0:06
Hi there. I’m going to briefly introduce you and then I’m going to give you the state. So doctors can make explores maximizing capacity and because you know scientists and defenders, I just love the polar scientist’s bad side incubator in fact economic intervention, and the co-founder and chief scientist she applies machine learning lessons. effects of chronic stress were named Max and why not? Just be

1:13
One wonderful chatting medication actually inspired me to start off going to have a somewhat colorful history calling myself a delusional mess. Truth in advertising I certainly didn’t start off by liking each other. My family was not wealthy in Silicon Valley since my dad was a doctor. My mom was a teacher. I grew up in coastal California I had a lot of advantages actually made pretty clear to me early on. I was supposed to win the Nobel Prizes somehow also be a sports star. Sure my parents had done that would all balance out but they weren’t cool. It was just and so I the same

2:36
Person unfortunately and so some that self-inflicted I found some purchase back and over the course of a very frugal 10 years. These are not my decade of academia graduates, I didn’t know what they are, and besides just talking about learning how to Yeah That wasn’t the coin you consider going to study economics but I am concerned about terrible failure.

3:26
Okay, can I be concerned

3:34
And this topic contains a single year but nonetheless, this was a delusion for my exploration just always say.

4:09
Never again, that was professors the professor the only program and he said teaching assistant next year, by the way, recommended working at this place called the machine lab on campus at the University of California in San Diego sponsored by the CIA to do research into real-time facial for lie detection. I think we can figure out why the CIA and the CIA were interested in such shady, ethically highly questionable technologies. But I’ll tell you, as someone that had no idea about when I was outside of science fiction, actually writing the code that could find a filter. I didn’t know what that little fold on your nose was called. That’s what it is. And pupils and from that, figure out whether someone was It was amazing. And I somewhat very quickly realized I don’t really want to stick wires, cat brains, or give macaques rabies retrovirus to trace out their motor pathways. I wanted to do this field-theoretic analysis, which again was just done. I mean, lucky that my life on the streets and then lucky again to just walk into this incredible opportunity, and the phenomenal data, and machine learning and programming. But I extend this story, to argue that its real value is its creativity. Its ability to solve problems and eventually we’re going to talk about ethics. And we’re going to talk about learning routine skills and everything in between. But in this case, being part of some of the early work in real-time facial recognition and analysis has given me an interesting perspective over the last few decades, watching that. Particularly because my undergraduate labs spun off as a startup and eventually got bought by Apple. So if you have an iPhone, it’s facial recognition as the output of things I worked on a couple of decades ago, as an undergrad. It’s an interesting thing to know that you are there at the moment that this thing that is touched 100 Million people’s lives and frankly, from the algorithms developed out of academia and beyond. Its touch millions because I’m sure we’ve all seen news stories of abuses, facial recognition systems, and probably many of them really, systems that can’t see dark-skinned faces, systems, that thing that most of the major professional athletes in Metro Boston are probably wanting criminals guessing because they’re either black or have tattoos. These are not good systems. And in fact, one company, in particular, thinks they can analyze your face and how will you actually get the Chief Scientist when it comes to the first company in AI in hiring, as well as someone who’s building connections. And I will tell you that someone who’s selling something in the market, which is inarguably wrong and unethical. Isn’t it possible that someday we might be able to use information about faces to advance our understanding? Higher it’s possible but as this case, example, we haven’t made great choices. And then maybe, but I also have some other stories to share about patients for example, very early. As an entrepreneur, I went off to grad school, and then was a faculty at the University of California, Berkeley, joining Stanford for a little while, and then my wife and I had an idea of education. And during that time, I also did some collaboration. They gave me a pair around the time to recognize. This is actually I think, the third pair that they gave me this not even a software person, be honest, I have written that I don’t know how many lines of code in my life, quite a few. I hated every second of it.

9:42
What I love about programming is gonna show you how to solve problems. So when Lou gave me an early class, he said, Can you come up with ideas of cool things to do? And so I did one of those analyses faces, read expressions. And there’s also obviously, maybe the CIA wears a pair to tell whether you’re lying or maybe a bank loan officer letters in to float everyone’s credit scores of their heads, or in a call up your Facebook and LinkedIn account whenever I see could be great. I built it because I wanted to give someone a superpower. I want to give them the ability to read facial expressions because it turns out they have autism. And one of the most common symptoms of autism is an inability to read patients. Something the rest of us get for free. They can learn it, it’s a bit like a foreign language they have to explicitly so I, I had this chance to build a system where instead of looking hard to flash, where a stick fake, smile, learn with them and have kids come and see an actual person who they were interacting with open the little heads up just tell them whether they were smiling faced or they angry.

