Big data and dangerous ideas | Daniel Hulme | TEDxUCL

thank you yes so I'd like to talk about actually Big Data and artificial intelligence before I do I'd like to give you a scenario so Jane lives in in London and a new sport shop has just opened on her high street and they're being an avid sports fan Jane goes down and wants to have a look around the shop she goes to the doors and browsers around doesn't really see anything she likes and leaves now little too Jane Jane's realization when she walked through the door she didn't really pay much attention to the terms and conditions that were written on the door she walked through turns on terms and conditions that stated that the shop has the right to follow her around to 24 hours a day to take notes about what she likes what she doesn't like and to the reason why they want this information this data is that so they can predict the types of products and stuff that she'd like to buy so John is a sign to Jane's case and John follows her around 24 hours a day making notes he realizes if she's 40 years old she has two kids she has Persian genes she likes sport of course she trains for triathlons at the weekend sheets power bars you know the protein bars she drinks glucose drinks she has a high-stress job and and actually John has followed quite a lot of women around in his life only for his work of course not not for and and he made a startling realization about 40 of the women that Jane has that John is followed around that had very very similar profiles to Jane turned out they had a high proportion of them had very aggressive ovarian cancer okay so a large proportion of these women that have very very similar characteristics as Jane have this this rare high aggressive form of ovarian cancer now I want to ask you a question should we should we let Jane know this information I want you to think about the consequences of not letting her know think about the consequences of her dying a younger age the cost on her children the cost of society but what if we do tell let Jane know let let her know that she has a high chance of a Varian cancer and she doesn't and that causes stress so just a quick show of hands who think we should tell Jane that she has a high chance okay it's hard to see I've got big lies in my face and who thinks we shouldn't tell Jane okay and who doesn't know kind of exactly right we are we have these these questions these ethical legal and moral questions that we're currently trying to deal with right now and I don't think anybody really knows the answer so what I'm going to try to do is present you with a framework a framework to under stress under stand data-driven decision-making and then hopefully use that framework to to raise some of these important moral and ethical questions so the framework has five components and I want you to really try and remember these the five components are data information knowledge understanding and wisdom okay what what is data we've seen lots of words a lot lots of references to data what is data in Latin data means given things which I interpreted as stuff data is letters numbers pictures it's it's all this stuff on the internet it's all the stuff that's connected via computers and two one zero two eight zero is data it's not that could be a date of birth it could be a it could be a sort code it could be how much money higher I have in my account it's not until you actually contextualize data does it become useful and just so you know big data just means lots of symbols it means more symbols and you can process on a single machine so if you think about all the stuff that's in Twitter all of the movies that have ever been created all the digital movies it's more than you can process and store on a single machines or all big data raises all of this stuff distributed across the internet in companies and your emails and all things like that so the information is data in context as I said it's when you can't contextualize data then it becomes useful so that number that I gave you two one zero two eight zero is a date of birth and there's a huge movement at the moment going on behind the scenes in the internet called link data where what we're doing is we're empowering machines to give meaning to tato we're actually giving giving a pieces of data meaning and linking it all together and I'd encourage you all to go away and look at blink data because it's really going to change the world and once we start to link data together and give it meaning then we can start to ask some very interesting questions of it we can organize it we can query it and I would argue that the definition of knowledge is the organization of information once we've organized some information once we've identified some patterns and trends we can start to know things so we might know that when the weather's warm that there's a there's a high chance of ice cream sales right so as such a huge amount of work that's being done in computer science to identify patterns and and predictions in in data find using visualization using machine learning to process all of this information to try and find these patterns but knowing something doesn't mean that we understand it and that's the fourth part of this framework so we know that when the weather's warm that there's more ice cream cells and we can use our intelligence to understand why that's the case because when warm weather humans are warmer we like to be cooled down so we equal cold things and actually computers find it very very difficult to understand to interpret knowledge it's a there's a huge amount of research out there trying to figure out how we can we can do that but just knowing just understanding that when the weather's warm there's more ice cream the being sold doesn't mean anything what we need to do now do we need to make some decisions based on that maybe if you're an ice cream company you can go and and and and predict that there's going to be warm weather next week and therefore try and sell more ice cream and so that final part of the the framework is wisdom is the utilization of understanding it's using our understanding to make better decisions and these are these five components are starting to be brought together now in academia we're building systems they aggregate data that mine and look at big data that find patterns in it and try to make decisions based on some of those patterns and and I would argue that that's the on the verge of artificial intelligence that's what we do right we look at all of the data we gather it all together and we build models and we execute on those models and we make decisions and once you start to bring all of these different components together in a computer then I would argue that that's true artificial intelligence in fact the best definition of intelligence that I've ever found is goal-directed adaptive behavior the idea is you've got a goal and you can behave or act in the world to try and achieve that goal and perhaps the most important word in that definition is adaptive that's what we do really well computers don't do it so well but in the next decade perhaps we're going to get computers that can automatically learn automatically adapt ok so