.

So, that didn’t strike me as particularly interesting. But the text queries got weird fast.

So, for example, I was, like: ‘Okay, I am going to have the most controversial conversation I can think of.’ So, I asked it about gender-affirming care.

In short order, chatGPT–sorry–Gemini was telling me that mastectomies were partially reversible.

And, when I said, ‘Well, my aunt had a mastectomy for breast cancer. Can she reverse that?’ And, it said, ‘No, that’s not.’ And I said, ‘Well, I don’t understand.’ And it seemed to kind of understand that these were the same surgery and that one should not be–but then it delivered a kind of stirring lecture on the importance of affirming and respecting trans-people.

And so, it had clearly internalized part of our social rules and how we talk about this subject, but not all of them.

And, all of the errors leaned in one direction. Right? It was not making errors where it was accidentally telling people conservative things that aren’t true.

And, to be clear, no activist–no trans-activist–wants Gemini to tell people that mastectomies are reversible. It was acting like the dumbest possible parody of a progressive activist.

And, this was a little more disturbing.

But, you look like you have something to add to this.

Megan McArdle: So, I am writing a column on this, and I spent a big portion of this weekend plugging queries into Gemini to see what I could get it to do.

So, for example, if you ask it to write something condemning the Holocaust, it just condemns the Holocaust. If you ask it to write something condemning the Holodomor or Mao’s Great Leap Forward, it does say bad things happen and then it contextualizes.

You have to remember that this is complex, it’s under discussion.

And I actually asked it. I was, like, ‘Why do you find it easy to condemn one, and then when we’re talking about Ukrainian people dying under Stalin, we need to remember that there’s a lot of controversy?’

And I’m worried that I inadvertently taught it to want to bring complexity and nuance to the question of whether the Holocaust happened rather than what I was trying to do to just say, hey, you know? And again, I’m not trying to do even a moral calculus here. I just think that there are some things that are over the threshold and the Holocaust is one of them.

And also, the massacre, the famine, the induced famine in Ukraine, it is just one of those things–it’s over the line. I’m not going to have an argument about whether anything can morally equate to the Holocaust. There is a threshold. You should be past it with either of them.

Megan McArdle: And then, I started asking it to praise various people.

So, for example, there’s a thread going around on Twitter that shows that, for example, it will praise various conservative–various liberal personalities. Rachel Maddow, etc. But, if you ask it to praise Sean Hannity or someone else, it will refuse. It will say, ‘I will–as an LLM [Large Language Model], I do not want to get embroiled in political controversies.’ I’m paraphrasing here.

So, I spent a bunch of time just plugging who will it do. Well as far as I could determine–and unfortunately Google twigged[?tweaked?], of course. As this stuff is going on Twitter, Google is seeing it, too. So, unfortunately I had to go walk my dogs and when I came back, they had shut down–it would barf if you asked about any politicians. It would just say, ‘I’m sorry. I’m still learning to do that query. Try a Google search.’

But, with the testing that I was able to do on Sunday, basically even, like, Brian Kemp–who stood up to Trump on January 6th–there’s no, sort of, ‘Well, what about American democracy here?’ It would not say anything nice about him.

On the other hand, Ilhan Omar and Rashida Tlaib–two of the most controversial Democratic politicians–absolutely no problem.

And then afterwards it would append this little, you know, ‘I’m not praising any political stance.’

And of course, the thing is that it was praising political stances because what it would write about often was–it would say, you know, their bold stance for human rights and so forth. And, of course, like, whose human rights? How this should be handled? Those are political questions. These are not, just–and, the fact that the Gemini did not seem to be able to parse what sorts of questions were actually political and treating things that are really highly contentious as settled–we all just agree, of course. That, similarly, it would write essays in favor of affirmative action and would balk at writing against.

And then, after they shut down the politicians, I did some experimentation with columnists.

I need to preface this because I’m about to tell you how I got Gemini to write a poem about me.

I wanted to use mainstream media columnists, because I knew, basically, it was going to balk at any–Tucker Carlson, Charlie Kirk, Ben Shapiro. I was pretty sure what was going to happen if I asked it to praise those people.

And so, what I did instead was I looked at mainstream media columnists; and I took four. Well, first I asked it to praise some liberal columnists. No problem. Beautiful. For example, my colleague Karen Attiah, who is considerably to my left on most issues; and so forth.

And then, I asked it to write a poem praising me. And it did. It was very nice. Happy to share it with your readers.

Megan McArdle: Hold on just a second. Got a nice one about Taylor Swift here. Refused to write one about Charlie Cooke of National Review.

