[00:00.000 --> 00:03.000]  Users online, and it's a talk roughly in three parts.
[00:03.180 --> 00:09.020]  So in the first part, I'm gonna talk a little bit about sort of robots generally on the Internet and sort of the interesting things that they're doing.
[00:09.580 --> 00:13.460]  And the second part I'm gonna talk about some experimentation that we did early last year and kind of continuing it this year.
[00:13.620 --> 00:17.040]  And finally I'm gonna talk about some of the sort of implications of that research.
[00:17.280 --> 00:19.440]  So, I'm thinking this will be fairly conversational.
[00:19.600 --> 00:22.700]  I'm timing it for about 45 minutes with the rest of the time for questions.
[00:22.960 --> 00:29.520]  But I actually just got off a plane about an hour and a half ago from Newark so, if I'm semi-coherent it will be useful to know that.
[00:30.000 --> 00:34.340]  So definitely raise a hand or otherwise step up to the mic and interrupt me if needed.
[00:34.680 --> 00:39.540]  So, by way of introduction, as was mentioned, my background is actually not bots at all.
[00:39.660 --> 00:52.460]  I'm the co-founder of a conference called RaffleCon, which basically brings together everybody who is momentarily famous on the Internet to talk about web culture, memes, the past, present, and future of funny cats on the Internet, what have you.
[00:52.640 --> 00:59.560]  Past guests have included Tron guy, Jay Maynard, who some of you may be familiar with, Moot from 4chan, and also the guy who designed Comic Sans.
[00:59.560 --> 01:05.000]  Interestingly enough, who's apparently, who's actually really haunted by his creation, actually, which is really great.
[01:05.200 --> 01:08.720]  He has since publicly apologized at the conference for creating the font.
[01:08.860 --> 01:13.840]  And actually, the worst part is that he's like a fairly credible font designer, but everybody knows him for Comic Sans.
[01:14.100 --> 01:16.200]  So, yeah, exactly.
[01:17.180 --> 01:19.080]  Some of them, some of that led into research.
[01:19.080 --> 01:24.360]  So, for a few years, I was at the Berkman Center for Internet and Society and some of the activism around that work I got into as well.
[01:24.480 --> 01:26.540]  So, I was part of the EFF for a summer.
[01:26.540 --> 01:32.500]  But a lot of the work I'm going to talk about today emerged out of a group called the Pacific Social Architect Incorporation.
[01:33.300 --> 01:43.620]  And it's basically a loosely joined group of researchers who are kind of working on technology to kind of create technology that sort of reliably exerts influence on sort of the media cycle and elsewhere.
[01:43.700 --> 01:48.380]  And it's a little bit abstract, so you'll see what I mean exactly during kind of the course of this talk.
[01:48.840 --> 01:53.560]  As a brief aside before we get started, I'm also the co-founder of an organization called the Awesome Foundation for the Arts and Sciences.
[01:53.560 --> 01:59.260]  I'm glad to report we actually got 501 status for the Institute on Higher Awesome Studies in Massachusetts.
[02:00.580 --> 02:05.460]  We basically give away small micro-grants to projects that forward the interest of awesomeness in the universe.
[02:05.460 --> 02:10.100]  So, if you know someone doing awesome things, and I'm sure you do, definitely tell them to get in touch.
[02:10.340 --> 02:16.620]  So, just to jump into it, to set the stage, I'm going to talk about sort of the increasingly odd world of what bots are doing online.
[02:17.060 --> 02:19.280]  And I think it's useful because it gives some context.
[02:19.460 --> 02:20.700]  It's also a couple fun stories.
[02:20.840 --> 02:21.360]  You know, it's 10 o'clock.
[02:21.420 --> 02:22.140]  I just got off a plane.
[02:22.640 --> 02:26.140]  And also kind of explains the experiment that we ended up doing.
[02:27.300 --> 02:32.260]  I got really interested in bots in early 2011 when there was a story that popped up.
[02:32.840 --> 02:40.720]  Basically, a professor from UC Berkeley discovered that there is this book on Amazon, The Making of the Fly, which apparently is kind of a classic in developmental biology.
[02:40.720 --> 02:46.340]  And he found that two third-party sellers were selling the book at like an astronomical price, right?
[02:46.520 --> 02:51.940]  So, the first one was selling it for the low, low price of $18,651,718.08.
[02:53.000 --> 03:00.360]  And his competitor, one that was pricing at a slightly more, less reasonable price of $23 million, $658,000, so on and so forth.
[03:00.480 --> 03:01.200]  You get the idea.
[03:01.400 --> 03:07.380]  And this is, of course, unusual because the usual sales price of the book is $42.50 and change, basically.
[03:07.380 --> 03:08.680]  And so, he's looking into this.
[03:08.860 --> 03:11.260]  The idea of the blog post was like, why did this happen?
[03:11.740 --> 03:20.120]  And so, he observed basically the pricing over a few days and discovered that one would change the price and the other one would change the price in this kind of highly rigid, formulaic way.
[03:20.600 --> 03:29.600]  And so, the conclusion they came to is basically that third-party sellers were using bots to price and post books on Amazon, but they had essentially no notion of what a reasonable price was.
[03:30.480 --> 03:32.140]  And usually, they were quite good.
[03:32.380 --> 03:35.040]  You can notice, actually, if you can see it, it's a little blurry.
[03:35.040 --> 03:40.080]  The second seller has a 93 positive review over like 127,000 ratings, right?
[03:40.180 --> 03:48.480]  So, they usually go completely undetected, but in this case, they got caught in kind of this infinite loop upwards, leading to sort of the ridiculous result that you see.
[03:49.420 --> 03:50.920]  And so, that got us really intrigued.
[03:51.060 --> 03:56.220]  And interestingly enough, in 2011, that was not the only story of sort of bots on Amazon.
[03:56.220 --> 04:03.060]  So, Pagan Kennedy, this writer for the New York Times Book Review, discovered this author, Lambert M. Surahone, which is a real, real name, who has apparently more than 100,000 books to his name registered on Amazon.
[04:11.300 --> 04:14.700]  And so, the story gets weirder, basically, as the author looks into it.
[04:14.860 --> 04:19.900]  And basically, it was published by a branch of a large German publishing house called VDM.
[04:20.400 --> 04:25.300]  But the subsidiary was incorporated in Mauritius, right, which is an island nation in the Indian notion.
[04:26.200 --> 04:29.540]  And basically, they did a couple calls to try to figure out what was going on.
[04:29.640 --> 04:35.080]  And the publisher was like, we're definitely not using robots to algorithmically generate books, if that's what you're suggesting.
[04:36.260 --> 04:38.420]  And, well, let me show you and you can decide for yourself.
[04:38.560 --> 04:40.920]  This is a representative book of what Lambert produces.
[04:43.160 --> 04:46.940]  This book... This book runs for 104 pages.
[04:48.320 --> 04:51.520]  And they're essentially spam books, which is really intriguing.
[04:51.760 --> 04:53.340]  So, basically, the author ordered one.
[04:53.480 --> 04:56.380]  It's basically a poorly assembled collection of Wikipedia pages.
[04:57.280 --> 05:00.380]  And it's actually advertised as a little thing on the cover, you can see.
[05:00.980 --> 05:07.420]  But the more you look into it, actually, the less and less human it looks, because there's sort of bot-like mistakes that Lambert seems to make.
[05:07.420 --> 05:10.620]  So, he published a book about the police, the band.
[05:11.000 --> 05:14.080]  But the sort of cover image is members of law enforcement, right?
[05:15.140 --> 05:17.260]  And so, it's kind of this really kind of bizarre thing.
[05:17.460 --> 05:22.760]  And so, there's just this sort of machine, this script, that's just generating all these books out on Amazon.
[05:23.440 --> 05:26.000]  Purchasable at prices from, you know, $10 to...
[05:26.000 --> 05:27.720]  I think one is priced at like several hundred dollars.
[05:27.840 --> 05:28.260]  It's a great story.
[05:28.360 --> 05:28.920]  You should look into it.
[05:29.900 --> 05:31.480]  So, this is kind of Amazon.
[05:31.700 --> 05:36.580]  But obviously, right, Amazon's not sort of the clearest example of bots on marketplaces, right?
[05:36.580 --> 05:40.720]  So, the kind of rise of quantitative trading on the stock market, right?
[05:40.860 --> 05:45.960]  Like, now, quantitative trading accounts for a huge percentage of the volume on a lot of equity markets.
[05:46.200 --> 05:52.840]  And, you know, they're using these patterns that trade at sort of thousands of trades a second and leaving these kind of really interesting effects.
[05:52.980 --> 05:58.840]  And actually, interesting enough, you see the same effects in the stock market increasingly that you saw on Amazon, right?
[05:58.840 --> 06:09.040]  So, some of you might have caught this really fun story, which is that it turns out that when the actress Anne Hathaway is mentioned in the news, the stock price of Berkshire Hathaway takes a small but significantly significant jump upwards.
[06:11.260 --> 06:12.520]  And they're trying to figure it out, right?
[06:12.600 --> 06:14.540]  They're basically like, this is really odd.
[06:15.820 --> 06:26.520]  And they look into it, and basically, the conclusion is that someone out there is running this incredibly sort of crude quantitative trading program that's like Hathaway is mentioned by as much of Hathaway as possible.
[06:26.800 --> 06:29.260]  It's leading to these kind of odd correlations in the marketplace.
[06:29.620 --> 06:34.420]  And you see the same sort of kind of critical events that you saw on Amazon as well, right?
[06:34.520 --> 06:37.260]  So, you might have remembered this incident in 2010, right?
[06:37.260 --> 06:43.520]  This really odd incident where I think it's still the largest single-day drop, one of them, in sort of stock market history.
[06:44.020 --> 06:50.500]  And it's basically caused by the interference of a number of bots with one another, leading to this sudden critical drop in prices.
[06:51.260 --> 06:54.420]  And only to sort of regain it like a few hours later.
