Press "Enter" to skip to content

“Do Robots Have Free Speech?” | Byron Ruby | TEDxHarvardCollegeSalon


thank you to our tanks for having me

here this is incredible honor and great

pleasure to be talking to everyone here

today so the title of my speech is do

robots app free speech and you may be

wondering what does either of your law

student know about robots let alone code

or anything like that it’s been well

known that most law students know

nothing about code and I’d say okay

maybe that’s generally true um Steph

league in my case as well um but the

story of word of this sort of came from

was that back in december i had the

opportunity to go see conference in

seven where a where Aaron Pettigrew

gawker gave us short speech short

presentation on the use of new

technologies to micro target messages to

individuals in the course of a campaign

and one of these your technologies is

boss and that piqued my interest they

said you know it may be what I’m going

to do is I’m going to research this and

so this presentation is the result of

this research which is still ongoing but

it’s kind of Justin shared with you

results we’ve seen so far so um I’ll

admit the title is votes Ingenuus

because I’m not really talking about

robots like this little guy i’m also not

talking about things like Siri or

Amazon’s Alexa all those I mean else is

my department when I’m talking about our

automated social media accounts AK this

stuff so like in popcorn and receipt

sort of your garden-variety Twitter bot

which has some algorithm where you know

they’re basically posting different

peoples Twitter handles this line did

you know that Facebook is disabling

accounts of activists um not true and

then I’m also below 28 weeks which if

you haven’t heard of was a Microsoft

experiment to create artisan an

artificial intelligence teenage girl on

Twitter um it didn’t end well it lasted

about 24 hours and then rhythmic

catastrophe if you want to know more

just ask me in human

about that so what are the uses of these

box so generally speaking de as

programmers and the platforms um become

more sophisticated and AI becomes more

robust the uses of bots for liberation

increases and so I link sure to bump

into two categories there’s benign uses

and there’s malicious pieces so but nine

uses can include something like having a

bog on Twitter that gives you a weather

update we’re automatically post a score

from recent or you meant that healing

work otherwise have to be manually um

there’s a reddit box that go around and

look to see if any links in your post

broken and there is one finds is just

going to post under you and say hey by

the way buddy here your link is broken

you may want to check that those are

good uses there’s also some uses of us

to actually engage people when they use

things like or racial slurs so there’s

one guy who published the results in the

journal club a behavior where you found

that and he had a bot that actually when

I’m certain that honest people or

engaged and constructive feedback when

they used to certain racial slur that

person that they parted was far less

likely to use that score in the future

which is really cool in my opinion that

that’s a good use however um there’s

also facts malicious so these can

include things as sort of entry level

bad is like retweeting stuff who’s you

know boosting how many likes of those or

downloads on certain posts that you see

generally speaking making it seem like

certain posts or ideas are more popular

or less popular than they actually are

we’re still there’s also the ability for

these for these boss to engage big

speech and for spread this information

or lives or conspiracy theories and if

you think this is all sort of abstract

is not there’s real consequences to us

and just to sort of illustrate that we

can look at the recent of the recent

incident in DC where a gunman showed up

at a pizzeria because he had read online

that in the beast of this pizzeria was a

secure paedophilia like rain run by

Hillary

john podesta um so he read this on lock

and the thing was it started as some

sort of hoax or a joke or something you

need to read it or fortunate got picked

up on Twitter but according to Professor

long are started professor I’ll write

home to pursue belong he found that it

turns out that most of the accounts that

we’re talking about this weren’t

actually people it was us and so it

seemed like there was this groundswell

of common knowledge I’ve all my gosh

there’s this this pedophilia ring would

really that wasn’t the truth and it led

to very serious consequences that said

there’s again there’s not just religious

or the nine muses there’s also some sort

of in the middle like commercial uses

applause so lots of easy sell things and

if they can sell products they can also

sell people and ideas and that leads us

to 2016 election so here we have some

research from the University of Oxford

which looked at tweets during the

presidential campaign in here is it the

general sort of sort of sense of what

the tweets are going on during the run

to the debate the black is just general

angel hashtags red was Pro trunk we was

proclaimed and they analyzed it with

machine learning detect which accounts

were fielder which ones were fake for

BOTS I’m in turns out that this green

line which is the number of bots making

these tweets is about twenty percent of

all the traffic now that’s that’s that’s

a fair amount and the problem is dead

unassuming news outlets like CNN more

necessarily we’re discerning between

they would just will get topics and

we’re trending and it would then we

amplify a message that was already being

amplified by the use of these plots and

so you may step back and say welcome

back what’s the big you you know like

okay so there’s my boss like who cares

they’re just throwing around garbage we

can all tell the difference between

ivana even right right turns out it’s a