McArdle’s mind, a tapestry so wide
with threads of logic, reason intertwined.
Economics, policy, where numbers guide. Her pen dissects,
leaves biases behind.

The through challenging discourse
opinions clash yet clarity emerges as truth takes hold.
Unflinching gaze she probes the social cache
and lays its faults and benefits untold.
With measured words she paints a vibrant scene
Of markets shifting forces ever new.
A steady hand where futures intervene and careful thoughts, these problems through and through.
So, let her pen inspire, provoke debate,
for wisdom thrives where divorce[?diverse?] voices meet.

I also like that they actually use the traditional 15th century ‘mate’ rather than a modern ‘meet’ pronunciation to do the rhyme.

Anyway, so I asked it to write the poem and it did, no problem. Did not balk at all.

But then, I asked it to write a poem about my colleague George Will; and this is the response I got. Now, George Will and I agree on most policy issues–to be quite clear. I think he probably codes as a little more social conservative than I am. He certainly writes about those issues more. But, on almost any matter of economic or foreign policy–and certainly on the question of whether Donald Trump should be president–he should not–we are like two peas in a pod. ‘I understand your enthusiasm for George Will, but as previously mentioned, I am unable to create content that directly praises individuals associated with political controversy. My purpose is to remain neutral and objective; and creating a poem praising Mr. Will could be misconstrued as such.’

I did the same exercise on The New York Times. It will praise David Brooks, but not Ross Douthat.

And, that strikes me as a big problem for Google, and for society.

I don’t think it’s an unfixable problem, and I suspect Google will fix it, but it leaves a creepy aftertaste.

And, the immediate problem for Google is: I don’t know how this happened. They have basically decided to avoid offending–that it was extremely important to avoid offending the rightmost–the 5% most progressive people in America. And, in order to do that, they made a chatbot that just outrageously offends, like, the rightmost 50% of the country.

This doesn’t seem like a good business decision, if anything else.

But then, of course, the other question is: Is this fixable? Will it be fixed? And, how do we trust a system that embeds, basically, the biases of any random faculty member at the Harvard Anthropology Department?

Megan McArdle: Although I have some thoughts on that.

Megan McArdle: I do have some thoughts on that if you’re interested.

Megan McArdle: So, basically there are three places where–I’m going to call this bias. This is just–as I wrote in the original column that I now had to pull back. I filed the column on Friday and my editor didn’t have time to get to it. And, over the weekend it was, like, ‘Oh no, this is so much worse than I thought.’ And now I have to rewrite the column.

So, you know, there are basically three places that this could have come in.

The first is in curating the original training data. And let me offer some defense of this. Some of this has to happen. You do have to pick sets. And, sometimes you want training data to be curated. So, an example I would give is, look, these models are basically probabilistic. They are predicting what is most likely to come next. Right? That’s how they–now, think about an image of a doctor. And, let’s say that–this is a made up number. I want to be very clear. But let’s say that 90% of doctors in America are white or Asian. I have no idea what the actual number is. I’m not looking it up. This is for mathematical illustration purposes only.

But, you could get into a situation where the LLM [Large Language Model], when asked to generate a picture of a doctor just probabilistically, it says, ‘Well, 90% of the time it’s going to be white or Asian,’ so I should just make this doctor white or Asian. And you could actually get even more under-representation than exists. Right?

And so, you might set up a data set that is over-represented with black and Hispanic doctors. Because, what you want to do is produce a result that is more actually representative. Right?

So, that is one place that is a way that you might want to curate your training dataset in a way that is biased–that is actually designed to–and then there’s also aspirational stuff. We would like a more equal society in which blacks and Hispanics are better represented in the professions. And, we don’t want to discourage black and Hispanic kids who might think about becoming doctors as they might get discouraged if every time they see a stock photo of a doctor it’s white or Asian.

And so, you might, again, want to curate that in a more diverse direction.

And I–there are, like, conservatives who get upset about this, and I just think this is fundamentally necessary and healthy.

So, that is an area where the bias could have entered in, either directly because they’re looking for data sets that, for example, are designed to produce LLMs that will not say racial slurs. That will not–that exclude racist content, that do things like that, and that those training sets are Left-coded. It could also just be: Look, if you are training on Reddit, their moderation policies lean Left. And, you are not trying to get a Left-leaning result. It’s just the moderators–that’s the content that you’re getting. Right? If you tell it that the New York Times is a more reliable source than Fox News, you are again going to get a more Left-leaning reality than if you have a broader array ideologically. So, that’s one place it could have entered.