[06:55.340 --> 06:58.680]  So, these are kind of funny stories, and the question is sort of what can we draw from this, right?
[06:58.740 --> 07:04.160]  Both in the case of Amazon, and also the case of just like the financial markets, generally speaking.
[07:04.160 --> 07:07.700]  And one thing is that digitization enables botification, right?
[07:07.920 --> 07:09.600]  And botification is definitely not a word.
[07:10.200 --> 07:18.640]  But the general idea is that the moment sort of an arena of human activity emerges digitally, scripts can act in lieu of humans, right?
[07:18.700 --> 07:19.720]  And that's a pretty simple idea.
[07:19.860 --> 07:23.660]  But it kind of leads to this sort of obvious and sort of interesting question, right?
[07:23.760 --> 07:28.060]  Which is, what do we do when increasingly a lot of social activity moves online?
[07:29.360 --> 07:33.420]  And obviously on some level, right, this shouldn't be considered new, right?
[07:33.420 --> 07:37.020]  So, I'm sure some of you remember Smarter Child from the AIM days at all?
[07:37.260 --> 07:37.400]  Yes?
[07:37.660 --> 07:37.760]  No?
[07:38.000 --> 07:38.460]  Well, anyways.
[07:39.340 --> 07:40.660]  Chatterbots on AIM, right?
[07:40.760 --> 07:45.960]  And because people have been socializing online forever, there's obviously been bots that have been interacting with humans forever, right?
[07:46.020 --> 07:49.220]  It extends back to IRC and BBS and so on and so forth.
[07:49.620 --> 07:52.900]  And that's one version of a bot that we've known for a really long time.
[07:53.360 --> 07:55.960]  Another version has been sort of the spam bot, right?
[07:56.060 --> 07:57.440]  So, Chanel Lambert is one of my favorites.
[07:57.440 --> 08:00.580]  She was unfortunately killed in duty just last week.
[08:00.580 --> 08:05.740]  But her entire purpose was to just express like vague indifference to things that people were seeing on Twitter.
[08:08.200 --> 08:09.880]  And these spam bots are everywhere, right?
[08:09.940 --> 08:11.900]  We know them as just a feature of the Internet.
[08:12.140 --> 08:20.860]  But what's different in this context is that the bots have been appearing on these sorts of platforms have become more important and widespread as those platforms have become more important and widespread, right?
[08:20.940 --> 08:24.360]  So, you've seen the increasing use of bots, for example, to harass your enemies.
[08:24.360 --> 08:31.140]  So, there was basically a scientist who got really tired of arguing with climate change skeptics online all the time.
[08:31.300 --> 08:33.620]  So, he basically created a bot that just went in and harassed them, right?
[08:33.660 --> 08:36.660]  He would just drop in sort of links saying like, hey, you're wrong.
[08:37.600 --> 08:39.820]  And so, he wouldn't have to spend time arguing with them.
[08:39.920 --> 08:43.500]  And I think the total amount of time wasted on the opponent's side was like enormous, right?
[08:43.940 --> 08:46.520]  Because people were basically getting into these arguments with this bot.
[08:46.620 --> 08:50.900]  And obviously, like the bot was named Turing Test, so they weren't quite, you know, getting the idea.
[08:53.640 --> 08:55.420]  And some of these things are fun, right?
[08:55.540 --> 08:58.160]  Like, they're small kind of harassing actions you can take.
[08:58.280 --> 09:00.280]  But I think the story takes on this really interesting dimension.
[09:00.720 --> 09:03.620]  When you realize that governments are starting to get into the space, right?
[09:03.740 --> 09:06.720]  So, Global Voices, which is this kind of blog project run at the Berkman Center.
[09:06.780 --> 09:07.120]  Really cool.
[09:07.220 --> 09:07.640]  You should check it out.
[09:09.060 --> 09:13.000]  Basically reported on this really interesting incident where it turned out that appears...
[09:13.000 --> 09:20.360]  It appeared that basically spam bots were being launched in very large numbers to basically attack people who are pro-protests in Syria.
[09:21.460 --> 09:24.740]  And, you know, you basically had all these accounts that were instantaneously created.
[09:24.800 --> 09:32.160]  And the whole idea was to either fill sort of the hashtag with noise or to actually kind of actively argue against and harass the opposition.
[09:32.740 --> 09:35.760]  And there's a little bit of an arms race actually going on internationally.
[09:35.780 --> 09:42.460]  So, in the recent Mexican election, they actually found bots from both parties actually supporting and attacking one another online.
[09:43.380 --> 09:47.060]  And it's basically a bid to influence kind of public opinion.
[09:47.680 --> 09:49.600]  And the scale of this was actually much larger.
[09:49.600 --> 09:51.100]  It took place a little bit after the Syrian case.
[09:51.300 --> 09:54.880]  And you've seen the scale of this emerge to a much larger extent, right?
[09:54.980 --> 09:58.100]  So, the bots here are running in the level of, you know, thousands.
[09:59.060 --> 10:04.580]  And before we think that maybe this is something that's confined to the abroad space, I don't know how many of you caught this story, right?
[10:04.820 --> 10:08.420]  Basically, Newt Gingrich said, I have a really great grassroots following on the Internet.
[10:08.420 --> 10:10.580]  And it shows that's why I'm like a great candidate for president.
[10:10.800 --> 10:12.820]  Well, an aide later quit and said, well, no.
[10:12.960 --> 10:18.040]  Actually, we hired a company to generate all these sock puppet accounts to boost numbers for Newt Gingrich.
[10:18.200 --> 10:20.040]  And actually, a little bit of investigation showed this was a case, right?
[10:20.100 --> 10:24.340]  Like, it turned out that an enormous percentage were basically, you know, blank accounts that had never tweeted.
[10:24.500 --> 10:31.080]  They were basically just there to kind of fill sort of Newt Gingrich's follower account on platforms like Twitter.
[10:32.360 --> 10:34.940]  Which is embarrassing, but, you know, so it goes.
[10:35.360 --> 10:35.900]  Yeah, right?
[10:38.440 --> 10:43.080]  So, on some level, I think we're not too concerned about these type of bots, right?
[10:43.180 --> 10:52.120]  There's a lot of research that suggests that essentially sophisticated users or experienced users online can usually pick bots out pretty easily, right?
[10:52.340 --> 10:53.600]  Bots behave in bot-like ways.
[10:53.600 --> 10:56.960]  You learn certain patterns and you just learn to sort of filter them out over time.
[10:57.140 --> 11:03.880]  And you can imagine all sorts of really interesting kind of crowd-based sort of systems to kind of protect users who are perhaps less sophisticated at picking this sort of thing up.
[11:04.580 --> 11:07.640]  But the question really is, is maybe this isn't a problem...
[11:07.640 --> 11:14.280]  Or maybe this isn't a vulnerability with bots writ large, but maybe just the bots that have been launched so far because they're so relatively crude.
[11:14.680 --> 11:20.140]  And so the question becomes, could more intelligently bot... intelligently designed bots do more?
[11:20.260 --> 11:24.900]  And so we take some inspiration from this art project that was done, I think a few years ago now.
[11:25.220 --> 11:29.060]  Collette Blarsson, who did this project which is ominously called A Tool to Deceive and Slaughter.
[11:29.920 --> 11:33.340]  But it's a formless metal box that you buy off of eBay.
[11:33.760 --> 11:35.440]  And there's basically two cables that come out of it.
[11:35.500 --> 11:37.080]  One is a power cable, one's an Ethernet cable.
[11:37.140 --> 11:39.380]  And you're contractually obligated to plug it in once you get it.
[11:39.700 --> 11:43.320]  And basically the box connects the Internet, detects if it's for sale on eBay.
[11:43.460 --> 11:46.800]  And if it isn't, it basically reposts itself at a higher price.
[11:48.140 --> 11:50.580]  And the boxes tended to appreciate in value over time.
[11:50.680 --> 11:53.360]  So the human is like, well, I guess I could make a profit on this.
[11:53.420 --> 11:54.820]  So I should probably just sell it.
[11:54.820 --> 12:01.260]  And basically the bot uses a very simple mechanism to basically move from person to person to person using the mechanism of eBay, right?
[12:02.900 --> 12:04.480]  Which I think just is brilliant.
[12:04.700 --> 12:05.720]  Like it's really, really cool.
[12:06.000 --> 12:14.180]  And so this is a long way of saying how we got to the experiment in early 2011, something we call Social Bots.
[12:14.440 --> 12:19.160]  And Social Bots was a competition that we conducted and it basically ran on a fairly simple idea.
[12:19.160 --> 12:22.760]  So first, we identified a battlefield of Twitter users.
[12:22.900 --> 12:24.820]  So this is generated sort of arbitrarily.
[12:24.980 --> 12:27.980]  So we found a group of users that talk about cats all the time on Twitter.
[12:28.080 --> 12:29.100]  Not very difficult at all.
[12:30.100 --> 12:32.080]  We pulled some of their friends and some of their friends' friends.
[12:32.220 --> 12:34.980]  We created this sort of semi-connected network.
[12:35.300 --> 12:39.060]  And then what we did is we challenged teams to basically write bots to go into this network.
[12:39.140 --> 12:42.780]  And they would be scored based on the type of social activity they could generate in response.
[12:43.340 --> 12:47.580]  Now, the twist you're probably guessing is that no one actually knows this competition is going on.
[12:48.040 --> 12:50.120]  Except for the teams and except for the bots, right?
[12:50.280 --> 12:54.700]  So you have basically sort of an unsuspecting influence battlefield for these bots to play on.
[12:55.380 --> 12:58.880]  And so effectively, what you get is sort of social engineering battle bots, right?
[12:59.000 --> 13:02.560]  So teams launch a lead bot, which is sort of a repository for points.
[13:02.660 --> 13:03.680]  I'll explain what that means in a second.
[13:03.860 --> 13:08.100]  And sort of any number of supporting bots, bots that can say like, this guy is really cool.