11:29
What we found is not much better. But it’s nearly impossible

11:41
To separate symptomology happiness and sadness simply learn to not believe really hard to learn by other people to see things the way they do. So you’re able to use patient to help some people that need a little help and another instance actually during that enacted by the UN and it’s a wonderful story itself, but I’m short here and say building the system learn how schemes forming in class, tiny interventions and they’ve done the demo that faces for the venture capitalists that was investing or starting to show our system, how we proceed

12:54
Just about the thesis, but how we got the system where you have an idea in your head, Southeast Asian Muttonchops and if you are actually consistent in how you made those judgments are guessed the faces in your category. And then we charge very coolly. We actually built another moment of the sexy face. Terrible. A promise was that for $5 We will find you, someone, you think the truth is, and trick people into playing the game because we confess to the mind-reading as they play the train that royally This is a bit of an ethics situation, tricking people to play a game. They think they’re gonna get laid. We’re not actually going to do that. But our action was to reunite orphan refugees with extended family. Building a system for which the UN’s existing system of choice literally with your family, couple tablets

14:29
Based on the training of our house that says a nice a little bit more like that. What we found was about three to five minutes, engaging with your nieces and maybe camera anyone that provides a very different sense about data and ethics than what we typically get typically the question about artificial intelligence seems to only have two answers, yes or no. Can we allow the abuses of artificial companies abuse of power governments have used that authority yes

15:23
To other abuses genuine, should be the answer. To these questions and can never be less so disgraced ourselves in our own that we can trust. We have been trusted to use our reasons why out into the market. But surely we can see that it’s possible to use data and artificial intelligence to change.

16:08
And let me say both of those programs so maybe you’re a merchant.

16:12
I, nowadays, get to pay other people to do that. Too expensive, actually theatre constantly affords me to write code, but I do occasionally just program. In fact, going into a program or sponsoring, think about what I’m about to say. Artificial Intelligence is one of the most powerful tools developed in the last 15 years. But it can’t magically solve problems. We cannot find the solutions for ourselves. There are only ever messy human hours, and they only ever have messy human solutions. It doesn’t matter that machine learning is based on hard numbers, or that the data is some object objective external resource that we bring to bear. At least at this point.

17:16
Exploring the unknown is still an issue.

17:21
And what’s amazing about being in a program using massive datasets which have been able to deal with many of my companies even nowadays, topic work is there really can allow you to take a good idea and reach so many fundamental changes. Now let’s have a smash, even if it can. Worms is an application, exclusively that AI can never be there’s your transcript for some reason is half an hour report.

18:04
No longer times in society transcript

18:10
So I know the chance to have a transcript that’s been accurate kind of timing here a little bit by here because we got started by touch eight happens but I’m going to show

18:20
For some students, those things might work. I’m not going to do it’s actually fun.

18:34
Right in the Data

18:36
Training machine learning. You can actually look at what aspects of faces and that was all play.

18:45
We have other things or less all the time. So

18:50
I wanted to understand it. Turns out I can build artificial intelligence I can analyze just a few segments. That is a powerful thing,

19:08
Even if maybe there isn’t as good as human, economics fabric. Twitter’s looking at three things on the recipe. Enable your school and your last job. I think we can all guess the name and it’s pretty cool we can be I do something that when I was the chief scientist and as I said it was the first time that machine learning at higher rates. We built a dataset of  22 available data but the amount of data flowing find out because I don’t think many people appreciate how much they have looked online as the storage of this data allows everyone to opt-out transcriptions for you. Put it on and then we add all those nouns on there. Not the time. Again, that’s really easy. I loved your name, your school. Your last job. I know your entrance gets the best kind of just the best transcription. Turns out that’s a very good turn out in fact, once I know what the names are they were able to easily you maybe get into Stanford is because it’s an interesting, true success is not self do. Because it’s the perfect document to stamp first. In our analysis, first in the world at every major university and the number of different things to YouTube and Stanford is a little extra investment doesn’t get damaged from plastic behind them. Let’s see. Stamp for food for everybody in the queue much more national samples. Can you believe it was essentially true social skills class turns out just as being able to understand other authors is the heart of your school? So I’ve used the video as valuable as it’s a degree as we saw this science conference and being done understand more of what makes these turn out to be things that are learning to create software and retails designers. And actually, as it turns out, for little kids that grow up long. Happy Girls is fascinating data science and sees fundamental truths that have nothing to do with your chapters or race that are predictive. of what it means to pursue it treasurer. Today I had the opportunity to face 20 years working in facial recognition algorithms to get a job or at least an interview and also working hours I developed the first artificial intelligence systems work hard

24:08
Every time someone says so much everyone every center someone shouldn’t be that person wanted us to call me

24:20
In five minutes then that’s a building for that one person to do the right thing.

24:27
When there’s so much money and your reputation right. a multi-billion dollar here. All you have to do is don’t tell me exactly what to do. It’s for their own good. You’ll be up for now.

25:01
That’s ideal, but you can’t make mistakes. No. Courage isn’t just something you have to suffer. You have to practice those moments you need to prepare. It takes all the traffic to YouTube that we don’t want. We don’t want it to go to YouTube. We want it to be on your website to see what your work is or whatever it is to them in its own tape that you’re trying to solve rather than just your the wall passes at a problem and never do we Amazon and Amazon famously force it if they’re hiring an entire woman. I guarantee you, not racial minorities. And just record Thank you embedded drama two it is in some ways YouTube video for sure Mark were interested has made a series of increasingly disgusting and even dangerous candies toggles or decks are just as confusing?

27:58
Yes, so they can’t do anything about

28:09
Ingredients wonder monosodium glutamate and the police officer says there should be massive warning label warning next hires Dennis for any tech company they told you how they know which company work about selling to go to have to have

29:00
The chance to have many elsewhere reduce suicide risk or understand hearing loss. People don’t go into jobs. They’re the villain in this very quickly. It turns out that these algorithms even removed a bias at least in theory, even doing exactly what they’re supposed to and no longer serving.

29:36
The person whose data so in that sense Thank you. I’m heading towards our time here but to actually make that sense. I wanted to try something very.

29:51
I mentioned that happened. Don’t necessarily now we will not add them to your salary. Or don’t compromise your website if you’re curious is just public health. Turns out there’s a much better business and you probably selling appeal for heart attacks or better yet, selling heroes for sexual dysfunction. We’ve been decimated one of the gates. There have been probably 40. Juicing accidents on us. They want to see bags of diet and exercise in hotels. They live in their name through areas of healthy people that go you want to make a change. That’s where you make for someone’s it’s possible so to that end, remember specifically the relationship between chronic stress and slowing heart disease, absenteeism at work, diabetes for bipolar disorder major depression left accumulation wherever you might take time do you want to be true that it is amazing and helpful things we can do about it was in the video or something before he can make a YouTube you have to collect a lot of agency whenever he collects a lot of data you’re doing all almost is present at all systems. aspirants look at it and oh my goodness, I can’t know what to do. If you only tell you there’s no this case, copy-paste trust this is a NumPy array that we’re spending our own money to it will be entirely independent of our own data as a trust for themselves to the trust in other words voluntarily did me because that’s not actually the value that we’re trying our best to live longer have anything about a political consultant can better target voter registration drive or consumer goods companies and figure out what kind of work maybe if it’s something worse. Why don’t we just take it completely right? If we believe that this is what we believe that it is so that people have insights into themselves

33:34
And what they that I would personally hate to see happen. Even if it never made it then surely I don’t need to do anything. So this is what we’ll do. Turns out oh my goodness, what a pain in the ass. This is turning up to a gigantic lunchbox of what you can imagine you’re trying to build something

33:59
To the entire infrastructure. out there that was nothing, Steven

34:08
Is entirely complete.

34:10
We want to share the data out. Never claim that it doesn’t matter. I just bought this is for Facebook doesn’t matter. Just do it. You should not figure

34:34
Out why you should own it. And that trust can be arranged in whatever way maybe that trust actually chooses to give that data away at barriques Maybe it only allows scientists no one else that makes judgments against whatever works. I actually don’t think there’s no right answer. Looks like there should never be a yes or no and that these actually love the idea that trusts themselves can be granted to many different funds as well what works in training.

35:10
We will start with how most of our users will decide how that works and how that data have also been signed.

35:24
And if we can change

35:27
Health and wellness and people will continue to trust the score because that’s the way I look forward to playing around the world. To get my start down, my main job is incredibly challenging when she doesn’t know what to do what the job is from burning nibble profound Valley has a scholarship to accomplish your highly in 32 Explore going tonight on what are we going to launch what we don’t hear from them again, you didn’t want to get out how do you accept that? I don’t do anything for my customers. That person makes the second set of things better for that turns out people that make more sacrifices during their lives actually, like a paradox. end up getting the daily structure in order for it to work up small The second was around that idea discovered data analysis of 500,000 employees and then generalize from their data and AI is a wonderful thing and emotional followers data for more than ever but it’s what you need in your mind. That’s really ready to send this message. When it comes time to send your time back to have letters serving the comments to the whole world. This depends on a lot of files, so many different ways that he comes back on that and I’ve never even one go ahead Absolutely. So hopefully because I did it all. You want to take get involved when I was a really friendly person and she would always do that. But here’s my recommendation. He was a huge fan of logic. I never saw any of China to kind of build websites so that for networking is really the thing amongst the most prepared doctors but so much about the realities they engage in trying to solve. So hand in hand will find that opportunity to learn to code, learn probability theory. Oh my goodness. Well, the man’s and I get to jump in and learn about brains or whatever you want. Just learned how to find something. You’re willing to make a sacrifice and use those tools to create a change. Don’t be afraid to come back to the pay battle because it’s a lot should do this because I will go back to the table it like everything in life will surprise I have to say that was one of the big things I have to learn over the course of my life. I thought those things because setups because I needed to show people how smart and that light lit literally led me to a life where I didn’t even know where my next meal was coming from.