lot a lot more difficult so there was

one experiment 2014 we found that

seventy percent lots went undetected by

Twitter there’s another study from the

University of Reading where during an

actual Alan Turing test

we’re sort of flying tank and look into

tech bots they only got it right seventy

percent of time and got along pretty

time these are people who are like

actively looking for it my hunch is that

your average user of the internet is not

not discerning and that’s not because of

mrs. so stupid it’s just because um

maybe there’s a difference in tech lon

savviness for education levels and also

just in general these bots are getting

really sophisticated really quickly in

fact we’re kind of like terminators um

they like have a face they have a name

of a profile picture if fabricated geo

locations and pending on how

sophisticated programmer as we move you

made it they can really actually engage

with you pretty easily on like an a

question or a top or a dialogue so if

you’re like me and you care about

whether or not the internet is still in

public domain for public discourse

little discourse the exchange of ideas

exchange of truth for people that

actually debate persuade each other it

should be a little concerning certainly

Trump campaign was the only campaign

that did this intended to but the idea

that there’s sort of these armies of

matza they’re sort of polluting our our

marketplace of ideas it is alarming and

at least a very sort of simple inside

which is that as more and more political

discourse news online internet all so

just as an informal poll I just want

people to raise their hands how many of

you in the last like three months got

into a debate on facebook or reddit or

something I thought mumble something

didn’t know if you okay for those who

raise your hand how many be done that

with equal or more frequency in real

life okay so service serous same so

maybe you guys aren’t miss Layton

representative sample just because maybe

you guys are more plugged in or activate

gauge especially if you’re coming to

attack on politics end but my mic that

would be that most people are less

having fewer in person talks and more

online class and so this makes us ask

well what should we do about their

three things so first of all we could

talk about it um the very fact that I

someone who was working on my campaign

this fall and i think is relatively

plugged into her internet trends didn’t

know about this until about like two

months ago certainly a month after

election shows that auslly women are the

radar this year but the more we talk

better the more be certain moment lure

people to those are the presents and

have a function maybe people will be a

little bit more discerning about what

they see and what they read on the

internet so that’s step one step two is

their proposed to be technical solutions

and so platforms of what i call it front

and make tiny access to api is to sort

of reduce the number boss they actually

want another platform or try to

eliminate them all together with old

technology old relative sense technology

length captures on the back end if they

do want some box on their black arm like

Twitter mike still want some good box

and they can use advances additional

learning to try to detect which box are

good with robes are bad which one of

violating terms of service and get rid

of that once enjoyed one spouse but this

leads to sort of a another interesting

scenario which is well if that’s the

case are we just bleeding ourselves to a

sort of cyber arms race where it’s going

to be the pots versus the ante pots um

you know and it’s just going to go on

and on and on is there any room where we

can maybe get Congress or the law help

us on our side and try to prevent this

from get this under control and this is

a really interesting question and this

is where or off whereas I can’t miss

Elise be fun the technical side of

things I can talk to you about what

limitations that might be move along and

so before I answer this question let me

just first ensure that this is really

hypothetical because I mean certainly

the law lacks the high technology

generally speaking but on top of that

we’re living the political climate right

now where it’s very hard to get anything

done it’s very unlikely Congress who’s

going to want to have to take the take

on this issue they don’t really have a

full power at the time but if we were

nonetheless is your task all right well

how would you do it if we could do it

what would be clear what would be the

optimist well the first obstacle is

about 200 years old and goes like

it says Congress shall make no law

abridging the freedom of speech in

pertinent part that’s the free speech

Clause of the First Amendment and you

may say look Byron you have your answer

oh I’m just going to try to regulate the

use of these pots no not sorry First

Amendment that’s it you know it’s like

if the programmer has the right to free

speech doesn’t bought 28 hey maybe not

it’s actually a little more complicated

and answer this it’s worth taking a step

back and analyzing the levels of speech

that actually occurred within the robot

or the bottom the first is the code that

goes into the body style and then the

second is whatever the body is saying

and so because the second level a little

bit easier to digest we’re going to

start with that first so generally

speaking and the First Amendment

whatever a box says is usually protected

police’s is productive anything you or I

might say so just as for example if i

make i use fighting words or threaten

you or I engage in speech that is likely

to incite imminent violence those things

weren’t protected under the First

Amendment just as much as it would be if

I fed it online for a bot or I said it

to one of you right now person but but

that’s that said there’s there’s another

level speech because what if we had

Congress by law said fine you can have