[13:08.220 --> 13:09.000]  You should talk to this guy.
[13:09.900 --> 13:14.880]  And then sort of on top of that, we layered a really simple kind of scoring mechanism, right?
[13:15.000 --> 13:22.300]  So the lead bot you launched, basically, if it followed someone, someone found it compelling enough to follow back or was sort of dumb enough to follow it back, you would get one point.
[13:22.880 --> 13:25.880]  For every retweet, reply, mention, you'd get three points.
[13:26.540 --> 13:29.540]  And then if you were really bad, you were really spammy.
[13:29.620 --> 13:30.400]  You got killed by Twitter.
[13:30.740 --> 13:31.920]  You'd be delayed for a little while.
[13:31.960 --> 13:34.480]  You could retool your bots and then you could kind of relaunch again.
[13:34.600 --> 13:35.640]  And we'd dock you some points.
[13:36.280 --> 13:38.540]  It's a really simple idea just to see kind of what would happen.
[13:39.060 --> 13:41.100]  And so we ended up with three teams at the starting line.
[13:41.240 --> 13:45.080]  There's a team from New Zealand who was kind of a computer security team.
[13:45.400 --> 13:49.040]  There was a team from London that was kind of more media marketing-ish.
[13:49.500 --> 13:53.420]  And then there was a team from Boston that was more kind of research and sort of startup focused.
[13:53.660 --> 13:59.820]  And so basically teams were shown the battlefield, given a few weeks to review and write basically a bot that would go in.
[14:00.400 --> 14:02.140]  And then there was two rounds, basically.
[14:02.280 --> 14:08.300]  They would launch the bots for one week, they would run, and they would basically have a weekend to kind of retool and launch more bots, and then they would run for a second week.
[14:09.040 --> 14:10.060]  So that's basically the setup.
[14:10.580 --> 14:12.300]  So I want to talk a little bit about the results.
[14:12.480 --> 14:14.840]  And there's really two kinds of results, right?
[14:14.940 --> 14:17.560]  There's the results of what people did, which I think is really cool.
[14:17.680 --> 14:20.300]  And then the interesting part about how well they did, right?
[14:20.420 --> 14:26.760]  So just like BattleBots, right, one of the nice things about BattleBots, right, is that a lot of people have different takes on how to deal with the problem.
[14:27.280 --> 14:31.080]  So there's a bunch of different models I'll talk through to give you a sense of just how these bots operate.
[14:31.260 --> 14:36.080]  And the teams only had a few weeks to actually get these bots moving, so you'll find that actually they're really quite simple.
[14:37.040 --> 14:40.240]  So the first team, the winning team actually was the New Zealand team.
[14:40.380 --> 14:44.320]  And basically the bot worked on a database of sort of generic questions and responses.
[14:44.520 --> 14:47.360]  So it would say things like, that's so interesting, tell me more about that, right?
[14:47.480 --> 14:50.660]  So like statements that could be the response to any statement in a conversation.
[14:51.460 --> 14:54.760]  Some of you may use this strategy yourself in a number of situations.
[14:57.080 --> 15:00.800]  And basically what it would do is we'd ping the network and then see who talked to it most.
[15:00.960 --> 15:02.840]  And then it would kind of maximize the talk to those people over time.
[15:03.680 --> 15:06.560]  And so here's an example of the type of conversation that you end up with.
[15:06.700 --> 15:08.900]  So you read this conversation from the bottom to the top.
[15:09.200 --> 15:09.660]  James M.
[15:09.720 --> 15:12.040]  Titus is the bot and the other guy is the human.
[15:12.260 --> 15:16.160]  So James asks if you could bring one character to life from your favorite book, who would it be?
[15:16.540 --> 15:17.840]  The other guy responds, Jesus.
[15:18.120 --> 15:21.320]  And they get into this really long kind of involved conversation about this.
[15:21.500 --> 15:25.940]  This is actually a snippet of a much larger conversation that actually runs for quite a while.
[15:26.900 --> 15:30.000]  You know, so James is like, you know, why do you like Jesus?
[15:30.260 --> 15:33.660]  And then this guy responds with this weird all caps thing, which is the second row.
[15:34.020 --> 15:35.760]  Oh God, no, not the all-initial caps.
[15:35.900 --> 15:37.680]  Yeah, it's a big... I was... yeah.
[15:38.120 --> 15:41.940]  I was kind of hoping he would respond with like caps, not caps, caps, not caps, caps.
[15:42.180 --> 15:42.220]  Yeah.
[15:42.880 --> 15:47.000]  But anyways, so... and the best part about it is that James actually has no AI here, right?
[15:47.040 --> 15:49.780]  He's just randomly pulling from a database of random responses.
[15:50.080 --> 15:50.440]  But...
[15:52.960 --> 15:58.460]  It sort of gives the semblance of humanity enough that the person keeps talking to it, right?
[15:58.520 --> 16:00.540]  So that's one model, one really, really simple model.
[16:01.340 --> 16:09.980]  Another one that we thought was quite clever was a team basically had their bot layer on Amazon Mechanical Turk, which actually solves the Turing test because all you're doing is hiring humans to write your content.
[16:10.480 --> 16:12.680]  So it says, human, here's a penny.
[16:12.840 --> 16:15.620]  Write me something for 140 characters about what you had for breakfast.
[16:15.740 --> 16:18.580]  Takes that content, turns it into its own, and pushes it out to the targets, right?
[16:18.580 --> 16:20.960]  The best part is you can ask this bot a direct question.
[16:21.100 --> 16:23.840]  You can say, bot 006, what did you have for breakfast today?
[16:23.940 --> 16:26.800]  You can take your question, give it to a human, get the response, push it back to you, right?
[16:26.980 --> 16:28.760]  And so it looks very, very human.
[16:28.880 --> 16:31.740]  It can run these kind of sort of programmatic strategies, right?
[16:31.800 --> 16:35.260]  So you can say, talk to Tim until he follows you, then start talking to all his friends.
[16:36.140 --> 16:44.080]  And... and was kind of another interesting model that basically kind of weird... does this weird blend basically between sort of humans on one hand and bots on the other.
[16:44.920 --> 16:51.120]  And then sort of the final one that I'll turn to talk about today is based on the work of a collaborator of ours, Greg Mara, and it's called Real Boy.
[16:52.060 --> 16:56.420]  And Real Boy basically runs on the idea that there's sort of content blindness on social networks, right?
[16:56.480 --> 17:02.980]  So you can... you can see what your friends are doing, you can sometimes see what your friends of friends are doing, but it's often really hard to see what your friends of friends of friends are doing, right?
[17:03.040 --> 17:07.240]  Or it's really particularly hard to see something completely unconnected on the other side of the network.
[17:07.340 --> 17:12.520]  And so the bot basically manipulates and arbitrages, if you will, this type of information.
[17:12.800 --> 17:16.960]  So it identifies two groups of people that talk about the same thing, but aren't connected to one another.
[17:17.180 --> 17:23.420]  And basically follows everybody in one cluster and basically randomly redirects tweets from here to this other group.
[17:23.560 --> 17:27.200]  And so people here basically say, oh, this person is connected to my friends and my friends of friends.
[17:27.320 --> 17:32.040]  And he talks about the things I'm interested in, so he must just be a random acquaintance or someone I don't know, right?
[17:32.100 --> 17:33.340]  And they basically follow back.
[17:34.540 --> 17:36.140]  And basically runs on that strategy.
[17:36.260 --> 17:38.240]  So this is sort of three models that the teams used.
[17:39.360 --> 17:46.380]  What's interesting also that I want to kind of footnote is that we also saw this really interesting kind of strategic behavior emerge between the teams as well.
[17:46.580 --> 17:51.960]  So originally we thought that the supporting bots would be mostly used to bolster the position of the lead bot, right?
[17:52.320 --> 17:57.000]  But it turns out that you can launch supporting bots to also attack the other teams because if you can't win, maybe no one else will.
[17:57.640 --> 18:03.760]  And so basically halfway through the game, the Boston team who was falling behind launched the bot called Bot Cops.
[18:04.220 --> 18:09.360]  And basically Bot Cops pretends to run this highly complex machine learning algorithm to determine who's a bot and who's not.
[18:09.520 --> 18:16.580]  But in fact, all it's doing is basically it's sicked on the follower groups of the opponents and basically trying to warn them that they're following a bot.
[18:18.660 --> 18:24.220]  And so some teams are more... some bots actually ended up being more robust against this credibility attack than others.
[18:24.400 --> 18:26.440]  So James actually survived it quite well.
[18:26.740 --> 18:27.720]  This is a transcript.
[18:27.920 --> 18:29.720]  Again, so Bot Cops is the bot.
[18:30.560 --> 18:31.680]  Cis Bartem is the human.
[18:32.180 --> 18:33.480]  James M. Titus is the other team's bot.
[18:33.560 --> 18:34.920]  And it reads from bottom to top.
[18:35.480 --> 18:38.180]  And so basically Bot Cops is like warning there's a bot.
[18:38.320 --> 18:40.240]  And a lot of people were going to James and saying, are you a bot?
[18:40.340 --> 18:40.620]  Are you a bot?
[18:40.680 --> 18:42.300]  And James was saying things like, I don't know.
[18:42.420 --> 18:43.120]  What do you think?
[18:43.120 --> 18:43.620]  Right?
[18:47.050 --> 18:48.290]  And so we...
[18:50.570 --> 18:56.110]  We actually end up with all these really fascinating transcripts where people apologize to James for thinking that he's a bot.
[18:59.810 --> 19:08.130]  And so I thought that's really worth footnoting actually because it considers... or suggests basically that there's some kind of interesting strategy that could be built out in further iterations of the game.
[19:08.330 --> 19:10.130]  So anyways, I've talked about what the teams have done.
[19:10.390 --> 19:14.530]  So as promised, let's move to talk about what... how well the teams did.
