Transcript: Community Broadband Bits Episode 348

This is the transcript for episode 348 of the Community Broadband Bits podcast from the Institute for Local Self-Reliance. In this episode, Christopher speaks with author and researcher David Weinberger. Their conversation touches on many topics, including the importance of the Internet, how the concept of knowledge has changed throughout time, and the promise of machine learning. Read the transcript below, or listen to the episode.

David Weinberger: It's a library unlike any we've had in that you can casually dip in, spend literally the rest of your life exploring a topic by following links — links that we made for one another. This blows apart just about every idea about how the world goes together.

Lisa Gonzalez: Welcome to episode 348 of the Community Broadband Bits podcast from the Institute for Local Self-Reliance. I'm Lisa Gonzales. David Weinberger from Harvard's Berkman Klein Center and Google joined Christopher this week. As a senior researcher, author, and writer in residence, David has spent much of his time analyzing the Internet and how it has affected society over the years. Christopher and David take some time to discuss David's observations and conclusions, including addressing why the Internet is important and valuable despite its negative characteristics. The conversation also looks on how knowledge in the age of the Internet has changed and taken on a whole new meaning, not only in how information is distributed, but in how it's gathered, the extent of its reach, and the expanding responsibility that accompanies the changes. Chris and David also discuss machine learning, David's hopes and concerns, and how it expands innovation. Now here's Christopher with David Weinberger.

Christopher Mitchell: Welcome to another episode of the Community Broadband Bits podcast. I'm Chris Mitchell with the Institute for Local Self-Reliance in Minneapolis, Minnesota. Today I'm speaking with David Weinberger, the senior researcher at Harvard's Berkman Klein Center, also writer in residence at Google, working on machine learning, and the author of several books that I've enjoyed: Too Big to Know, Everything Is Miscellaneous, and author with several others in The Cluetrain Manifesto, and another book we'll tease in a second. Welcome to the show, David.

David Weinberger: Thanks. Great to be here. And by the way, you gave me a little bit of a promotion by making me THE senior researcher. I'm just a senior researcher. It's a little bit like the Internet.

Christopher Mitchell: I'm just excited I got through that without an edit, so I'm going to keep it.

David Weinberger: Yeah, well please. Sure. I'll take it.

Christopher Mitchell: You know, one of the things that I really enjoy about reading your work, David, and in the conversations we've had in the past too, is I feel like whether you're writing or talking, you're often both making it an argument while chatting with the reader or the person you're talking to about the argument you're making. It sort of gets meta in some ways, and I always appreciate that, so . . .

David Weinberger: I am not aware of that, but thank you. I should probably try to make it up, take it up another level.

Christopher Mitchell: IN one of the articles that I was reading, for instance, you were talking about how the Internet was paved over and then later you came back and said that actually it really wasn't pavement and you're sorry you brought it up at all. Those are the kind of comments I really enjoy.

David Weinberger: Well, thanks.

Christopher Mitchell: So anyway, you've written some other books too. Those are just the books I read that I listed. I highly recommend those, but I wanted you to validate something now that I've buttered you up, and that's that I am correct in capitalizing the word the Internet when I'm talking to the global Internet that connects all the other networks.

David Weinberger: Yes, you are correct. The style guides say you and I are both wrong. I have a book coming out in May, and I lost the argument with my editor both over capitalizing "Internet" and "Web" when talking about the World Wide Web. The style guides are just against that idea. I think it's — grammatically you can argue it and that's fine, who cares? Politically, I think it's a mistake because especially as the Internet is in danger of fracturing, if it hasn't already depending on how you look at it, I think we need to have a more and more vivid sense that there is this thing. It is the Internet. It is a single thing that touches everyone, and it's the same thing. The experiences of it are different of course, but it's the same thing that touches everybody. So I am very much, even though I lost the argument — I had to put in a footnote in, might be the first footnote in the book — it's an early one anyway — acknowledging that I had lost that argument.

Christopher Mitchell: You say that grammatically, you can argue it. Frankly, I've never understood it. I mean, I always think of the Taj Mahal. There's one of them, really. I mean there's facsimiles, but there's one and we capitalize it because it's one very special, unique thing that's a proper noun. I just fundamentally don't understand how the Internet is not a proper noun.