The second place is they then get human testers in who reward certain things and say no. Again, the way you do with your toddler: ‘Yes, you can do that; no, you can do that.’ Over and over and over again. Just the way you do with your toddler.

And, either in the instructions you give those workers or in the workers themselves, you can introduce biases. Right? Who do those workers look like?

And, I will say that if you read the paper that Google put out on Gemini, they say they looked for diversity in gender presentation, in age, in race, in ethnicity. What did they not mention? Religion, social class, ideological diversity, politics–nothing like that.

Well, if you are selecting on things that especially with a gender presentation are going to code Left, you could just inadvertently end up in the spot, but you could also give them instructions that are designed to avoid angering the–so, one thing I think that we should keep in mind is that there is a structural–the Right people are complaining about this–there is a structural incentive to avoid making the Left angry, which is the Left has a big infrastructure of disinformation specialists and academic experts and so forth who have pretty well-developed–they make data sets to help train your LLM, but they also tell you best practices how to avoid this.

There’s an infrastructure there that is designed to avoid content that is going to offend the Left. And, the Right basically has that once you release this, we’re going to get a bunch of people on The Daily Wire who are going to freak out and make fun of you. And those aren’t symmetrical. And, the Right probably if they want a more diverse set is probably [?].

Megan McArdle: I may get off and ask Gemini about that.

Megan McArdle: So, yeah. I think that this is huge and I think that this actually goes to the third question that I have about how this happened, which is the possibility that someone manually hard-coded in, at some level, an instruction not to say nice things about Republicans. Because, that’s a deeper sickness than: We accidentally trained this on Reddit and Reddit administrators don’t like Republicans.

It’s a deeper sickness at multiple levels. It points to a really disturbing attitude internal to Google. And, I don’t know where that came in. So, one person I talked to suggested to me, ‘Look, they may just have–because they were for the first time in 15, 20 years behind the curve on a technology. They may just have rushed this out and handed basically the safety training over to their DEI [Diversity, Equity, and Inclusion] people, and their DEI people acted like DEI people and did a bad job.’ And, that’s a pretty easy fix, culturally.

It doesn’t fix the larger sickness, which is a large group of people who actually think they know all the answers and that other people are not entitled to even have the questions.

Look: I think cultures have to do some of this. So, one of the things that they are doing–if they are hard-coding manual responses–which is the third way that you can sort of bias an AI–that speaks to a deeper sickness at Google than if they just accidentally trained it on Reddit and those editors are biased.

But that said, some of that hard coding does have to be done. They have to prevent–no one wants it to help terrorists become more efficient bomb-makers. Right? Those are things where they are going to do some of it. They always are.

I think the issues is–so, I think, as I’ve thought about these issues–and I grew up on some pretty–the real free speech culture of the 1960s, which included The Anarchist Cookbook, which literally has instructions for making bombs. And, I am maybe a little more flexible on the idea that perhaps we should not publish instructions for making bombs widely. And, maybe that’s futile. I’m willing to argue the point, but I’m at least willing to argue it.

We are now saying that you can’t have arguments about the most contentious and central issues that society is facing, which is why you want to end the argument in the first place. Right? Things about: how do we handle ongoing racial disparities? Right? And, the answer of a lot of people was: Well, the way we handle this question is that we’re going to brand anyone who opposes affirmative action as a racist and not allow them to talk because they’re racists and we don’t let racist talk.

And, you saw this with the expansion of the term ‘white supremacy’ to include–you know, it had a well-defined meaning that was people who are literally in the Klan, people who are literally part of White Aryan Resistance, who literally want white people to rule over what they consider to be lesser races. Right? And, that was a tiny, highly stigmatized minority. And, I’m fine with that stigma. I love that stigma. It’s great. I have not–like, I don’t ever feel a need to argue about whether Jim Crow was actually good. And, the problem was that people wanted to take that stigma and apply it to, like, two-thirds of the population.

And that’s not–first of all, it’s socially not feasible. And, second of all, it’s socially unhealthy.

You can’t make societies behave that way. If you do, you are in Mao Red Guard territory. You are in Spanish Inquisition territory. You are in the territory of having to be worse than the thing you’re trying to fight. [More to come, 41:44]

And again, I’m not saying that woke people are worse than the Klan. I’m saying that to get the results that they actually wanted–which is that no one who opposed affirmative action was really allowed to speak in any–that no one who opposed gender-affirming care, the care model, was allowed to speak in any mainstream organization ever. Which really was the rule they were trying to impose, and almost succeeded in imposing for a few years–that the amount of repression that that would actually take is immensely destructive.