now you have the framework let's start to understand some ask some questions of it so the first set of questions are I think around data we've heard from previous speakers that the more data that we have perhaps the better decisions that we can make from it so let me ask you a question what if what if all data was open what if your bank records were open your your medical records everything that you did was completely open there was complete transparency I reminds me of a story from from Plato there's a there's a story from one of his books called the Republic and it's the ring of guide use and the guide uses a is a shepherd and he finds this ring he puts on the ring and realizes he has the power of invisibility anonymity what do you do when you have that power of anonymity what those guys do here he rapes a queen rapes a Queen kills the king and takes control of this the kingdom and and so we're starting to see some of this this behavior on the internet so people that are anonymous are trolling people and doing really heinous things so let's say I might made my account details to you and actually I gave you the ability to take money out of my account that would probably happen very quickly they could have by student debt as well if they want it but imagine they want to take somebody out of my account but let's say that there anybody that twitches me all that data is also made of available so if you take some money out of my account then that data is made publicly available to the rest of the people so do you think that you would take data out of my account sorry take money out of my account if you knew that that was gonna be public I think in a world where there's more data more transparency more public knowledge of data then perhaps we would behave better towards ourselves I don't know the answers maybe there's some research over the past decade about being able to look at people's brain brain electrical brain activity and try to understand what they're thinking so whether thinking about an apple or a house imagine if some clever dude comes along and is able to do this at a distance so he creates a device that's able to look at people's brainwaves and identifies patterns to understand what they're thinking or you gather together all of the all of the the gadgets that you're wearing and you can start making some predictions about again what people are thinking should that data be made available I don't know I heard a story recently about terms and conditions where some some people in Canary Wharf were providing some free Wi-Fi to people but to get access to this free Wi-Fi you needed to sign up to their terms and conditions in their terms and conditions it said that we we now have the right to your firstborn so I think they're about 250 people that have now signed their rights their firstborn into this organization to get free Wi-Fi so I would encourage you to to make sure you read the Terms and Conditions and actually this raises a good point and it's been raised by some of the other speakers which is perhaps some of the things are inside these terms and conditions should be made more transparent to you guys what can the companies do with the data they're collecting who are they selling it to in the same way that we had we had complex food labels and it's very very difficult to understand whether these foods that were consuming are good for us or bad for us the government has legislated that we have to have these labels on there now to see how much salt and sugar and things like that are in our food and perhaps we also need the same thing for our terms and conditions so the second set of questions that I want to ask are about inferences about extracting inferences from from data who owns these inferences when I sign up my my my data away to an organization I tell them my name and my address and things like that and I actually have the right to know what information that organization has about me but let's say that organization make some inferences at me it makes it inferences about when I might die or whether I'm having affair or something like that should I have the right to that information - I don't know imagine if we are if we were able to - in James case help people understand their their health problems before they become a detriment to society society so this question about who owns the inferences I think is a very interesting one what happens if we create algorithms that are prejudiced who's responsible for creating racist algorithms right so if you if you want to go and get some insurance that there's some algún they're deciding whether or not you should have insurance but that algorithm is using the color of your skin your ethnicity perhaps your sex to determine whether you should have in insurance or not it's actually very difficult to look inside algorithms and see how they're making these inferences so who's responsible for those inferences I don't know the answer to that and finally and finally decision making ok so a decision making is actually really hard it's hard for humans it's hard for computers and at the moment I guess date big data is used to decide what to show you in terms of what products you might like to buy or not buy but at some point we're going to start to get systems computers making some more complex decisions for us and eventually we're going to start to embody these systems in these artificial intelligence systems in machines in robots and so maybe we'll want to try and get our robot to help us understand how to end poverty or to end cancer does anybody know a simple answer to ending poverty and ending cancer and humans so we have to be very careful about how much power we give we give to these things because of course the Machine can't contextualize things in the way that we do they can't they can't understand the question we want you to end cancer I end poverty with all of these other constraints don't don't destroy the world and and and don't kill people we also basically assume those things because of lots of lots of evolution and of course all of this prior knowledge that we have so so big data I think is the catalyst for artificial intelligence and I think that what was one science fiction over the next decade is going to start to become science fact rather than us intelligently interacting with the world what will happen is the world is going to start intelligently interacting with us and and it's going to change everything it really is I can't imagine some of the innovations and amazing things that are going to happen in our generations lifetime it's going to be insanely powerful and I do think the bigger the big data and and artificial intelligence is one of the powerful things that is going to happen since the Internet but my uncle once told me very wise man he said Daniel with great power comes great responsibility thank you for listening you

Channel: TEDx Talks

Views: 84,678

Rating: 4.7755103 out of 5

Keywords: English, tedx talk, ted talks, tedx talks, ted x, Technology, ted talk, Public Policy, TEDxTalks, ted, Philosophy, tedx, Data Science, United Kingdom, Computer Science

Id: tLQoncvCKxs

Length: 14min 39sec (879 seconds)

Published: Thu Jan 15 2015