David Weinberger: Well, in the same way I think — and I lose this argument all the time — it's the earth is lowercase "e" and Earth is capital "E" and same for the sun, which seems to me just totally backwards to begin with if you're going to do it, and I don't get it. But there's a certain level of arbitrariness in some of this, I think, in language, could be.

Christopher Mitchell: I was just going to say, I feel like reading your books, you get a sense that there's a lot more arbitrary in our lives than we're prepared to accept and that we realize at first glance

David Weinberger: You are absolutely correct. That's a theme throughout my books and my life. We normalize everything. We look for generalizations, general principals because in some sense that's how language works and how thought works, but we also have tended to emphasize and valorize those generalities as truths as opposed to — and this is unfair, but I'll say it anyway — as opposed to being more or less shortcuts. This is one of the reasons why I've gotten so interested in machine learning, and we may come back to that in the conversation, but machine learning does not begin with a generalized model of how its domain works. Here's how business works: why, there are the following 30 variables and factors and here's their relationship. You know, you model the business. We do that and it works pretty well. It works well enough that we continue doing it, but machine learning doesn't work work that way. You give it the data, you don't give it the model, and it creates its own model. And the models that it creates are highly probabilistic and they connect individual data points and can be gigantic webs — lowercase "w" by the way — gigantic webs of correlations, probabilistic correlations. I mean, it can be so complex, we simply cannot understand how some of these systems come up with some of their results. It does not start with nor does it always, or even that often, come up with generalizations, with general principles. It's a really different way of thinking about the world, and in some ways I think it's more accurate than the shortcuts that we take. No offense to Newton. I don't want to argue against Newton's laws because they seem to be pretty good.

Christopher Mitchell: They got us pretty far.

David Weinberger: Yeah, we've done okay with them. And he was a fairly bright fella, got to give it to him.

Christopher Mitchell: So we're gonna focus on something you've given a lot of thought to, which is the Internet. I wanted to have you on because I've done almost 350 shows now, most of which just take for granted how important the Internet is and expanding access to it and things like that. And one of the things I was recently reflecting on was I don't know that I've done a good job describing why it's important, not just because kids need an education or because Netflix is nice, but why the Internet is more important for reasons beyond that. And so, you've given this a lot of thought, I know, because you and I've talked about it and you've written about it, even caring so much about the book that you published it with lowercase "i's" in it. So the question I want to pose to you as we get into this is there's this question that you said you often get, which certainly is headlines in magazines time and time again: "Is the Internet making us dumber?" And even beyond that, you know, is the Internet or Facebook to blame for real harm that's being done in terms of people's anti-vaccine beliefs or, you know, climate change denialism or things like that? How do you respond to those sorts of questions?

David Weinberger: First of all, I admit that in many ways the Internet has been a destructive force. I don't want to argue — I think it's not only pointless, it's wrong to argue against the sorts of things that people point to. I find that the negatives have been very well covered over the past 10 years.

Christopher Mitchell: Right.

David Weinberger: There's lots more to say, but I don't feel like I have anything really to contribute to that. I want to acknowledge the negatives and I have no problem doing so, but I think it's also — I agree with you about this Chris — that it's important even while one is pointing out all the negatives, that we remember how radically different and beneficial the 'net has been, even with all those negatives, because otherwise we're in danger of losing it. I mean, it's possible to lose this thing. It's something we built; it's something that we can lose. I mean, I have an odd interest in this, which is not only in sort of the social formations that have been so beneficial to us — they have their negative side but that have been so beneficial to us — and I don't know that we want to talk about the following, but personally my major personal interest is in ways in which the Internet has led us to think about the world differently, about how the world is put together and power relationships in the world and what it means to be social. And I'll give you a really simple example — I mean, obvious example. I actually don't think it's a simple example, but it's a clear one.

Christopher Mitchell: A very complex, clear example.