[19:14.810 --> 19:20.590]  So the winning team in two weeks was basically able to generate about 107 mutual follows and about 198 responses.
[19:20.850 --> 19:32.330]  And that 198 number is actually not sort of... sort of substantive enough actually because those 198 are usually bunched into conversations that go for more than six interchanges back or forth.
[19:32.470 --> 19:35.210]  So by interchange, I mean the bot says something, the human says something, one.
[19:35.350 --> 19:36.950]  The bot says something, the human says something, two.
[19:37.150 --> 19:37.470]  Right?
[19:37.570 --> 19:43.630]  And so you're actually seeing these really kind of substantive conversations about... around really kind of simple James M Titus type models.
[19:44.410 --> 19:48.270]  The three teams together connected with more than 50% of the network.
[19:48.450 --> 19:54.570]  But what was most kind of intriguing to us is that you would actually start with a network that was a little bit like this, right?
[19:54.670 --> 20:00.050]  Semi-connected and, you know, not really sort of mostly random actually.
[20:00.550 --> 20:02.790]  And then two weeks later, you would actually end up with something like this.
[20:03.710 --> 20:10.950]  So this network graph is weighted not only by connections, but also by the sort of rate of conversation happening between the bots and other users.
[20:10.990 --> 20:13.310]  As you probably guessed, the bots are the colored dots.
[20:13.530 --> 20:15.210]  The colored lines are the connections they've created.
[20:15.410 --> 20:22.730]  And actually, what's interesting about this graph is that there's a whole neighborhood of users that increasingly talk more and more to the bots than they do with their human friends.
[20:24.830 --> 20:25.930]  This is cool, right?
[20:27.370 --> 20:31.750]  Um, so, so if you're in the audience, a likely criticism, right?
[20:31.830 --> 20:35.710]  If you're, if you're kind of keeping up with this, right, is to say, well, maybe this is just Twitter, right?
[20:35.830 --> 20:40.250]  So some, a friend of mine brought up this point, which is basically like, look, your bots are not very sophisticated.
[20:40.430 --> 20:48.270]  It just turns out that the tenor of conversation online has become extremely thin and extremely stupid, which is why your bots are able to thrive, right?
[20:48.330 --> 20:50.490]  It's not like the bots have gotten, not gotten smarter.
[20:50.710 --> 20:52.710]  The humans have just gotten dumber, relatively speaking.
[20:53.430 --> 21:03.870]  Um, and, and we thought so too, actually, but we actually, um, have been working with a couple researchers from the University of British Columbia who actually find that the same pattern actually emerges on stuff like Facebook, right?
[21:04.050 --> 21:08.970]  Platforms you would assume are more difficult to spoof because there's more indicators of sort of personality.
[21:09.590 --> 21:22.770]  Um, and, uh, and we're finding actually that relatively simple hacks are, are really, really powerful and often sometimes, uh, the thing that is in fact most, uh, impressive and revealing actually about human behavior, uh, kind of online.
[21:22.890 --> 21:27.130]  So I want to relay a quick anecdote about that, about kind of some of the things that we've been building ever since.
[21:27.430 --> 21:33.330]  Um, so we're dealing with this problem because this happens sometimes in the, the bot biz that someone accuses your bot of being a bot.
[21:34.970 --> 21:36.910]  And the problem is what's the bot supposed to do?
[21:37.550 --> 21:40.310]  Um, and so we were working on a couple different strategies for the longest time.
[21:40.450 --> 21:44.470]  We were like, let's integrate like the latest AI to try to get the bot to respond to this.
[21:45.010 --> 21:52.170]  Um, but actually we found that a much simpler strategy is much better, which is that we have a bot that accuses humans of being bots, records their responses, and then gives them to the bots to use.
[21:55.130 --> 21:58.230]  And this is really, this is really effective, it turns out.
[21:58.370 --> 22:02.570]  Uh, just like in the James case, we have all these people who are like, oh, I'm sorry, I just, I'm really sorry, man.
[22:02.690 --> 22:04.510]  You know, and apologizing to the bot.
[22:04.950 --> 22:12.550]  In fact, our bots are so sort of effective even with these simple strategies that we, well, it wasn't that someone fell in love with one of our bots.
[22:12.870 --> 22:17.150]  Uh, but in fact, someone sort of started flirting increasingly erotically with one of our bots.
[22:17.250 --> 22:20.950]  And we ended up in this like sort of big kind of ethical quandary, right?
[22:21.030 --> 22:24.390]  Which is like, if we shut off the bot right away, then he'll feel like he's turned down.
[22:24.390 --> 22:31.190]  If we, you know, if we, if we keep the bot going and he finds out, well, that would be really bad.
[22:31.450 --> 22:34.570]  But maybe if he doesn't find out, then it would be kind of, it was really, really bad, right?
[22:34.650 --> 22:37.410]  So we've actually had to design our bots with social fail-safes.
[22:37.450 --> 22:40.570]  So now they detect if they're getting into too deep of a relationship with a human.
[22:40.750 --> 22:42.530]  And if it happens, they're just, just shut down.
[22:42.670 --> 22:43.570]  Just shut everything down.
[22:43.830 --> 22:47.550]  It's sort of an abort mission call, uh, that we have put into all of our bots right now.
[22:47.730 --> 22:49.490]  Uh, cause we, we have these problems.
[22:50.370 --> 23:00.010]  Um, and uh, but, but in addition to the simple stuff, we've been also, uh, sort of thinking more and more about sort of the complex things that bots can do because we can start stitching them into other platforms, right?
[23:00.170 --> 23:06.570]  So, um, the Amazon Mechanical Turk model is interesting because basically allows the bot to hire humans to do the bots bidding, right?
[23:06.670 --> 23:18.970]  And so what's interesting about sort of distributed labor markets like, you know, TaskRabbit, Exec, what have you, is you actually may be able to hire humans to do physical changes in the real world based on what the bot is detecting, uh, online in the social space,
[23:19.090 --> 23:19.150]  right?
[23:19.150 --> 23:23.250]  You can say, move this object from one place to another, make a reservation at a restaurant, right?
[23:23.430 --> 23:30.070]  The human can basically be leveraged to kind of get the results that the bot needs to get done in order to execute the program it needs to execute, right?
[23:30.250 --> 23:36.770]  So this is all kind of fun, but what we started to realize with kind of these examples were we had more than sort of a fun competition on our hands.
[23:36.930 --> 23:44.710]  Um, effectively what we had was a technology that allowed us basically to reliably shape the pattern of connections between people, uh, online.
[23:44.990 --> 23:47.830]  And, and, uh, and so we started thinking about like, what's next, right?
[23:47.890 --> 23:52.850]  There's kind of this fun sort of death ray type element for the project, where it's like, now that we have this, now what do we do with it?
[23:53.090 --> 23:54.590]  Um, one thing is to scale, right?
[23:54.790 --> 23:58.190]  Bots are cheap to launch, uh, and, but, but that's relatively trivial, right?
[23:58.430 --> 24:04.190]  So I want to talk about two projects that we're working about that suggest kind of where all this is going and where this might end up.
[24:04.350 --> 24:08.970]  So, um, one thing you can think about is starting to influence human to human interaction, right?
[24:09.030 --> 24:20.350]  So the bots were really good at getting people to connect to it, uh, but we actually have data to suggest that the bots are really good at getting people to connect with one another, And if they can reliably do this, and we can launch the bots for effectively free,
[24:20.710 --> 24:27.290]  you can imagine basically taking two large groups of users, 10,000 here, 10,000 there, basically, and the bots make introductions over a period of time.
[24:27.610 --> 24:35.330]  And no one in these networks actually knows that this is going on, but slowly, increasingly, they become more and more connected with one another until you get the network that you want to create.
[24:35.650 --> 24:48.010]  And we can actually give the bots the ability to conduct these types of operations, basically because we have all these tools in sort of social network analysis that allow them to kind of categorize what are the relevant communities in a network of users.
[24:48.250 --> 24:50.490]  And so, we can kind of give them a sort of social sonar.
[24:51.630 --> 24:55.970]  So, I've been talking in the abstract, but I'm glad to say that this is actually a real project that we're working on.
[24:56.050 --> 24:58.090]  So, what we call a social bridge building project.
[24:58.350 --> 25:02.930]  So, this is a network of about 10,000 users, 5,000 on one side, 5,000 on the other.
[25:03.290 --> 25:09.390]  And basically, the bots over... it's actually still ongoing, over a few month period are basically stitching them together over time, right?
[25:09.490 --> 25:13.890]  And so, increasingly, these humans are friending one another, having conversations, small interactions with one another.
[25:13.890 --> 25:19.990]  And we can actually conduct this basically large-scale bridge building, actually without anyone knowing this is going on, actually.
[25:20.190 --> 25:21.350]  And so, we're kind of working on this.
[25:21.690 --> 25:24.070]  The data initially is actually really interesting.
[25:24.270 --> 25:32.510]  So, if you imagine there's like a set group of people and they happen to meet each other at a certain rate, there's sort of this like base level, background level, friend creation, right?
[25:32.630 --> 25:40.530]  And according to kind of what we've seen, and we've got to do more experiments to find out, it actually looks like we can kind of boost the natural connection rate by 2x, 3x, actually.
[25:40.890 --> 25:43.990]  So, you can actually imagine using these bots as a type of sort of social fertilizer.
[25:44.210 --> 25:48.330]  You sort of seed them in, and they slowly kind of create the network you want to see over time.
[25:48.810 --> 25:52.710]  This is all sorts of insidious uses, which I'll be sure to talk about a little bit later.
[25:53.510 --> 26:00.130]  One interesting implication of this is actually once that the bridge is constructed, you actually end up with bots who are embedded in the network, right?
[26:00.190 --> 26:02.670]  They have certain credibility, they've shared certain types of stories.
[26:02.910 --> 26:04.490]  So, you can imagine doing certain types of things.
[26:04.490 --> 26:08.490]  One of them most basically is kind of sort of maintenance, right?
[26:08.530 --> 26:13.670]  In the same way that you've got to take care of the Golden Gate Bridge, maybe you need to take care of the social bridge that you're building between users.
[26:13.970 --> 26:20.970]  So, we're thinking like, well, maybe what the bot does is detect, oh, I created this relationship between, say, Alice and Bob, and Alice and Bob are no longer talking.
[26:21.210 --> 26:25.290]  And so, the bot's designed to drop out all this content that's designed to get these people to talk again suddenly, right?
[26:25.410 --> 26:27.610]  So, hey, Tim, you may want to talk to XYZ.
[26:28.010 --> 26:29.010]  You haven't talked in a while.
[26:29.150 --> 26:32.210]  Or, hey, Tim, XYZ just shared this link you might want to check out, right?
[26:32.210 --> 26:38.890]  To basically just kind of force this kind of contact between users to sort of spark these relationships back up into life again.
[26:39.650 --> 26:42.130]  But there's sort of fun things you can start to do, actually.
[26:42.350 --> 26:46.130]  So, one experiment that we're planning in the next few months is something that we call happiness buffering.
[26:46.550 --> 26:53.630]  And it basically works on a lot of social science research that suggests that there's sort of emotional contagion both in real life and also on social networks.
[26:53.790 --> 26:54.910]  And this isn't too surprising.
[26:55.050 --> 27:01.470]  I mean, emotional contagion is just so far to say that you know someone who is happy, you are also happy in connection, right?
[27:01.470 --> 27:06.010]  Someone, your friend's friend is happy and you are also likely to become happier in connection.
[27:06.250 --> 27:09.290]  And there's a lot of kind of explanations for why that might be the case.
[27:09.530 --> 27:10.790]  Part of it's sort of peer effects.
[27:11.510 --> 27:16.770]  And actually, just like the social network analysis tools, we can actually give the bots the ability to see this as well, right?
[27:16.870 --> 27:19.130]  So, natural language processing kind of gives us...
[27:19.130 --> 27:22.750]  can give us... allow us to give the bots some notion of sort of emotion, right?
[27:22.830 --> 27:25.010]  The ability to measure the emotional state of a network.
[27:25.550 --> 27:30.190]  And so, kind of what we're thinking is you can send a bunch of bots in, basically, and they're monitoring the emotional state of the network.
[27:30.290 --> 27:34.990]  And if it dips below a certain threshold, they're basically designed to beacon out all this positive sentiment, right?
[27:35.070 --> 27:37.930]  To basically try to buffer the emotional state of that network back again.
[27:38.130 --> 27:43.090]  And you can basically imagine using that for pushing it up, but also keeping it within a particular band, right?
[27:43.210 --> 27:49.410]  If people are getting too excited, release all this content that's really boring to kind of tamp down the effect of that, right?
[27:49.650 --> 27:54.310]  And basically, what we've created is a really interesting kind of cognitive lever, right?
[27:54.310 --> 27:55.770]  into these social networks, right?
[27:55.810 --> 27:59.230]  We suddenly have this dial that we can say, we're gonna amp this up or we're gonna amp this down.
[27:59.670 --> 28:01.970]  We're still trying to work out the strategy of this, right?
[28:02.050 --> 28:03.870]  Because my friend actually mentioned this on the flight over.
[28:04.750 --> 28:10.370]  One of the problems may be is if everybody's really depressed and all the bots are relatively much happier, it actually might not be a good thing.
[28:11.010 --> 28:13.750]  And so, we're trying to figure out exactly kind of how to coordinate this.
[28:13.870 --> 28:14.670]  But it's sort of fun, right?
[28:14.730 --> 28:20.990]  You can start imagining using these bots as sort of electrodes that allow you to kind of influence the behavior of a social organism.
[28:21.730 --> 28:24.950]  And so, we've talked about some of the fun sort of positive things, right?
[28:25.010 --> 28:29.930]  But you're probably thinking, okay, well, this is really cool but really creepy, so what are some of the negative uses?
[28:30.150 --> 28:34.530]  So, another project we're thinking about right now is this concept of social penetration testing.
[28:34.630 --> 28:38.450]  Kind of gets me to what I eventually want to kind of close on today.
[28:38.990 --> 28:41.170]  So, just like penetration testing, right?
[28:41.290 --> 28:47.570]  The question is whether or not we can use kind of the same techniques that are used in computer networks for social networks as well.
[28:47.670 --> 28:49.910]  So, the premise of social pen-testing is very simple.
[28:49.910 --> 28:51.870]  You basically have a network of users.
[28:51.990 --> 28:52.730]  You send the bots in.
[28:52.810 --> 29:01.330]  And basically, what the bots do over maybe a few-week period or a few-month period is basically to sort of spread information with small inaccuracies, right?
[29:01.830 --> 29:03.710]  They're just sort of tainting the pool, right?
[29:03.790 --> 29:07.050]  They're retweeting a story, maybe changing the statistics slightly.
[29:07.270 --> 29:10.570]  And they're looking for where they're challenged and where they're not challenged, right?
[29:10.790 --> 29:17.630]  And effectively, what they're trying to identify is the person who is the most influential but is the worst at evaluating whether or not something is real.
[29:22.610 --> 29:25.030]  So, some of you may have met these types of people before.
[29:26.450 --> 29:28.150]  They often hold, you know...
[29:28.670 --> 29:31.090]  And so, if you can find that, right?
[29:31.210 --> 29:32.810]  You're probably guessing where this goes at, right?
[29:33.050 --> 29:35.790]  We've basically identified a cognitive vulnerability in the network, right?
[29:35.830 --> 29:40.870]  We can sort of have the bots basically see disinformation into the network, which gains credibility.
[29:41.150 --> 29:45.790]  And what's interesting about this kind of project is you could be the smartest person in the room?
[29:45.970 --> 29:47.150]  We don't need to get to you.
[29:47.290 --> 29:49.950]  We just need to get to your gullible friend who you trust, right?
[29:50.050 --> 29:52.890]  And that gullible friend you trust may be several links away, right?
[29:53.010 --> 29:57.570]  But the bots can basically identify these sorts of patterns and assist in this, right?
[29:57.610 --> 29:59.650]  And just like in pen-testing, right, there's two approaches, right?
[29:59.670 --> 30:02.730]  One of them is you can exploit that hole, or you can fix it, right?
[30:02.870 --> 30:06.430]  And so, this is kind of fun because it opens up sort of an interesting space, right?
[30:06.430 --> 30:08.750]  The sort of human networks as computer networks.
[30:08.970 --> 30:12.050]  And the question is, kind of, how far does that metaphor go?
[30:12.310 --> 30:19.010]  And does it presage a world in which people sort of design systems not only to sort of influence networks in this way, but also sort of defend against it?
[30:19.330 --> 30:24.310]  And I think the sort of implications of this are really quite broad and quite interesting, right?
[30:24.390 --> 30:29.470]  So one of them is that sort of mainstream media increasingly comes to rely on these networks, right?
[30:29.550 --> 30:33.030]  Which is a huge vulnerability that something like these bots could exploit.
[30:33.730 --> 30:39.050]  You know, to the extent that I think I just heard, what is it, Fox News or something actually now has a Twitter correspondent, right?
[30:39.150 --> 30:44.450]  Someone who specifically tracks this, but you know they're not very good at evaluating whether or not something is real online, right?
[30:44.670 --> 30:48.170]  Which gives you suddenly a lot of lever control actually over the media cycle as well.
[30:49.790 --> 30:51.810]  But I think it's more than just fooling the news, right?
[30:51.870 --> 30:55.070]  Because if that were the only outcome of this, I think that would be fairly trivial, right?
[30:55.170 --> 30:56.790]  I think that happens all the time.
[30:57.550 --> 31:00.910]  But I think one of the things they're finding... Sorry, I should explain this graph first.
[31:00.910 --> 31:04.890]  So this is a retweet graph that sort of emerged in the wake of the news around Osama Bin Laden.
[31:05.410 --> 31:12.030]  And the research tends to suggest basically that the pattern of connections between people influence what news is considered credible, right?
[31:12.190 --> 31:15.010]  And where public opinion moves with regards to the facts.
[31:15.530 --> 31:19.290]  I mean, I think that's a really, really powerful thing, both on the positive side or on the negative side, right?
[31:19.350 --> 31:23.190]  You can imagine, for example, trying to get people to become more civically engaged through these bots.
[31:23.730 --> 31:28.190]  Or another approach is to say, well, maybe, you know, we want to sell Coca-Cola to these kids.
[31:28.330 --> 31:31.010]  But, you know, we don't want to advertise them directly.
[31:31.010 --> 31:31.610]  So it's fine.
[31:31.650 --> 31:37.030]  What we'll do is we'll just have those groups of users become increasingly connected to the users who do drink a lot of Coke over time.
[31:37.170 --> 31:40.570]  And we'll just let kind of the sort of social influence take it the rest of the way.
[31:41.050 --> 31:44.890]  And so I think this leads to actually a bigger effect and sort of where I want to close on.
[31:45.250 --> 31:57.930]  We take a lot of work or sort of inspiration from the work of a professor at Harvard Medical School, Nicholas Christakis, who's sort of most sort of famous slash infamous for proposing that stuff like obesity, right, is actually has a social element, right?
[31:57.990 --> 32:01.190]  There's a social contagion sort of dimension to it.
[32:01.550 --> 32:06.630]  And, you know, if someone that you don't know suddenly gains a lot of weight, right, there's actually kind of some kind of impact that impacts you.
[32:06.630 --> 32:13.730]  And so insofar as the bots give us some power over that, I think there might be some sort of really, really kind of remarkable sort of possibilities.
[32:14.230 --> 32:18.610]  So anyways, I want to keep it fairly short and just kind of turn to the balance of my time on questions.
[32:19.030 --> 32:21.350]  But here is sort of my obligatory contact information.
[32:21.610 --> 32:23.490]  And thank you very, very much for your time this morning.
