BSidesCharm- 2018 -To AI or Not to AI What the US Military Needs for Fighting Cyber Wars Ernest Wong

at the united states military academy at westport new york um the top of the big day is uh it's a play on really shakespeare right hamlet uh you're not speaking about or not to ar and uh trying to figure out what the us military needs uh in order to fight our future wars right sidewalk uh the title i did not give the speaker committee was well what i learned about mainline from hollywood because again as jessica mentioned from the keynote speech whenever i'm in a meeting i usually think i'm just this guy's work so i'm actually in a very good place in our idea uh posture and tale i like using military analogies a lot by talks we've seen a lot of my

other

where the enemy or really the victor has some type of physical abandonment usually the size could be speed could be agility but support this then we have this industrial revolution right here we have resources are better right it's no longer just a physical attribute if your nation or your company or your team generate more resources you're going to win worse nutrition then in the long run knowledge right physics if you have the atom bomb you could dictate that's why there was space ranks race then now we're thinking about information in cyber warfare my argument is that innovation is what we need right with justice our attack speed of defense is the critical aspect it's not just knowledge how you use that

volume how fast jessica she did a great job right she mentioned obstacles to attack right signals symbols i give flags that come up that's his whole notion of innovation now with this background uh what this talk really the genesis of this talk came out of the sponsor not only are my research scientists at the arms diversity i teach as well in partnership journey and i collaborated with peace [Music]

graduates with a bachelor science so the interesting is not the you get to graduate with a factor science and math or bachelor of science in engineering disease right your science american politics master of science and laws that's because we have a required curriculum a portfolio volunteers have to take and in my case i teach curriculum for system engineering and this is the capstone experience that they undergo uh in this case it's not a year-long project provided semester-long project and so when the connection trying to redefine this problem statement that chief wedding gave to us how do we use artificial intelligence to combat really perform terrorist organizations financial industries what the guys came up with as

part of this uh prolonged research project their first revised strong statement they actually drafted or during what she was already identified as a potential force fashion using ai to combat terrorists i have to talk they didn't have not have that expertise they needed to clear this project so they went back to drawing work and came up with the spinal cord themselves to encounter bible extremist organizations and they wanted to find the best ways right not just technical as chief buddies due to artificial intelligence the best ways both can't plan on to counter violence groups of terrorism so with that background right basically that's right they require a bunch of these systems towards realities we have

a qualitative oil later on this afternoon with here about a map it took this a quantitative model he did can i stick to some quantitative modeling right trying to get some metrics to figure out what's the best way of managing financial streams and very quickly some value measure graphs for those key metrics a at least a more quantifiable way to figure out what's the best way of doing this problem or attacking yourself taking maybe those three courses right technical option uh option two non-technical option one in the middle of some hybrid now what i learned from this right so i'm taking what the guests did and the great thing as a instructor at west point they don't

give us tas right all professors

i was able to bring these projects in uh for undergraduate destination attackable and maybe team like tas surprised no one else from these guys figured that out some cadets that worked on your products so what i learned really is really the difference and extension between human and machine what the advantages are for each such a very low level approach this again from a human perspective right about critical thinking creativity right social intelligence

we saw a lot of that from jessica's pockets right processing logic uh possibly capabilities procedurally population graphic automation and so really what ai does for us but ai is trying to put the weight of the scale right he's trying to compute human characteristics onto that machine side to give us our official pledges so my call machine learning subset of that and this is a quote from one of the dartmouth former director pre-department records his third way of taking election donation starting and speaking machine learning protecting us and if you don't know the third way what it's really referring to the first with nuclear weapons right the u.s game dominance and the uh august race just quietly the second wave was willing

to get darker led by dark a lot of efforts with ways to combat the russian ussr's uh mass superiority right usr the way they thought of the warfare was they still want to focus on international minds we're going to win we're going to have bigger armies with or a bigger army there are bigger allies or deter you from doing a lot of action that are against us so how do we as a nation how do we defeat their integrated air defenses how we get well we can't send bombs anymore like we did in world war ii right now how do we get our aircraft or a cheap facility in the ocean well darkly advanced members right

grenade ones being stealth right now we can penetrate the radiuses yeah no detectors we have precision vibrations we don't need carpet bombing anymore yeah you should have

our own locations we know more information about ourselves not only about ourselves but again so really those three things form uh what was the second one a third way of innovation is this motion of perhaps augmentation of computers with human characters more efficiently now wired magazine has a very nice companion talking about rookies of the i started back in 1956 with a cool uh dark metal professor during their summer future activities hey how can we actually get these computers to start being more effective

a lot of famous universities mit stanford uh law organizations mostly educational institutions

network healing self-defense recovery we have ibm's watson what about your cell phone your alex who's got alexis he's got i'd go back to halloween you didn't hear the confessions