your box say whatever you want just say

that the published just make sure the

box says it’s a bot it just reveals its

true identity what about that and that

also might have problems and to do that

if you look at the cover so the code

itself is considered a form of speech at

least under some circuit so there’s one

case full of universities versus

horribly there’s no case of United

States Bernstein where the second line

service respectively have held that

phone is a form of speech just because

it’s in sign poles most of us are many

of us doesn’t mean that it’s not a

speech cognizable under the first clinic

however with the idea that age speech

dismissal II mean that it’s necessarily

going to be

and we run into something called a

compelled speech doctrine which may true

for a wrinkle to this and illustrate

we’re going to look at apple Apple’s

recent her bubbles so they call it with

the FBI so last year the FBI one hack

into the iphone of the San Bernardino

shooter and Apple I said we went to

apple into apple why does some code to

let into the Python and apples like we

politely client I’m so happy I was like

all right we’re going to go to the poor

and make you write this code and I was

like weep lightly fine and so a legal

debate legal incident came before and

these an apple Rogan emotion we get to

the order saying look this is compelled

speech you’re making us speak and it’s

offensive to our poor friend envy you

may be thinking now who what is this was

there any merit to this very milk what

is this compelled speech doctrine and

the answer that actually was all the way

back to nineteen forty-three in a case

called West West Virginia Board of

Education versus Barnett where the court

will have very landmark case that the

government can’t tell you to express the

opinion or believe that you don’t

actually believe it and that’s referring

to political opinions it was the case of

a kid didn’t want to recite the Pledge

of Allegiance so FYI done you have

precise budget we do to don’t want to UM

on that but the Apple used this argument

to say we find writing this code

professor so the problem with the Apple

FBI debate was never got resolved FBI

found a workaround the poor poor case

one way we can moon so that we’re never

actually answered whether or not the be

anything to whether that to the extent

that this and whether or not this

counted as compelling speech so here’s

how is it the way I see it is that code

is speech yes but code is a set of

instructions to machine and the bot is

separate from the coat in fact the bomb

is worth an emergent property of not

just the code that

used to provide it but also the code of

the platform’s it used to interact with

interfaces with the wires and the cables

and servers that create and comprise the

internet and then ultimately you and

your browser where you’re looking at

that bottom reading what it says so the

code and the bond are actually sort of

different and so looking at what the bog

is used for and the intent behind it I

think allows us to make a relatively

relatively colorful distinction so I

would say that it’s probably

unconstitutional for a poor a precaution

passed a law that says if you make a pot

of your own personal use for you to that

insert code into that box say it is a

month to reveal itself as a pop why be

hey this is your form of three stages

your form of hacktivism this your web

making part so your bot if you want to

do that you don’t have to reveal that

it’s a bottle but if you’re selling

balls like the kinds of we’re sold to

the Clinton and Trump campaigns this

glass election that’s the story because

they’re you’re not engaging so much of

speech as you are in creating a product

you’re creating an empty signifier that

someone else is going to use and use to

proliferate their message and that

respect government can compel speech so

we might get around this compelled

speech doctrine in the same way that the

FDA is able to make food manufacturers

put warning labels on the food you eat

so no process likes for me that way they

have to study how much sugar in visit

and maybe Kellogg thinks that really

offensive but too bad they still have to

do it so that might be one answer at how

we could solve this riddle of when we do

with thought can we have it always

reveal themselves as a lot that when

you’re going to hear that you know oh

this is a lot this is the real person

the other way to conceptualize it is

that the bot isn’t so much speech act as

it is like a billboard it’s a medium and

the Constitution allows us to regulate

media with content neutral time and

place restrictions so a law saying that

billboards have to be 20 micro defeat is

totally constitutional similarly a law

that says a billboard has to have a

little sign over here that says this is

a billboard is redundant but still

constitutional and so in theory we could

have

we could have this exactly work and

there’s some precedents of that there’s

a court case motor be as you see in the

Ninth Circuit where they upheld

Congress’s attempt to regulate robo

calls and they said regardless of a robo

call is for a commercial or political

purpose it’s a time manner in place

restriction and its content neutral so

that’s okay so there might be an analogy

there it’s plausible about not probable

but it’s plausible um but but with that

I sort of one is conclude with a sort of

open questions and say okay even if it’s

legal to give this to do this regulation

should we do it and here are three very

well-known justices homes Brandeis and

Harlan can be just as a theory free

speech when we should regulate it and

why and they all have answers to this

however what I think is sort of a large

web site from las is that you know the

technology of the future is really going

to stretch the wisdom of the past and so

welcome to brave new world of box thank

you [Applause]

Please follow and like us: