Whose job does AI automate?

foreign [Music] it is so exciting to be here and we're gonna have some fun I'm delighted to answer that question that everyone these days seems to be wanting an answer to which is whose job does AI automate so a little bit about me this is me I service Google's Chief decision scientist that's my Twitter you'll find my handle there in the bottom right so if I say something useful or spicy I would appreciate a little tag if you guys tweet that and if you like learning from me I have this whopping six and a half hour course free on YouTube all about machine learning and AI it's designed to be friendly for everyone so you don't need any special background but this talk is not about me this is my colleague Paul and this isn't about Paul either here is Paul's wife's Aunt Dora and Dora is a computer how do we know that well this is Dora's marriage certificate from 1950 and it says that very clearly that her profession is computer and to make this even more fun her husband's profession is listed as Engineers so this is one of those really early times where a computer has managed to marry her engineer excellent now I'm sure I don't have to remind you that Dora's job does not exist today no human has a CV on which it's written that they are working as a computer so let's ask ourselves who automated Dora's job I think we know the answer to that this might be a little walk down memory lane for some of you some things happened since 1950 right who's responsible for this who automates jobs well that is what Engineers do Engineers are responsible for this Engineers are the people in our society whose job it is to automate other jobs anything that we can do faster or better or in a more fun way in a more creative way in a cheaper way whatever it is Engineers want to have a go at making that happen now that's not the question of this talk though this talk wants to know whose job does a I automate and here's a marvelous little comeuppance the answer to this question is that it is the software engineer's job behind the scenes that is the thing being automated so let me explain this to you very briefly so what's the purpose of a computer it's supposed to take an input provide an output how via some recipe in the middle right if I have some software so where does that software come from in traditional programming it is instructions that are painstakingly written by the engineer right handcrafted thought about step by step and then translate it to the computer in the language that the computer can understand now the key difference between machine learning and Ai and traditional programming is where that recipe that takes the inputs turns them into outputs comes from it no longer comes from handcrafted instructions that a software engineer writes instead here's where it comes from examples now what you put into the system is a bunch of examples and it is the AI algorithm's job to go and find the common patterns in those examples and turn those into instructions that the computer can use to turn inputs into outputs again so basically this is a communication Revolution a different way to explain what we want to machines with instructions versus examples so if in one slide you want the difference between traditional programming and AI it is this examples not instructions and what's that fancy word for examples it's data now honestly I can't get over this use of that article and AI in front of AI like like it's some kind of entity and what is artificial intelligence anyway what are we talking about when we say that word intelligence have we defined it this bothers me I would much prefer if AI stood for something more reasonable and closer to the truth like maybe automated instructions which is really what it is behind the scenes other jargon that I wanted to put here for you in four words or less when you see around AI certain words what does it mean nothing magical when you see supervised the goal of that system is to give you the right answer unsupervised the system wants to group things like with like reinforcement what is that learning by doing like when you play a video game you don't know the rules of the game yet so what do you do you try it out and your score changes and you see how that's going and that changes your behavior that's also how you learn on the job you get dropped into the job you don't know what the rules are you're going to learn by doing and get some feedback from your boss and what is generative that's where we're making plausible fakes now some of the secret tricks behind what is being called generative AI systems today is that they're not purely generative systems if they're not only making plausible things but they're also making fakes that are correct in some sense it means we're now combining systems so these are quite complicated systems being made for your enjoyment which brings me to what I want to tell you in this talk I've told you that traditionally what AI has actually been about is changing the software engineer's job giving us a second mode of communicating with machines but we are going to have a lot of confusion as a society if we don't own the fact that the conversation around AI is changing and different people are using this word in different ways now I'm straight up going to dismiss any entire usage of AI which is the science fiction sense like we don't need that anymore let's talk about the three practical ways that we might be referring to Ai and these are like nesting dolls two decades ago the people talking about AI the epicenter of the buzz and hype what was that all about who was talking about it that was researchers and what did researchers intend to do what was it for them this was about theoretical capabilities to turn data into instructions right automated instructions like I told you then a decade ago where was the epicenter of the buzz and hype there who was it who was talking about Ai and remember the researchers don't go away like nesting dolls the original is still encapsulated in the next one but the one on the outside's a little bigger there's more people doing this so who is talking about AI now it is the builders who are taking that theoretical capability and turning it into a real at scale products and now in this decade who is doing most of the talking it is the users a different group entirely and so when different people talk about AI please listen who is it and how are they referring to these systems otherwise you will start to become very confused in a way that is not healthy in fact I'm now going to say the spiciest thing of the talk this AI Revolution that's going on today that's going on the last few months that so many people are talking about here's my spicy statement it's not an AI Revolution it is a design Revolution a user experience design Revolution let's talk about chat GPT that tea stands for Transformer the first paper academic paper introducing Transformers that was published six years ago by a team at Google for six years this has been exciting to the researchers and the builders but now only now in the last few months everybody's talking about it so if you were really excited by these capabilities by Transformers a year ago congratulations you were excited by AI capabilities if you're getting excited now you're getting excited by Design and that's no small thing but let's see it for what it is here's what happened if you've used a Google product five years ago you were interacting with AI most of Google's products had AI in there somewhere but it wasn't in your face or let's take another company who here's use Netflix hands up Netflix lots and lots of hands right when you are using Netflix you are interacting with AI as well and you have been for years but again that's not