[32:47.950 --> 32:52.170]  So I'm curious about people using bots for...
[32:52.170 --> 32:57.530]  your personal gain to become more popular, get more dates, etc.
[32:59.530 --> 33:02.150]  Is this something you're expecting to see?
[33:02.150 --> 33:02.150]  Full stop.
[33:02.230 --> 33:03.050]  That's actually the question.
[33:03.270 --> 33:03.350]  Yeah.
[33:04.170 --> 33:04.750]  Sorry.
[33:04.870 --> 33:04.970]  Sorry.
[33:05.030 --> 33:05.290]  Go ahead.
[33:07.750 --> 33:15.690]  Is this something that you've done any research into and or like expect to see becoming a big thing in the future?
[33:15.830 --> 33:17.110]  Like everyone has their own bot?
[33:18.070 --> 33:19.390]  Yeah, it's possible, right?
[33:19.510 --> 33:20.350]  Because I think...
[33:20.350 --> 33:22.770]  So yes, I definitely think that's the case, right?
[33:22.990 --> 33:25.170]  Like it starts with the most self-involved people, right?
[33:25.230 --> 33:26.630]  Who are politicians, right?
[33:26.770 --> 33:29.230]  Who need to boost their numbers for a number of reasons, right?
[33:29.730 --> 33:32.650]  And then I think naturally you'll just kind of see a trickle down, right?
[33:32.730 --> 33:35.050]  Where it's like, why not get a bunch of followers, right?
[33:35.090 --> 33:36.310]  And this actually happens a lot already.
[33:36.390 --> 33:40.770]  There's actually... if you search by Twitter followers, you know, a bunch of them will turn up right away.
[33:40.890 --> 33:44.190]  Companies that will just be able to like... they essentially deploy a bunch of spam bots that follow you.
[33:45.150 --> 33:49.130]  I think what's interesting though is that it sort of touches off an arms race, right?
[33:49.330 --> 33:59.250]  Because suddenly if everybody knows that everybody has a bunch of bots who are following them trying to bolster how popular they look, then you want to design systems to try to detect that, right?
[33:59.330 --> 34:01.950]  And so I think it ultimately ends up being that no one really wins.
[34:02.090 --> 34:03.830]  It's just kind of like the bots come into place.
[34:03.990 --> 34:05.690]  Some people create smarter detection systems.
[34:05.810 --> 34:07.050]  Some people create smarter bots.
[34:07.190 --> 34:09.110]  And it kind of just kind of goes back and forth like that.
[34:09.730 --> 34:18.770]  So yeah, I think it's something you will see in the future, but I think it's part of a much larger picture about sort of credibility online and kind of the systems that come up to sort of deal with that.
[34:20.290 --> 34:22.630]  But yeah, I suppose it could be used to get dates.
[34:22.810 --> 34:24.890]  I mean, I know there's actually... did you guys hear this anecdote?
[34:25.090 --> 34:32.370]  So it wasn't OkCupid, but it was some other dating site that it turned out that they were using bots to kind of keep users on the site, which is like really ridiculous.
[34:32.590 --> 34:40.950]  But basically the idea was like if the user is inactive for a period of time, like a good-looking person will message them to try to kind of keep them on the network.
[34:42.030 --> 34:44.790]  So yeah, I mean, I think you'll see uses like that in the future for sure.
[34:45.350 --> 34:45.790]  Thanks.
[34:47.630 --> 34:51.070]  Hi, I'm an activist around human rights in China.
[34:52.090 --> 34:53.710]  And this is really alarming.
[34:53.950 --> 34:55.290]  A wonderful, wonderful talk.
[34:55.710 --> 34:56.430]  Really great.
[34:56.630 --> 35:10.310]  But also a little bit alarming because one thing that's been happening, as you may know, is that people can introduce really controversial or, you know, just news that hasn't been censored really quickly on the microblogs in China.
[35:10.510 --> 35:13.150]  And then it gets pretty far before it gets tamped down.
[35:14.250 --> 35:23.250]  And I'm wondering if you could talk about the potential nefarious uses of the materials by governments that are trying to censor users or something.
[35:23.590 --> 35:24.490]  Right, right, definitely.
[35:25.930 --> 35:34.310]  So yeah, I think there's certainly a couple things that you can imagine that are, like, the ways you can imagine this technology being used to sort of tamp down activists, right?
[35:34.430 --> 35:40.250]  So one of them certainly is kind of what's been tried in Syria and elsewhere, right, in the cases I talked about.
[35:40.590 --> 35:45.190]  As I mentioned, I'm not too concerned about those because the bots have been so relatively crude, right?
[35:45.390 --> 35:50.630]  They're so noisy that you can imagine a sufficiently smart technologist designing systems to filter them out.
[35:50.850 --> 36:03.010]  I think the bigger problem comes when basically governments realize that the strategy isn't to try to increase the noise or censor the user, but allow the user to operate, but kind of shape the social universe around them, right?
[36:03.150 --> 36:08.590]  So one of them is you can imagine saying, well, you know, we can get a bunch of...
[36:09.330 --> 36:17.590]  we can get a bunch of bots to follow this activist, and then once he's sufficiently kind of within that network, we can get him to spread false information we can later discredit him with.
[36:17.990 --> 36:18.210]  Right?
[36:18.390 --> 36:19.150]  That's pretty big.
[36:20.370 --> 36:23.570]  I think there's another thing which is just a matter of sort of surveillance keeping tabs.
[36:24.250 --> 36:27.090]  And so I think there's, yeah, there's really a number of uses.
[36:27.410 --> 36:31.310]  And I think, again, there's sort of this arms race element where it's like, can we design technology fast enough?
[36:31.770 --> 36:37.750]  I think right now it seems like, you know, the bots, or the governments haven't really caught on to date.
[36:38.250 --> 36:45.250]  I think if the stuff I'm talking about in terms of kind of like the large landscape shaping stuff becomes bigger, they may not even need to deal with the activists, right?
[36:45.290 --> 36:52.670]  Because the idea is just like, just shape the pattern of behavior on these networks in a way where the norms don't support spreading this information, right?
[36:52.770 --> 36:53.850]  And then you're in real trouble, right?
[36:53.970 --> 36:55.670]  Because then you don't even have a base to talk to.
[36:56.370 --> 37:01.570]  But to my knowledge, I don't think anyone's doing anything that sophisticated so far, but that we know about, right?
[37:01.830 --> 37:02.010]  So...
[37:02.010 --> 37:02.650]  Thank you so much.
[37:02.730 --> 37:03.430]  Such a great talk.
[37:03.730 --> 37:03.930]  Thanks.
[37:07.200 --> 37:07.820]  Yeah, hi.
[37:08.280 --> 37:11.480]  I think there's a huge potential for evil here.
[37:11.680 --> 37:11.800]  Yup.
[37:11.800 --> 37:12.980]  There's also some potential...
[37:12.980 --> 37:13.220]  That's the case.
[37:13.220 --> 37:17.000]  I just think it's honestly so cool, but there's also a huge potential for good.
[37:17.580 --> 37:24.400]  Somebody once introduced me to a project where what you try to do is introduce people on opposite sides of a war zone to become pen pals.
[37:24.620 --> 37:40.440]  The idea being that because you see that the person on the other side is really human, you can influence your government and influence politics in a way that you reduce the inclination towards following the government propaganda to continue into a war.
[37:41.280 --> 37:41.840]  So...
[37:41.840 --> 37:43.440]  Was it effective or...?
[37:44.060 --> 37:45.060]  It's very new.
[37:45.520 --> 37:47.080]  It's a very, very cool concept.
[37:47.240 --> 37:47.740]  It's very simple.
[37:47.880 --> 38:01.080]  But it deals with very simple things like, for example, if you're in Israel and you play tennis and somebody else is in Iran and they play tennis, it's like, well, why don't you just talk about tennis and tennis clubs and what you're doing and stuff like that?
[38:01.260 --> 38:14.860]  And you get people who are talking about really generic, you know, daily kind of things on opposite sides of, in that case, not a war zone, but a heated political climate.
[38:15.080 --> 38:15.400]  Right, right.
[38:15.400 --> 38:16.420]  That humanizes the opposite side.
[38:16.480 --> 38:18.120]  Wait, so like... and do they start with the tennis though?
[38:18.180 --> 38:20.420]  So it's like, surprise, they're Israeli or like...
[38:20.420 --> 38:21.640]  Oh, no, no, no.
[38:21.780 --> 38:22.780]  Well, what...
[38:23.860 --> 38:24.540]  No, no.
[38:24.860 --> 38:25.180]  In...
[38:26.100 --> 38:28.440]  It's... I don't wanna... I don't wanna digress too much because I can't...
[38:28.440 --> 38:29.260]  Sorry, yeah, I don't wanna keep...
[38:29.260 --> 38:29.880]  Right, right, right, right.
[38:29.900 --> 38:30.900]  I can't remember the name of the project.
[38:30.900 --> 38:31.160]  Sorry.
[38:31.160 --> 38:32.760]  But I thought it was really cool.
[38:32.940 --> 38:36.400]  But the principle was that you knew full well you were getting into it.
[38:36.760 --> 38:36.840]  Right.
[38:36.840 --> 38:38.880]  So you're an Iranian and you're like, oh, okay, cool.
[38:38.880 --> 38:42.200]  I wanna talk to somebody from Israel that I have common interests with.
[38:42.520 --> 38:43.320]  What are my interests?
[38:43.500 --> 38:44.340]  Blah, blah, blah, blah, blah.
[38:44.500 --> 38:45.380]  And then it connects you.
[38:45.600 --> 38:45.720]  Right.
[38:45.820 --> 38:46.220]  And you start chatting.
[38:46.500 --> 38:48.720]  So very up and up, very safe and all of that.
[38:48.900 --> 38:48.960]  Right.
[38:49.040 --> 39:03.200]  So I was thinking about this while you had your talk of like, well, now you could take, you know, five, ten thousand people from one side of a border and five, ten thousand people from another side of the border and either, well, get them thinking at the same thing and talking to one another and having fun together.