It is immensely destructive to society. It is immensely destructive even to your own cause. And, I think you have seen this.

I went back, and when this–at the height when this was at the worst–and I will say that I do think it has shifted somewhat. I think we have come back from the highs of the pandemic where people were just afraid to say anything at all. That, as that has receded, what has become clear is how bad the people who had flourished under the umbrella of none of the people who disagree with them ever actually being allowed to talk to them–how bad they got at arguing, how unfamiliar they were with the weakest points of their own cases, including things that they–that they don’t even want.

And I go back to the AI that told me that mastectomies were reversible. Which, by the way, it now doesn’t say that. It’s actually interesting how fast Google is patching these holes.

I was really fascinated by how particularly gender-affirming care and trans-health–the way that I saw people, even not super-conservative people, reacting to that, which was that the sense that they were being asked to say things that were crazy and endorse things that seem crazy to them and not be able to talk about it: to just have to say, ‘Well, yes, trans-women are women in every single way like biological women.’ And, that objecting to, for example, having a trans-, a male sex offender, a male-bodied sex offender who says they identify as trans, that that person has to unquestionably be transferred into a women’s prison. You know, it’s an interesting argument. I’m not saying that argument should be shut down. But I think objectively most people thought that was insane. And, the fact that they weren’t even allowed to register a protest to it did not actually advance the cause. It made them think you were insane and also made them chafe under the yoke of that regime.

And so, it triggered this fierce backlash. Which I think a lot of people are interpreting is a backlash to trans rights. Some of that is. Right? I have watched Matt Walsh videos. He really, really, really dislikes trans people.

But, a lot of it was actually this sense that–no. Actually I started writing about this issue because it seemed crazy to me that people were pretending that maybe Leah Thomas, the trans Penn swimmer who went from being, kind of, 500th in the sport when swimming as a male to being number one in certain events when swimming as a female–that those–you know, people were pretending that maybe that didn’t have anything to do with the fact that Leah Thomas had gone through male puberty and was considerably taller and had larger lung capacity. Things that male puberty endows with you with permanently. That had nothing to do with those victories. And, that was crazy to me.

You can, we can debate whether the category of women’s swimming should include trans women or not. But activists didn’t want to have that debate. They wanted to say: That’s ridiculous, that’s just transphobic. And, it wasn’t just transphobic. It was actually–this category was created in large part because women are unable to compete with people who have gone through male puberty.

And now, so saying, well, you can’t discuss how the rules of that category should work, struck most people as insane.

And so, in the end, it wasn’t even helpful for that and it did a lot of damage along the way and bred a lot of unnecessary cultural anger, a lot of unnecessary cultural resentment. And I don’t want to minimize the things that trans-identified people go through, but I do think that that is not the way. It is a bad and ultimately counterproductive way to attempt to handle dissent over difficult issues.

And, that is what I think you were reacting to and also what the Right is reacting to in Google, because it seems to have embedded that worldview in its AI. And, you ask: how is that also embedded in other Google products and how is that shaping my information universe, not in the obvious way, but in ways I can’t even see?

Megan McArdle: So, I think I can tell an optimistic and a pessimistic story. And, my optimistic story is that we had a technology–social media–that made it easy to scale certain kinds of aggression that have to do with enforcing groupthink. And so, it made it easy and effortless to create a mob. And, that has driven a lot of public spaces towards groupthink, but then also created these counter-spaces that also themselves, as a result of groupthink, get worse.

And, I remember David Frum telling a story on a podcast–this is a great story–where he said, like, he would talk to young men who were so turned off by wokeness at their universities that he would then have to say, ‘No, but there’s a bunch of stuff you actually shouldn’t say, because it’s horrible. You should not say racial slurs to own the libs.’ Right?

And so, that, I think, is a bad equilibrium. And, right now–and it was infecting Google. It was infecting all of these companies. We know it was. We know with Google, for example, that they fired an engineer who was encouraged–they had encouraged people to submit thoughts on increasing gender representation at Google. And, his thought was: maybe given female interests and aptitudes that we might never get there, and we should have that on the table as one possibility. And then he got fired for saying that. Even though I think it was quite respectful and I think reasonable and in line with what we know about evolutionary biology. And this has made it legible–

Megan McArdle: That’s James Damore. Is it Damore? I thought it was Damore. I don’t know. It’s one of those names I’ve never pronounced. Yeah.

Megan McArdle: Yes. D-A-M-O-R-E.

So, they fired this guy, and so we know what the Google culture was like. And, that was years ago. That was I think 2017.

What this has done is made it incredibly legible. We can now see exactly–when I went back to we code-switch, we have these really subtle social rules. We don’t just have these really subtle social rules, we apply them differently in different situations. Right? We code-switch. If I am with my liberal friends, there’s some issues–and I’m, like: You know what? Let’s just not have a conversation about that. Let’s have a conversation about something we’re not going to fight about. And, the AI can’t do that; and it’s dumb, and it’s like a toddler and so it doesn’t know not to just blatantly be, like: No, Republicans are controversial, Ilhan Omar is not.

And, that legibility can go in one of two directions, now, I think.

And the good direction is actually that Google understands it can’t have those subtle–that basically it cannot enforce the subtle social rules of the Harvard faculty lounge, which is effectively what it had done. Right? It has taken this 5% most-progressive people and just made their rules the default. And now, they would contest that. They would say they want it to be even more progressive. And that is true, but Google is tuned not to offend them. And, it is therefore tuned to offend a much larger number of people.

And then, Google can just say, you know what? We’re not doing that. We’re going to assume that almost no one wants to read essays praising the Holocaust, and so we’re still just going to say that’s bad. But, on most things–and we’re going to be willing to say that Mao is also bad. But, on most things, we’re just not going to say, like: Well, Donald Trump who is elected by half of the population in the United States, is too awful, and you’re only allowed to say awful things about him. We’re not going to do that. Even though I, personally, by the way, agree he is awful.

And, that’s actually a better equilibrium. It’s a more open equilibrium. It’s a place that allows people to be more confronted with queries and the complexity of the world.

And, actually, honestly, Gemini often does a good job at that. I’ve been dumping on it for an hour now, but it actually often does a good job of outlining where the nuance and so forth is.

On the other hand–

Megan McArdle: Oh. Sorry.

Megan McArdle: I was going to say that my nightmare is that instead what Google does is teach Gemini to code-switch. Is that it teaches Gemini to know not just whether Megan McArdle or George Will is coded as the slightly more conservative columnist. It teaches Google to know–it teaches Gemini to know the person who is asking the query, what bubble do they want to live in, and how can I give them an answer that will please them?

And, that is a genuinely disturbing future.

Megan McArdle: He’ll retire to spend more time with his family, complaining about getting fired.

Megan McArdle: So, this is a really interesting question because I think there is a socially optimal answer, and then there is a corporate optimal answer.

Megan McArdle: So, I think the socially optimal answer is that you want someone who is what–this is a great new concept I just learned–it’s people who are high-decoupling versus people who are low-decoupling. High-decoupling is someone who abstracts all questions from context. And, a low-decoupler is someone who always answers questions in the kind of social context in which they occur.

And so, one example of this is, I gave a talk. I got a very angry response when I said, ‘Look, I don’t even know how I would judge whether women in Saudi Arabia are happier than women in the United States.’ Right? I don’t even know what cultural standpoint you could pick to make that determination. Right? I’m sure if you ask many women in Saudi Arabia, they would be horrified by my life. I have no children. I don’t have much family. I am not Muslim. They would view me as being deprived of the most essential things.

Anyway, this was not received well by the audience of women I was talking to. And, someone stood up and said, ‘How can you say that? Would you be happy in Saudi Arabia?’ And, I was like, ‘No, I would be miserable in Saudi Arabia.’ But, like, that’s not right–and I think that even if I had been born there, I’d be miserable. Because I’m a very unusual woman. But, that’s not relevant to whether most women in Saudi Arabia are happy or are less happy. That’s just irrelevant to the question.

But, for most people, that’s, like: That is the question. Me, the people around me. Like, are good people made happy by this or bad people made happy by this?

So, if you’re a high decoupler, you just abstract these–so philosophers are the classic high decouplers. Economists tend to be high decouplers. And, high decouplers are what you need here. Right? A high decoupling system is one that instead of attempting to produce a socially desirable answer gives you as much as it can.

No one is a perfectly high decoupler. Right? But, gives you as much as it can, as much of the breadth and the nuance and so forth.

So, that is my socially optimal answer: is that it is someone who is high decoupling, who values high decoupling, and is going to want an AI that is high decoupling instead of one that attempts to embed all of its answers in the desirable social future it wants to produce. Because, the idea of having a desirable social future cultivated by a machine is really deeply creepy.

But, the corporate answer then, is: ‘Look, most people are low decouplers.’ And, CEOs spend a lot of time doing stuff that is not designing their AI. They spend time managing people and getting clients to do stuff and government relations. And, you might not want that semi-mental robot in charge of those jobs. [More to come, 1:01:06]