pushed on you loudly why because the user experience approach of the last decade was seamless correctness deliver a good experience to you don't make you worry about what's actually happening there under the hood is it AI isn't it is it done with instructions or data or what just give you a good experience that works first time today something else is happening now for the first time ever the user is encouraged to directly Tinker with AI as AI so it's not right first time the way that a big company like Netflix says you should have the experience it's here's something that might be useful we don't know what it's useful for go enjoy it it's a raw material for solving your individual problem that is a completely different approach to design and it's massive it opens up so many Avenues of creativity we shouldn't dismiss it but we should see it for what it is a change in how we think about design and let's look again at the problems that were being solved over these decades by each group first the problems were theoretical Channel purpose tools someone else might possibly use when they have enough data and computing power then a decade ago the problems were Enterprise scale most of my talks until this one actually were aimed at that Enterprise side of things large teams automating things at scale at the business scale now today what we're looking at is individual problems being solved by these Technologies now these folks they're not inventing new AI capabilities they're not building Enterprise scale systems but they're using those tools like the amazing tool using species that we are to solve their own problems in a creative way and now the thought process has moved from is it correct to is it useful for the individual to get amazing stuff done it's huge we should be excited about it let me take you a little bit behind the scenes of this talk because I want to point out just how individual this is my talks before this and you know maybe if I were talking about AI five years ago I would be talking about Enterprise scale things look at Google scale machine learning Solutions now let's look at the individual scam my individual scale it's a huge honor for me to be here at web Summit on Center Stage so I better make a talk for you right first thing I need for a talk is a script but the thing is despite having so much writing I get the worst writer's block I see a blank page I can't start writing a script I can't work like that so I need some a little bit of help so let's play app Bingo here is it going to be screenshots from my mobile phone so the first one who recognizes this user interface this one is otter AI so what did I do here I rambled into my mic and then instead of playing it back and transcribing those words by hand I use text to speech with this app and I've got a script but it's a ramble it's going nowhere we don't know what I'm doing with it so I'm gonna take that whole text jumble and put it into a chat bot so what's this one this is Google bard and I'm just going to ask it for some talking points that I'm trying to get to in my ramble and what does it tell me I have to read it for you here if you can't see it carefully it says to me sure here is the cleaned up transcript of the speech with the points to cover in bullet points introduction introduce yourself and your credentials tell the tell the story of Paul's wife's Aunt Dora who was a computer in the 1950s ask the audience who automated Dora's job then answer to the question Engineers automate jobs the purpose of AI is to do what software Engineers do by hand traditional programming involves creating a model that tells the computer how to do something and you keep scrolling there on the phone and you get all the Talking Points not all because I still made some judgment calls and adjusted my speech but that helped me start getting a script down then of course if I'm going to make slides for you I'd better illustrate the slides so Bingo what's that one mid-journey and in mid-journey I asked it for those smiling robot nesting dolls did it get me the perfect thing first time of course not I had to rerun that 70 times and pick and choose and get you the illustrations that were best but that was much faster than me who's not a good artist trying to do this by hand myself and then finally I've got chat gbt here that's the fourth one I had to use China gbt because at the time that the slides would do Google Barn had not yet released equivalent capability now you could use either for this I I told it I have a folder of PNG images on my Mac desktop give me the Python 3 code that takes all the images and turns them into a gif and it says sure go pip install pillow put in that code and there you go it is running a little GIF of a bunch of different robot nesting dolls that I was choosing from so that is me personally taking advantage of a little productivity boost all of this would have taken me longer it would have been more annoying and fun folks there is so much writing about these Technologies it is so quick to just try it yourself stop reading about it and just use it these are screenshots from a mobile phone you don't need to be any kind of software engineer or Specialist or researcher to do this this is an individual revolution in productivity that is being handed to you another way I love to think about AI is accelerated inspiration and it is that but maybe maybe we should start thinking of AI instead as augmented individuals which is terribly exciting there are so many problems in this world that remain unsolved big ones ones that we say we couldn't possibly tackle we don't have enough funding enough people enough time we need a productivity boost we need to cure diseases we need to think about climate change we need to think about resource allocation there's so much we have to do so we need that boost we do but besides the very exciting bits of it what I worry about not the Sci-Fi stuff but what does it mean if we have unequal distribution of these powerful Technologies in society what does it mean if a few people are able to afford the subscription fees to boost their own productivity relative to others those are going to be some serious economic issues some serious labor force issues that we all together need to think about we need to step up and take the time to have those conversations and not be distracted by the robot component and also here's the thing with augmenting yourself as we enlarge ourselves with technology it unfortunately becomes easier to step on the people around you so just like if we all stumble around at five kilometers per hour it doesn't really matter how good we are at walking but if we suddenly have vehicles that can go at 100 kilometers an hour all in the same road together now it's time to step up and be a bit more responsible about how we do things so I hope that we will take this to Heart I'm excited about how the conversation is going I'm excited by these three phases solving theoretical problems to solving Enterprise problems now to solving individual problems but I hope that we will quickly go to the fourth level which is solving society's problems there is a lot for us to do there are big problems to solve we can make our world much better but we have to do it together and responsibly thank you very much [Applause]

Channel: Cassie Kozyrkov

Views: 41,692

Keywords: DataScience, Data, DecisionIntelligence, MachineLearning, Statistics, AI, Analytics, Google, GoogleCloud, Education, ArtificialIntelligence, Decisions, Leadership, Technology, Cloud, Cassie, Cassie Kozyrkov, Tech, Google Cloud, Cassie Google, Google Cassie, AI Cassie, Cassie AI, Cassie ML, Data Cassie, Cassie Data, Data Science

Id: gjDRDrOrH94

Length: 18min 54sec (1134 seconds)

Published: Fri May 05 2023