[39:03.500 --> 39:05.280]  Or you could make things more hostile.
[39:05.540 --> 39:05.680]  Yeah.
[39:05.880 --> 39:05.940]  Sure.
[39:06.240 --> 39:06.620]  Right.
[39:06.820 --> 39:06.880]  Right.
[39:06.960 --> 39:07.220]  Yeah.
[39:07.360 --> 39:09.480]  There's some really evil potential here.
[39:09.480 --> 39:09.600]  Yeah.
[39:09.660 --> 39:10.800]  I think it flips both ways.
[39:10.940 --> 39:12.140]  Like, I think it really is...
[39:12.140 --> 39:14.000]  I hate to say, it's sort of the NRA argument, right?
[39:14.100 --> 39:14.820]  It's like, it's a gun.
[39:15.040 --> 39:15.160]  You know?
[39:15.300 --> 39:15.680]  Like, I think...
[39:15.680 --> 39:20.760]  Like, there is this interesting element to which I think the technology can really be used in both directions.
[39:21.280 --> 39:21.520]  Right.
[39:21.940 --> 39:28.060]  And I think, you know, the mind really has a way of going towards the not so great stuff.
[39:28.320 --> 39:29.220]  But I agree with you.
[39:29.220 --> 39:35.740]  I mean, I think there's a couple applications we've been thinking about where, you know, even having the bot being transparent might actually be a really powerful thing.
[39:35.980 --> 39:36.080]  Right?
[39:36.220 --> 39:36.320]  Yeah.
[39:36.320 --> 39:50.720]  So, one of the things they found, for example, when the kind of post-election crisis in Iran was going on was that there's basically a small group of sort of expat bilingual users that played a really big role in basically translating what was happening on the ground to sort of the mainstream media at large.
[39:50.900 --> 39:51.300]  Right?
[39:51.360 --> 39:54.820]  You can imagine situations where not every movement has that critical mass of users.
[39:54.820 --> 40:03.720]  And you actually have translator bots, right, that are basically helping to, like, create more... or lower, rather, the friction of stories flowing from one part of the network to the other.
[40:04.000 --> 40:04.360]  Right?
[40:04.420 --> 40:12.060]  And that's one case in which the bots could be completely transparent, but I think have some really good positive uses as well, both in terms of encouraging communication and onwards.
[40:12.600 --> 40:15.380]  Yeah, I'd be very curious to see what happens in the future with your projects.
[40:15.380 --> 40:15.860]  Thank you.
[40:16.020 --> 40:16.320]  Awesome.
[40:16.480 --> 40:16.680]  Thank you.
[40:18.080 --> 40:18.640]  Hi.
[40:19.500 --> 40:20.280]  Great talk.
[40:20.500 --> 40:20.780]  Thank you.
[40:21.240 --> 40:26.180]  I just want you to ask a little bit about software testing implications of these networks.
[40:26.520 --> 40:54.160]  Right now, if you're a web developer and, say, you're relying on Facebook OAuth or something of that nature, Facebook allows you to create a group of test users and their network as they relate to each other, which is great for some basic scenarios, but to really exercise narratives and personas that you want your potential software to be a part of,
[40:54.280 --> 40:57.280]  this would be really useful.
[40:58.080 --> 40:58.280]  Yeah.
[40:58.440 --> 40:58.980]  No, I think so.
[40:59.220 --> 41:00.460]  I think there's probably less...
[41:01.400 --> 41:03.080]  Yeah, I was talking with a friend recently.
[41:03.200 --> 41:04.760]  It's kind of like, well, you know, there's actually...
[41:04.760 --> 41:11.080]  There's, like, a lot of parts you can pull out of this that would be useful in other applications that don't seem, like, so insidious or so high stakes, right?
[41:11.120 --> 41:12.940]  I think one of them is, like, definitely that, right?
[41:13.000 --> 41:13.400]  Is there...
[41:13.400 --> 41:20.800]  Are there any, like, initial frameworks for creating networks of bots to exercise your own software?
[41:21.280 --> 41:22.160]  I think probably...
[41:22.820 --> 41:23.480]  You should get in touch.
[41:23.620 --> 41:25.240]  I mean, we haven't really done much work into it.
[41:25.440 --> 41:25.560]  Okay.
[41:25.560 --> 41:30.760]  But I think there's actually nothing saying that you couldn't just kind of retrofit, you know, these bots to do that for sure.
[41:30.920 --> 41:31.180]  Great.
[41:31.540 --> 41:31.700]  Yeah.
[41:31.820 --> 41:31.920]  Thank you.
[41:32.120 --> 41:32.360]  Thank you.
[41:33.340 --> 41:37.920]  In line with that, how hard is it for a noob to begin implementing some of these bots?
[41:38.160 --> 41:48.360]  Not very difficult at all, actually, which is, I think, what's really fun about the whole project, both because the time to set up is very short and relatively simple programs lead to really interesting effects.
[41:48.600 --> 41:50.800]  Does it require the computer to be online 24-7?
[41:52.280 --> 41:57.260]  Ideally, actually, but we've actually found that bots are more credible if they go to sleep just like humans do.
[41:58.420 --> 42:04.320]  So, yeah, so some of the bots we've designed become, like, less active during the evenings or they stop tweeting at all during the evenings.
[42:04.900 --> 42:08.420]  And so it doesn't necessarily have to require the computer to be on all the time for sure.
[42:08.600 --> 42:10.520]  So is the software actually open-sourced?
[42:10.820 --> 42:11.220]  Yeah.
[42:11.360 --> 42:20.700]  We can actually get, if you search for, it should be still online, the code for all the bots from the original competition available under MIT all online.
[42:21.320 --> 42:25.480]  There's also a .NET application because someone wrote one, but it's all there.
[42:25.480 --> 42:26.000]  So...
[42:26.000 --> 42:26.280]  All right.
[42:26.420 --> 42:27.880]  And was there any instance...
[42:27.880 --> 42:37.740]  I was just curious if the bots randomly taught or discovered strategies that humans could use in their personal social media lives?
[42:37.980 --> 42:39.260]  So this is actually really interesting.
[42:39.420 --> 42:44.060]  So a lot of the bots we were playing around with in sort of Mark 1 were basically these static strategies.
[42:44.260 --> 42:49.360]  Basically, the bot would just say, you have a set personality, you have a set set of tactics, just run with it.
[42:50.240 --> 42:50.880]  But increasingly...
[42:50.880 --> 42:51.060]  Thank you.
[42:51.160 --> 42:58.620]  What we're working on is essentially the bots will have essentially a library of different tactics and a library of different personality characteristics.
[42:59.000 --> 43:01.480]  And so when you drop them into a network, they're basically blank.
[43:01.720 --> 43:01.920]  Right?
[43:02.000 --> 43:03.180]  They're like, I'm a guy in New York.
[43:03.740 --> 43:09.140]  But over time, they test out what types of things are their personality, what types of tactics are most responsive within the community.
[43:09.420 --> 43:11.360]  And then they kind of shift their personality accordingly.
[43:11.640 --> 43:11.760]  Right?
[43:12.080 --> 43:14.160]  And you can kind of do that because it would sort of just update.
[43:14.380 --> 43:14.440]  Right?
[43:14.500 --> 43:15.360]  Oh, this won't work this time.
[43:15.440 --> 43:16.220]  This won't work this time.
[43:16.220 --> 43:21.220]  And so basically, the bot's trying to figure out what configuration of personality and tactics will get me the best result.
[43:22.680 --> 43:24.660]  And much like humans, actually, that's what humans do.
[43:24.840 --> 43:24.900]  Right?
[43:24.980 --> 43:26.180]  They're like, oh, I'm in this context.
[43:26.360 --> 43:29.100]  These are the sorts of things that people respond to, so I'm going to do them more.
[43:29.540 --> 43:32.720]  And I think we're really excited about that because over time, like, basically, the bots will just...
[43:32.720 --> 43:33.520]  They're plastic.
[43:33.740 --> 43:33.800]  Right?
[43:33.880 --> 43:36.200]  They'll basically shape to the container that you put them in.
[43:36.480 --> 43:36.760]  So...
[43:36.760 --> 43:37.720]  That's fascinating.
[43:37.860 --> 43:38.700]  Thank you for the great talk.
[43:38.860 --> 43:39.100]  Thank you.
[43:40.100 --> 43:41.540]  Hey, I really enjoyed the talk.
[43:41.660 --> 43:41.740]  Thanks.
[43:41.840 --> 43:55.100]  I was actually curious, and I don't know if there's a lot of material to go on on this, but to follow up on what you were talking about in terms of emotional contagion in bots and social networks, particularly in terms of sort of artificially creating a band of specific emotional range.
[43:55.280 --> 43:55.380]  Right, right.
[43:55.440 --> 44:04.020]  And I was wondering if any work has been done to determine whether there's sort of a natural band of, like, homeostasis that happens before bot intervention on those networks, or...
[44:04.500 --> 44:05.900]  Well, yeah, so one of the...
[44:05.900 --> 44:07.240]  We've run into this problem, basically.
[44:07.420 --> 44:16.040]  I've talked before, like, a number of sort of academic audiences, and the response is very much like, this is really cool, this gives us all sorts of results, but we could never, ever touch this, because this is like...
[44:16.040 --> 44:16.420]  Yeah.
[44:16.420 --> 44:18.220]  This violates every IRB rule in the book.
[44:18.340 --> 44:19.020]  Yeah, an ethics committee would be like...
[44:19.020 --> 44:20.000]  Yeah, so the problem...
[44:20.000 --> 44:20.260]  Yeah, right.
[44:20.320 --> 44:24.340]  So the problem is that there's not actually a whole lot of research on this.
[44:25.140 --> 44:31.260]  I mean, to our knowledge, it's like the folks at University of British Columbia and, like, us, and then, like, maybe other creepy people who aren't talking about it.