um

these

we all know these movements at first right how was the protagonist right the computer and outdoor population alien terminator one matrix i robot then right

good too

then that one single um and then we have disney right disney game but then i get it all the way back right going back to disney i will hold back the uh

of that right that's the promise and the fear of artificial intelligence no one talks about the ugly the ugly of a i is because if you really think about it otherwise we should be uncertain right from the galaxy we have marvin right morgan the what they call it the uh uh they actually called it i didn't understand that i went on wikipedia this morning i just wanted to make sure it's like marvin the paranoid yeah the reality was he was super depressed and he was uh stupid and again so if we think about it but if you really can't we're hoping ai does something but reality is the variance in these systems we need to be concerned and ugly as well

because really there's some great quotes in marketing marvin any ideas really they all point to certainly yeah martin's [Applause]

the first 10 million years were the worst and the second 10 million they were the worst two the third 10 lead i didn't enjoy it all after about i went into it so again we have to consider it ugly okay if you're playing with artificial intelligence i'm not really talking machine learning but if we're talking about artificial intelligence the possibility of getting to is a possibility now the my system should hear this actually across the three points maybe ai is not the only way of doing this right they talked about some human activities that are taking place i also talked about less about this look at this this uh model the other way instead of having human interactions

right and becoming machine you can do the other way right bionics you have you know engineering we have cyber cybernetics this notion right yeah we get some epic launches here us military does i can't confirm that i don't have enough um um but there's this other possibility right we have uh human augmentations for this case uh superiority this type of warfare tools and for me again i'm looking at this from a concerned point of view again if you go with righteousness uh biotics or instagram man but that's the good side bad side and right there's always this network left and we have again from there i think tournament coverage

because for me the best movie that minimizes this concern of human augmentation is you won't hear this but i always like putting in star wars i can base all my thoughts and star wars so this notion of probably star wars 4 right they uh [Music] this whole notion that uh right as darth vader becomes more and more mechanized it gets more pulled through the dark side that we can certainly

now yeah we've heard all these stories right you can google just do a google search for top ai failures these all these pop-ups they haven't had right since the 2000s it's amazing right you learn about all these things right they have robots learn racism and it's based on right all the data how you train your day life marco chain gives a great uh precise talk on finding case decks and it's really about this the abi at this point is not technically developed enough to where we can train it to do what you want you have to train our in a proper way to make sure you function problems but even justice talked this morning right we have

strong adversaries who are adjusting their referrals on a daily basis to support potentially a possibility yeah that's the reality you just need to be aware uh other other things where reality is greater than fiction right remember facebook's artificial language that creates some language well the other side of it but again ai did something right the new program did something they did something that researchers did not intend her to do very introduced bearings in the system and it's not just in the us okay here's china china's attempting at uh they were actually attempting to censor increase the censorship of their content and they're allowing eye to

but here's an interesting point i actually like putting on this china one because um i give another talk on innovation i talk about innovation and speed of innovation it's really the way i look at it is uh i think the chinese are learning when they need these things like this just like we are i just like facebook just like microsoft just like people never make mistakes china's learning you start looking at more and more reports of ai i've seen more coming out of china stanford and china china researchers just came up with a language program that's actually doing much better against the more and more trades

[Music] so that's a concern when i see them failing that's actually not a good song it means they're experimenting and they're learning

another way of learning more about the artificial intelligence we do a thing called breakcasting twice a year and uh we have in our audience really actually two of my colleagues are the key members of this threat casting group and one of the foods that we talked about well one of the sessions was on artificial intelligence and what

[Applause]

artificial intelligence and this pop-up guys so if you want to read me more about it it's a pretty good read as i tell my cadets when i show them my books it's generally a pretty good read when i have tabs all over the place uh or i'll tell you guys yeah it could be my social media way of just telling

pick up this talk really as an introduction to uh hey earlier because he's going to be talking about a qualitative event that he's used to better participate what our future needs are in terms of artificial projects so hopefully you'll come in for that and uh hopefully you sound better now i revisit this i know i'm getting short time now i'm going to revisit this note um of what i can decide what about what we should be concerned about and the reality is how do you think cyber warfare will if that ever comes to pass my concern i actually i'm not taking this to myself

[Applause]

but for me again [Applause]

[Applause] looking into is fiction i've done that a lot i can't lose science fiction instead of getting a spa there's really more data right in the next generation it's really damaging instead of looking at these guys we should be looking at especially for example voting [Applause]

[Applause] um [Applause] [Music] [Applause] [Music]

[Applause]

next feature set up with that i'll

[Applause] okay [Applause]

that was not related to the head logic i'm sorry i can't have you to walk past

[Music] are you ready

so my apologies it looks like this bike is busted all right i'm gonna do my test also [Music]