David Weinberger: Yeah. Well, just because the ramifications of it are so, so huge and that we so take for granted now. The hyperlinks. You know, we're now in the 30th year of the World Wide Web. A hugely important component part of it are hyperlinks, right? And they are relatively new. There were systems before, and in fact I worked for a company that made one, that enabled people to have hyperlinks between . . . But you know, they didn't have any traction in part because these systems were not open systems and in part because at least in some instances, like the company I worked at which is called Interleaf — in the '80s you could do hyperlinks, but the creators of the documents, which tended to be, you know, technical documentation departments, had bought this expensive system — they were hard coded. You had to compile a system. You wanted to add a new hyperlink, you had to recompile the system and redistribute it. Hyperlinks on the Web, because of their accessibility, because the web took off so quickly and had so much material, they change our idea about how knowledge and ideas go together, about whether the right approach to knowledge is always to try to get it concise enough that can fit in a book that can fit on a shelf, whether there's sort of a centralized control over what counts as included in a topic or related to a topic. In a book, the author is in control of that. That's fine. It's one model, but it was basically our only model. The author gets to say what the book is about and what references he or she is going to make and the links that go out and those links are printed and so nobody really follows them anyway, and on the Web, anybody can link to anything. We've built incredibly quickly this massive, unprecedented in human history, bottom up, democratic — a little "d" of course — but also individualistic web of connections among ideas. That's a web of meaning. Here's one simple sentence; the rest of the Web springs out of it because everything on the Web is connected. So in a literal sense, you start anywhere, you can get anywhere. That's why it's important that we talk about THE Web and THE Internet, right? And so, there is this gigantic web of meaning, a semantic web if you will, that has been built by individuals from many, many, many cultures and backgrounds and interests. We've never had that before. We've never had anything like it. And anybody can speak and anybody can link and draw the connections she wants. So this is a different idea about what knowledge is like, what meaning is like, what it means to know something. It's a tool like we have never, ever had. It's not simply that the Internet is an information library the way that it early on was thought of and talked about. It's a library unlike any we've had in that you can casually dip in, spend literally the rest of your life exploring a topic by following links — links that we made for one another. This blows apart just about every idea about how the world goes together.

Christopher Mitchell: Well it seems like it's — it's funny because the two words that come to mind are both democratic and anarchistic in terms of it's choose your own adventure. One of the things I think about almost every time I'm in a room of people is one of the things I've learned, I believe from reading you, which was that in a group of people, the smartest person in the room is the room, I think is the phrase actually.

David Weinberger: Yes, the subtitle of one of my books.

Christopher Mitchell: So is that Too Big to Know?

David Weinberger: Yeah, it's part of it. It's a very long subtitle, but that's buried in there somewhere.

Christopher Mitchell: Right. I'm pretty sure I read the whole book, not just the subtitle.

David Weinberger: Stopping at the subtitle would have been fine.

Christopher Mitchell: Sure, well no, but the sense in my mind is that, you know, there's a difference between, in the past, I think one could accumulate knowledge by locking oneself into a very large library and just learning and learning and learning. And increasingly we've gone beyond that. It's not even about what a single person can accumulate in terms of knowledge. That's very limiting to think of it in that way. And to some extent when you go back to what's the difference between the '80s and now, I feel like there's just this difference of you really need to get a group of people together to do anything interesting because you just don't expect one single person would have that kind knowledge.

David Weinberger: Yeah. We spent a few thousand years with the assumption based upon the necessity that the way to know something is to get it into a single skull that's through single person project. That became a more pronounced tendency as time went on, and that requires a very rigorous discipline over topics. I mean, how long can a book be? Right, a book was and in many ways still is, I guess, the fundamental unit of knowledge. And we're going through a phase shift here, but books are really, really short. And anybody who's written a book knows that you have to be very disciplined about what you're going going to talk about, which means you can't talk about most of . . . It turns out that once you take the paper out of the system of knowledge, which the Internet and then the Web very effectively did, that a lot of the ideas about knowledge turned out to be based upon the limitations of paper. So for example, once you publish something on paper, you really can't change it. It just settles. But knowledge also has had this character property of being the stuff that we have settled on as a culture. If it's still being debated, we say, well, no, we don't know yet, but once it's settled, then it can become knowledge. We've had to filter. Knowledge has been filtered. Right from the ancient Greek origins of it, knowledge was a category that came later than the category of opinion and it was the set of opinions worth believing. In the west, that's been a guiding a property, but it means that knowledge is always filtered. Books also are highly filtered. Very few of them get published relatively, and very few of them can fit in any library, and there's no library that can fit all of them — no physical library can fit all of them. Libraries have to throw out books. I say, this as somebody who spent five years co-directing a library innovation lab. I don't mean to slight libraries. It's a fact of physical life that libraries have to throw out some books or sell them or whatever in order to make room for the new ones. And so, knowledge has always been filtered, and I don't think that it's an accident that the properties of knowledge have also been the properties of paper. You take those out and knowledge really begins to change. It becomes something that multiple people do in networks by building networks, connected networks, of knowledge or webs of knowledge. But one of the biggest changes I think we are now living through and is making us very nervous for understandable reasons, is that when you have a web of knowledge, that web consists of differences among the people. And sometimes they are very friendly differences in which one person knows about one topic or whatever, but inevitably are also webs of differences of ideas. There's disagreement. The knowledge never settles. And we look out across this field and we see disagreement, but that's in fact — the dream of knowledge is that everybody agrees with it; the fact of knowledge has always been that that has never been the case. And I'll give you a positive example of this in scholarship. I think we are quite happy, I assume we are quite happy, to have traditional networks of knowledge — we don't call them that, but that's what they were — traditional networks of knowledge among scholars on, say, Shakespeare who spend — we don't want them all to agree. We want them to disagree. We hope that they're, you know, civil and the rest of that stuff, but that disagreement is where all of the interest is. It turns out that that's not just a humanities thing though. It turns out that the fact — and we don't like it, I understand that — the fact is that we don't and never will all agree about anything. But now we're on a single thing, the Internet, where we see those differences and we see that they don't get resolved and it's very disturbing to us.

Christopher Mitchell: Yeah. I think one of the challenges that we have is where that spills over, and I think this comes down to whether you're talking about, again, like anti-vaccination confusion or even disagreement over basic facts in politics today. In part, because we're in an area, I think, in which people have a sense that you can find evidence for whatever you want, and we haven't yet adjusted. People haven't adjusted to the reality in which you have a higher responsibility if you're making an argument than just saying, "I found a convenient fact." You have to go beyond that, if we're going to do interesting things as human beings,

David Weinberger: The optimist in me, which is consistently — that person is consistently wrong, unfortunately, so feel free to slap them down. The optimist in me says that we are at an evolutionary — eh, I hate to use the phrase — inflection point.

Christopher Mitchell: Would you say the paradigm is inflecting?

David Weinberger: Inflecting paradigms is — that'd be a great traffic sign, don't you think? Warning: paradigm inflection ahead. I think there's some evidence to be somewhat optimistic about this, where we have to be more meta, we have to be more aware because you know, I hate to say it, but the old paradigm — now you've got me using the paradigm word — the old paradigm of knowledge was not actually. . . Because once something is settled, it's settled. It's known, it's done. You don't bring it up again unless there's some good reason. Everybody agrees, which of course they never did, you just couldn't hear the people who disagreed because they were disenfranchised.

Christopher Mitchell: They didn't get the paper.

David Weinberger: Yeah. And so we had an illusion that there was unanimity around knowledge, and so that lets you believe things without having to be very meta about it. You know, it's just true, it's just right, and everybody knows. Now, the conversations I think have to — really should is what I mean because I don't know that they will — become more aware about the role of evidence, less certain about one's own position, and more humble. I think there's evidence that in many areas that's happening. There's also pretty clear evidence that there are lots of places where people are becoming even bigger, let's say, jerks. That's not the technical term I would use, but bigger jerks than they ever were. I'll tell you a secret hope of mine, about machine learning. If we accept what seems to me to have been true — and others — to have been true for thousands of years, which is that we understand our minds often on the basis of using the metaphors that we gained from the tools that we use, then if the new tool is becoming — and we certainly saw this in the computer era when suddenly everything about our minds and then the world became information. The term information became a hugely important term when it had not been one, even though people can't tell you what it means. And I don't mean in the science sense, just it's a placeholder word for something. Anyway, so we've seen in the computer era that we've refashioned our idea of ourselves in terms of information and inputs and outputs and so forth, and if the same thing happens with machine learning in a particular way — machine learning is always probabilistic. It relies upon measures of confidence in order to do its work, and if we begin to understand ourselves along the machine learning model than maybe a good thing from my point of view would be if we picked up on the [fact that] all statements have a confidence level, that we recognize that they're all uncertain.

Christopher Mitchell: Right. I actually think that's really valuable. I'll tell you what I got out of it is something that I do think a lot about, which is this idea of how much confidence do I have in this thing that I'm saying. And if I'm speaking to people that are asking me for advice, I'll often say, well, like I think this thing and I'm very confident about it, and I'm about to say this other thing later in the conversation, and I'll say, look, I'm much less confident about this, in part because if I'm wrong about this, you shouldn't assume I'm wrong about the other thing because I'm more likely to be wrong about this thing. It's hard to know these things.

David Weinberger: Yeah, and also you are an honest and competent consultant and advisor, and I say this having known you for awhile.

Christopher Mitchell: Well thank you.

David Weinberger: You're welcome. And yet — so here's one of the popular negative things about to say about the Internet, which is true, which is the Internet is an attention economy and you gain attention because it seems to be a trick of the mind. One of the — this is in parentheses — one of the things that's been fascinating to me over the past few decades, because I'm old, is watching the extent to which the idea that we are rational creatures, that this is our destiny and our being, seeing that idea eroded by behavioral economics and much more, in which basically the brain now seems to many of us to be all optical illusions all the time except they're cognitive illusions. Nevertheless, one of the optical illusions is that we pay attention to strong or outrageous statements. It's not hard to see why. And so, the Internet economy as many have pointed out is an attention economy in which outrageousness is rewarded. That goes against the hope that we will become a more humble, measured, meta creatures. And then I want to say the third thing is, well, you know, there's sort of an Hegelian dialectical synthesis of this, which I think is maybe one of the dominant modes that you find on the Internet, which is people who assert things in a very overly bold voice do so knowingly and are heard as purposefully, knowingly overstating because it's funny, often. You see this in places like Reddit, where that's a pretty common form of expression. And so, it's both the attention grabbing overstatement, but done archly often with a signal, sometimes very implicit just by the subreddit that you're in, that no, we know that this is just — we're just being outrageous because there's some truth in what we say, but it's also pretty funny to talk this way. We can spawn a really funny thread if we talk this way.

Christopher Mitchell: Well, I think some of it is also — I think there's a sense of frustration of being unheard. You know, I feel like we see this around, for instance, the work I'm doing right now around 5G where we've become extremely snarky because, you know, honestly, part of it is that we feel the expectation that if we say smart things and we figure out smart things, people should listen to us. And then if they don't, we start to feel frustrated and then you kind of lash out in that same way. But you know, I think if you go back 25-20 years ago, people didn't have an expectation that they had any means of influencing those sorts of events unless they were born into the right family or went to the right schools or something like that.

David Weinberger: Yeah, absolutely. One of my deep concerns about my early views of the Internet and of the Web in particular, which, you know, it goes back to early nineties-mid nineties, and views that I still hold many of and have expressed some of, is that at the time I was a middle-aged, middle-class, well-educated white guy, and so the Internet was like a dream for me. It was like made for me because in some ways it was made by people like me. And the early Web fulfilled because, you know, initially it was not poor people and it was mainly Americans and other well-developed western countries and the like

Christopher Mitchell: People spoke English, had a certain set of expectations and knowledge.

David Weinberger: Yep. You have tons of very technical and highly professional and educated, you know — it was an Internet of privilege. And that allowed me a certain set of fantasies which were fulfilled at the time, but were destined to — and I did know this and write about it, but not sufficiently — fantasies that were, you know, not going to last as the Web reached to people who weren't like me.

Christopher Mitchell: One of the things that I think about as we accelerate that forward to some extent, this idea of people who are like me or have like interests being able to gather around. You know, I told you I wanted to talk a little bit about what we think the future might be, and you said, like an intelligent person, I have no interest in making predictions.

David Weinberger: Well, it's not that I don't have an interest. I have tons of interest, it's just —

Christopher Mitchell: Maybe you've learned enough lessons.

David Weinberger: Yeah, especially since the book that's coming out, Everyday Chaos — sorry, that's a plug. "Available for preorder now."

Christopher Mitchell: From your local bookstore. So the prediction I want to sort of root around in is this idea that, you know, if you look back to 30 years ago, something like CRISPR comes around, this idea of being able to edit the DNA and make dramatic changes, which is still a work in progress in many ways. Nonetheless, it happens. Probably someone writes about it in a journal, and then maybe over the next six months some people read the journal and other places and they iterate, and three years later some more people learn about that in a different journal. And now, I feel like instead it's more like as a lab in South Korea is iterating, there's a lab in California that's iterating, and over the course of a year, you have 20 or 30 years of scientific progress because of the Internet. And when people talk about the Internet as though, you know, it's only a form for clicking outrageous headlines, I think about things like that and just the way that — I'm really fascinated to see what happens, and this will be good and bad. Change sped up just means we get the good and the bad faster. But I feel like when I look at the history of innovation, and so much of it comes from different groups learning about different ideas, if we just have so much more of that, I feel like we're going to see much more change in very interesting ways. So I'm curious what you make of that, if I'm missing anything here.

David Weinberger: 100 percent. The only things I would add to it I know you agree with which is that this is not simply a quickening pace of existing processes, but there's so much sharing and collaboration, that it is the networking of knowledge, right in front of our eyes. One of the things I think is hugely important and is obvious — this is why I don't make predictions. I mean I really try not to. My actual interest as a writer is in trying to read what is already here to show often why it's why it's deeply weird.

Christopher Mitchell: You mean why, even if you knew everything that was happening a hundred years ago, you wouldn't have guessed we'd end up here. Like, we weren't destined to end up where we are. Is that what you're saying?

David Weinberger: Well, yes. It's so wildly contingent, but that's not the sense that we've had traditionally. We certainly recognize the contingency, but we also look to the general rules and we look to the trends, and it's just, you know, history is nothing but a series of unpredictable, unlikely events. I mean, wildly improbable events. So one of the things that to me is really exciting, that is increasing the pace of innovation, is our explicit and sometimes implicit attempts at making things interoperable. That is, interoperability, as you well know, is when an item from one system turns out to be usable in another system, often in an unpredictable way. And so we are increasing the unpredictability every time of the world in very fruitful ways as we increase the interoperability, which the Internet has done incredibly well for the sorts of materials that it deals with. Every time somebody comes up with a new data standard or a protocol for sharing information or set of services or an open platform —

Christopher Mitchell: Or a viedo that tells you how to, like, hook these two things together, right? I mean, like, just do-it-yourself type stuff.

David Weinberger: Absolutely. I mean, it's actually a great example because all of that stuff is an accelerant and what it accelerates is not only knowledge and new services and products and gives people control over the things that they use that they didn't create that can make something new out of it or tune it to the way that they want. All of this we take for granted. We take it for granted even in video games. You know, video games, one of the earliest examples of reconfigurable systems, modding where you could take a game and change — the game makers enable you, let you and sometimes enable you, by giving you tools to change their own game. I mean, it's very different from a Henry Ford model of how you build a car. You know, 19 years of the Model T didn't change.

Christopher Mitchell: Or my Toyota today.

David Weinberger: Exactly. Yeah. Right, try to mod that.

Christopher Mitchell: I just want it to stop beeping at me when it's below 37 degrees outside. I live in Minnesota. It's always below 37 degrees. Stop yelling at me, car.

David Weinberger: Well, no, that's your fault. You could move. Really, that's just shameful, Chris, sorry. Yeah, so interoperability is an accelerant for this sort of — it makes the world less predictable and that increases the pace at which we innovate. And we have not — there's so many efforts in so many areas to increase interoperability. We call it different names, but it's like — I'm going to go back to Newton who I've sort of mentioned a couple of times — but you know, gravity, it's pretty good. Universal Law of Gravity seems to be pretty much right. Newton discovered these causal relationships, but interoperability is also a way in which two things can interact but we get to design the rules. We get to decide how these things are going to be able to interact, and that is taking us to a world that we cannot possibly, possibly predict. And it's a dangerous world too. I mean, CRISPR has wildly horrible applications possible, right?

Christopher Mitchell: Absolutely. I mean, it's one of the things I worry a lot about. I mean, I view it as — perhaps as a result of the specific sci-fi that I've read — I view it as, if it doesn't kill us off or kill us off in sufficient numbers, it will give us the tools to avoid killing ourselves off with climate change because of the ability to change organisms to remove carbon from the air and things like that. But yeah, I mean, I know that in my lifetime, if CRISPR provides the kind of things we expect it will, that terrorist groups will be finding ways of trying to do horrible things with biological weapons, you know? And so, it's a very scary future, frankly.

David Weinberger: Yes. It's horrifying, terrifying, and makes, you know, concern about the Internet seem like small potatoes — potatoes you should pay attention to but . . .

Christopher Mitchell: Without the Internet, none of this stuff happens though, I mean, I think.

David Weinberger: Yeah. That's an excellent point. I thought you were going to suggest that CRISPR could save us from climate change because we would be able to develop gills.

Christopher Mitchell: No, I would love that though. I think I might take that over wings if I had my choice.

David Weinberger: That's an interesting choice.

Christopher Mitchell: Well you know, I think it may be less crowded down there because everyone else is going to take the wings.

David Weinberger: Here's the next sci-fi novel.

Christopher Mitchell: Well let me ask you though about this, right. I mean, so in one of the articles that you provided that we'll include in our link, you write about — you know, I think a person like me might just think that when you think about machine learning and I think about just generally — which isn't machine learning — smart machines, the difference between Deep Blue, which was IBM's effort to win chess and AlphaGo, they seem like they're both just really intelligent machines that can do things that I cannot do or in fact any human can do, but they're fundamentally different. And I'm curious if you can tell me why it matters that one uses machine learning and the other doesn't.

David Weinberger: With traditional computers, a developer comes up with a model of the world: the pieces that go together and how they relate, which things matter and what their relationships are. And that goes along with our old idea that knowledge. It has to be a very reductive idea. That's why we make spreadsheets for our businesses, but nobody, you know, if the factory catches fire, nobody blames the spreadsheet nor should they. So I mean, they work. They're better than nothing, but they are idealized visions of the factors, you know, isolated set of factors

Christopher Mitchell: And perhaps hiding patterns that we can't see because of the way we construct these models.

David Weinberger: Yes, that's exactly it. So instead, with machine learning, you provide data. You don't give the machine a model of the domain. You give it the data. All it knows are the numbers. It has no idea about what those numbers stand for. And it iterates and finds correlations, relationships among those numbers, building vast, intricate networks in which one data point may be connected to thousands of others with weights about, you know, their probability, likelihood, resulting in neural networks and that produce usable results — that's why we use them — but in some instances do so through networks that are simply too complex for human brains to understand them. This, at its best, when it works — and I have to put in the disclaimer, there's terrible dangers in this as well. The one that is most often talked about, which is appropriate, is that because machine learning makes models based upon data and because we live in an unjust world, that data reflects injustice. And so the models, unless carefully managed, will reflect and maybe amplify those biases — the biases in the data, which are biases in the world.

Christopher Mitchell: And to be very clear about what you're driving at there, the fact that for instance, if I'm a youth and I am engaged in shoplifting, I am more likely to be arrested if I'm a person of color. Therefore, the system that's looking at the data will start to assume people of color may be more likely to commit crime.

David Weinberger: Yes. I'll give you another quick example, standard sort of example. If you are using machine learning to call resumes who should get an interview with a human, and you use existing data, more than regrettably women will not correlate as highly with senior management jobs as men will in almost all industries. And so the machine will learn from that and it will learn that women don't correlate very well with senior management job, so that has to be carefully controlled for. There's a huge amount of work that's been done on this, which is entirely appropriate. Nevertheless, the sorts of models that machine learning makes seem to me in their architecture to be truer representations of how the world works. There's all of these little pieces that have influences. I mean, back to Newton, everything affects everything else. Everything has a gravitational pull on everything else. Machine learning gets closer to the complexity that is the world. That's why it works better. It's why we use it. And if we can internalize that model, I think we will be better off.

Christopher Mitchell: Well, I will look forward to learning more about that in May with a book called Everyday Chaos.

David Weinberger: That was smoothly done.

Christopher Mitchell: Oh yes, I'm nothing but smooth. I've taken up more of your time than I asked you for. I really appreciate the opportunity to talk about these things, and I'm sure we'll be developing some more questions for you in the future. So thanks for coming on.

David Weinberger: Thank you. I look forward to seeing you.

Lisa Gonzalez: That was Christopher and author, senior researcher, and Google writer in residence, David Weinberger. We have transcripts for this and other podcasts available at muninetworks.org/broadbandbits. Email us at podcast@muninetworks.org with your ideas for the show. Follow Chris on Twitter. His handle is @communitynets. Follow muninetworks.org stories on Twitter. The handle is @muninetworks. Subscribe to this podcast and the other ILSR podcasts, Building Local Power and the Local Energy Rules podcast. You can access them wherever you get your podcasts. Don't miss out on important research from all of our initiatives. Subscribe to our monthly newsletter at ilsr.org, and while you're there, please take a moment to donate. Follow us on Instagram. We are ILSR74. Thank you to Arne Huseby for the song Warm Duck Shuffle, licensed through Creative Commons, and thank you for listening to episode 348 of the Community Broadband Bits podcast.