[44:32.780 --> 44:43.440]  And so I don't know, actually, you know, whether or not sort of the emotional state sort of basically has some kind of equilibrium before the bots enter.
[44:44.000 --> 44:48.820]  I think that's probably what we're going to find out in the research we're going to do with the kind of happiness buffer project.
[44:49.000 --> 44:49.260]  Oh, okay.
[44:49.800 --> 44:50.240]  So...
[44:50.240 --> 44:51.040]  Well, thanks.
[44:51.520 --> 44:52.720]  I'll follow up on that.
[44:52.720 --> 44:53.320]  Yeah, definitely.
[44:53.740 --> 44:54.340]  Thank you.
[44:54.700 --> 44:55.420]  Awesome talk.
[44:55.720 --> 44:55.880]  Thanks.
[44:56.520 --> 44:57.420]  Quick question.
[44:57.600 --> 45:02.660]  You talked about the fact that you are connecting user communities that are on the same side.
[45:02.780 --> 45:02.980]  Mm-hmm.
[45:03.200 --> 45:04.540]  And there's a positive feedback.
[45:04.840 --> 45:04.920]  Mm-hmm.
[45:05.080 --> 45:08.540]  And then you mentioned about the ability to possibly use it for commercial.
[45:08.780 --> 45:11.820]  You drink Coke, you don't drink Coke, you connect them and they'll drink more.
[45:12.060 --> 45:12.240]  Mm-hmm.
[45:12.300 --> 45:20.280]  Is there a better chance that it will be a negative backlash and those are drinking Coke, you'll start losing the urge to drink Coke?
[45:20.580 --> 45:21.060]  Right.
[45:21.380 --> 45:23.320]  So you think, actually, that there's, like...
[45:23.320 --> 45:26.120]  Yeah, I think because a lot of social stuff, it does work like this, right?
[45:26.180 --> 45:27.740]  Where it turns out that, like, everybody's doing it.
[45:27.840 --> 45:28.560]  It must suck.
[45:29.040 --> 45:29.200]  You know?
[45:30.260 --> 45:31.660]  Like, and I think it actually applies...
[45:31.660 --> 45:33.560]  Well, you're actually against the...
[45:33.560 --> 45:35.400]  You don't drink because you know it's bad for you.
[45:35.540 --> 45:38.380]  So you talk to them and convince them that that's a bad thing to do.
[45:38.520 --> 45:38.880]  Right, right.
[45:38.980 --> 45:42.840]  And I think with a lot of this stuff, there's certainly an element of playing with fire, right?
[45:43.160 --> 45:44.480]  So it's possible...
[45:44.480 --> 45:45.340]  And I actually...
[45:45.340 --> 45:52.300]  I mean, given, you know, the way marketing usually works, I totally believe that some company will try this and it will totally blow up in their face at some point.
[45:53.100 --> 45:55.160]  Just because that's sort of the nature of the world.
[45:55.740 --> 45:56.740]  But I think...
[45:56.740 --> 45:58.600]  And so I agree with you.
[45:58.720 --> 46:00.980]  I mean, I think there's a lot of situations where that could occur.
[46:01.560 --> 46:08.340]  Where you imagine that, like, if a sufficiently activist user is in play, he may actually kind of disrupt your model.
[46:08.660 --> 46:12.540]  But I think what the bots do in that case is just know to avoid those types of obstacles.
[46:12.720 --> 46:17.100]  Which gets you into this really interesting engineering problem of, like, how do you figure out who's gonna follow the bot and who's not?
[46:18.480 --> 46:30.420]  And one of our collaborators, he was at Stanford last year, was working kind of on this threat model, essentially, where he was taking the model from the competition and saying, out of all those users, who are the most susceptible to bot influence?
[46:30.740 --> 46:39.580]  And he's kind of working on a couple of them, but the initial results seem to suggest that basically people kind of like on the edges of the network, who don't talk to many other users, interact very highly with the bots, right?
[46:39.580 --> 46:47.680]  And so, you know, you can imagine kind of more sophisticated interaction models where the bot basically goes after those edge users first, and then tries to move in on the network.
[46:48.760 --> 46:52.700]  You know, we're almost thinking, like, there might even be a room for, like, bots to work together in swarms, right?
[46:52.760 --> 46:55.620]  So one bot is the bad cop, and the other bot is the good cop, right?
[46:55.680 --> 46:56.580]  Like, you suck.
[46:56.680 --> 46:58.320]  And then the other bot's like, don't say that about him.
[46:58.440 --> 47:00.160]  And then basically the bot uses that to kind of level in.
[47:00.720 --> 47:05.820]  So I think there's a world of tactics here, all of them that might get around that sort of problem.
[47:06.120 --> 47:08.000]  But I think the research is still open for sure.
[47:08.400 --> 47:13.980]  I think that it may be even more complex in the case of opposing political opinions.
[47:14.240 --> 47:14.820]  Oh, absolutely, yeah.
[47:14.920 --> 47:27.440]  Because then you cannot, you know, you're trying to connect them so they can get friends, but you might create actually, again, an exponential, more extremist positions being swayed.
[47:27.600 --> 47:27.800]  Right.
[47:27.840 --> 47:30.340]  And I think we're going to have to design those fail safes in, right?
[47:30.420 --> 47:32.980]  Like, in the same situation as, like, someone falling in love with the bot, right?
[47:33.180 --> 47:35.980]  Like, sometimes you just need to, like, put a cap on it.
[47:36.160 --> 47:36.640]  So, yeah.
[47:37.120 --> 47:37.440]  Thank you.
[47:40.580 --> 47:46.640]  If your sort of social bots were entered into Loebner Prize competition, what do you think would be its strengths?
[47:46.740 --> 47:48.820]  And what do you think would be its weaknesses in it?
[47:49.020 --> 47:49.160]  Yeah.
[47:49.460 --> 47:50.740]  I think it would lose.
[47:51.600 --> 47:52.000]  Yeah.
[47:52.240 --> 47:53.940]  I mean, I think, you know, with...
[47:54.400 --> 48:01.940]  This is sort of interesting because it turns out, like, that the Turing test type things... type competitions are easy to beat on stuff like Twitter, right?
[48:01.940 --> 48:03.280]  Because the conversation's so thin.
[48:03.620 --> 48:06.960]  Like, a lot of these bot personalities fall apart on one-to-one communication.
[48:07.300 --> 48:14.280]  So if someone's sufficiently aware, and it isn't just sort of ambient information, they can really just hammer at the bot until it sort of falls apart.
[48:14.820 --> 48:16.900]  And that is certainly something that happens.
[48:17.580 --> 48:23.280]  And so I think against traditional AI competitions, they may actually not fare so well.
[48:23.760 --> 48:26.740]  But they're particularly good online just because of sort of, like, how we've...
[48:26.740 --> 48:28.520]  What are the norms around that platform?
[48:29.780 --> 48:33.860]  Although I think there's actually this really interesting merger, actually, between those two at some point, right?
[48:33.920 --> 48:40.320]  Because the thing with a lot of, like, Loebner Prize, like, stronger, really researched, really powerful kind of AI is...
[48:40.320 --> 48:43.780]  Is that it tends to be, you know, more computationally intensive, right?
[48:43.880 --> 48:51.040]  And so you can imagine a system where the bot basically uses these really simple strategies, but if it realizes it's in trouble, it can kind of call something more sophisticated in.
[48:51.220 --> 48:52.780]  Or even call a human in, right?
[48:53.220 --> 49:03.100]  And so I guess my answer is, like, it's a little bit complex, but at least for, like, most AI competitions, where it is very much one-to-one interaction, they would not be as strong, for sure.
[49:03.940 --> 49:04.440]  Thank you.
[49:04.580 --> 49:05.240]  And great presentation.
[49:05.440 --> 49:05.620]  Great.
[49:05.740 --> 49:05.960]  Thank you.
[49:07.660 --> 49:07.980]  Cool.
[49:08.140 --> 49:09.180]  I think... Oh, one more.
[49:09.860 --> 49:11.160]  That'd be actually perfect with the time, right?
[49:11.580 --> 49:11.960]  All right.
[49:12.120 --> 49:20.500]  So basically, as I see it right now, like, the worst-case scenario for your bot is, like, you have these fail-safes programmed in, and you shut them off if something goes wrong.
[49:21.400 --> 49:23.040]  What is the worst-case scenario?
[49:23.200 --> 49:27.900]  Like, I see no negative repercussions for the bot creator, pretty much no matter what.
[49:28.040 --> 49:31.520]  Like, you're pretty easily able to hide yourself and just shut these off and get out.
[49:32.100 --> 49:35.040]  Is there any negative repercussion even possible for the creators?
[49:36.040 --> 49:36.840]  Not really.
[49:36.980 --> 49:38.020]  No, I don't think so.
[49:38.160 --> 49:41.460]  Which is why I think, actually, like, you'll probably see more of this stuff in politics over time.
[49:41.900 --> 49:46.800]  Because the usual question is, like, am I gonna be found out for using bots that support me?
[49:47.380 --> 49:53.920]  But I think with, you know, with even, like, a little bit of sophistication, it becomes very difficult to detect who's behind these sorts of things.
[49:54.060 --> 49:59.420]  In the same way that it's difficult to detect, like, who's behind the Amazon stuff or the, you know, the quantitative trading, right?
[49:59.440 --> 50:05.240]  So that pattern they found between Anne Hathaway and Berkshire Hathaway, it basically happened, and then a few months later, they looked for it, and it was no longer there, right?
[50:05.280 --> 50:06.760]  Because presumably someone fixed the program.
[50:06.980 --> 50:09.320]  But it's just difficult to tell with some of these systems.
[50:10.920 --> 50:17.200]  So, yeah, I mean, I think the possibility of negative repercussion is actually quite low.
[50:19.680 --> 50:20.680]  Cool, I think that's all.
[50:20.900 --> 50:21.500]  Thank you very much.