Skip to main content

tv   Bloomberg Business Week  Bloomberg  May 19, 2018 3:00pm-4:00pm EDT

3:00 pm
♪ >> welcome to bloomberg businessweek. jason: we are here at bloomberg headquarters in new york. carol: this is the sooner than you think issue. >> a deep dive into the future of technology. >> it's in cars, and your computers, and boats across the ocean. >> and we take it to our editor with a preview. >> we want to go deeper into the world ai is making, what the big
3:01 pm
grail quests are our reporter went to switzerland. they spoke to the godfather of ai who has been written out of a lot of the history's. >> often at bloomberg, we are talking about ai and machine learning, but you guys dig into some unusual suspects. >> yeah, we tried to look at parts of the ai, folks in china who are making the default chips for bit coin miner rigs. jason: why is it that there are so many interesting characters here?
3:02 pm
what draws people to this field, ai seems to attract a motley crew. >> for sure. what we cap seeing is that a lot of these folks have been working in the background for as long as 30-40 years. the baseline theory behind ai, known as neural networks, designed to mimic human thought with big schemas of computers, it's an idea that's only been practical in the past 5-6 years or so. a lot of the guys who have done the baseline research. carol: what are the conversations you have had getting it to this point? >> we started with a handful of
3:03 pm
stories that seemed ai focused. in a larger package that would have been broader looking. at a certain point, we realized we had so many deep stories that it made just as much sense to just spackle in the gaps and more ai centered issue. jason: when we want to understand coded things, we turn to paul. carol: i love this guy. remember, he did that wonderful story about coding, explaining programming to the masses. here's what he had to say. >> you take a whole lot of data, like a billion pictures and captions for those pictures, and you teach the computer patterns. and the computer will say this has a shape that is like a cat, the word underneath is cat, and it does it a million times. now you give it a picture but no caption.
3:04 pm
and you say what do you think the caption is? it'll go back to the database that it builds, and it will say it looks like an airplane. sometimes it is, sometimes it isn't, but the models and statistics have gotten better. they can say with some confidence that is kind of an airplane. that is when you hit reply in gmail and it says here are some replies you might want to use, that's because they have looked at civilians of emails and -- bazillions of emails and replies. and email before got a reply that was like this one. the first time you see it, you think what are you doing, then you think it is useful. jason: you have seen this of all over a longer -- evolve over a long period of time. where are we right now? >> we are in a moment where it is viable for the first time. it has always been an expensive process, and there is the sort of amazing confluence where's the cards that you put into your computer for graphics processing
3:05 pm
like if you're into 3d gaming, turn out to do that sort of recognition task. they are really good at that because they can split it into thousands of discrete processes. the processor in your laptop, too slow. on a typical web server, too slow. but the processor in a 3-d card is really fast, so you have all of these cards sitting around, jason: you have been sitting -- jason: you have them sitting around, we don't. [laughter] >> you can do this all of the sudden. without spending millions of dollars on specialized hardware. now what is happening is that places like google are building their own processing unit. and you can rent them by the hour, they're more expensive than regular cloud equipment, but they let you do that kind of task. carol: is it becoming more
3:06 pm
productive because you have the technology that can do this processing, and you've also got companies like google which have so much information. he put them together and it is powerful. >> they've got the expertise and the channel. they say, wait a minute, we have billions of emails so we can learn from it. and then we can make a product out of that were people click and reply more quickly. if they like that, they will keep using gmail. give us more data, and we can learn more. where it gets more serious is google maps and self driving cars, google maps is at some level a way of understanding and perceiving the world, they can train themselves and look for patterns. jason: and so you essentially applied machine learning to your life? tell us about this mini hack you did. >> i was on a tight deadline and people said we need a machine learning in the issue, and i said great. i have wanted to do this for a while. i'm going to give it all the writing i've done on the
3:07 pm
paragraphs of my sentences. i'm getting paid by the word, so i thought this would be amazing. it turns out that that is nights and nights and nights of computers running. it was clear that it would be 3-4 days, and again i'm on deadline. before i could get any results at all. so instead, i just sat at my google calendar and try to see if it could generate meetings for me. carol: did it work? >> kind of. you got there a little bit. the first few passes, it was nonsense, but by the end, it was like drinks with gina, and i said, sounds about right. it's more about my life being a little boring, and less about programming. i would just running two scripts. carol: so what changed about your thoughts about machine learning, going through this process? >> i knew it was a big deal and clearly it is guiding the
3:08 pm
strategies of the big whatever's. not just apple, microsoft, and google, and amazon, but also like uber. and land o'lakes. agribusiness is very interested. jason: like the butter. >> yeah, agribusiness is very interested. they wanted to crop productions. for me, what changed is that it is more accessible. google released a huge open package which is become the standard. there is a way to learn it now. carol: they are sharing that? >> oh yeah. they want people coming into the tent. they need engineers and a community and ecosystem around this stuff. jason: ai is obviously a lot about science, but art as well. carol: writes, this meant use artificial intelligence paintings.
3:09 pm
>> we really wanted something that could show you what these deep learning networks are capable of, and there are lots of technologies around for teaching them how to make images. we reached out to this guy who taught his ai how to do landscape paintings. he had us do the cover. jason: so this is painted by a machine? >> yes. he shows the machine a bunch of paintings, which trains it. carol: that's what it is all about, you feed in machine a lot of information and get a result. you can apply it to art. >> what you usually think about our data sets and analyzing numbers, but beauty is a whole another picture. >> and the ai will change how sort of styles the painting. then it changes over time to be more psychedelic.
3:10 pm
you can see the sort of brushstrokes come out in a weird digitize way, it's interesting. carol: up next, big ambitions for the world leader in crypto currency mining. jason: and a company that says they can solve uber's car challenges. ♪
3:11 pm
3:12 pm
3:13 pm
3:14 pm
3:15 pm
♪ carol: welcome back. i'm carol massar. jason: i'm jason kelly. you can also find us online. carol: and on our mobile up. -- app. now we have a pair of stories by max chafkin. we really turned him loose on a lot of different things. carol: what i love is that he uncovers people and companies that nobody is talking about, but in a few months everybody will. jason: and everybody will be talking about bitmain. carol: they want to be a leader in artificial intelligence. >> they are a beijing-based semi conductor company. not super well known unless you
3:16 pm
are in the world of cryptocurrency, where they are basically the only game in town. we are talking about a market share around 80%. they are selling the most popular chips for bitcoin mining and operating a great percentage of these mines. what is interesting is that they are are now moving into ai. carol: they deftly are looking down the road and at the united states. >> they are in a great position because the prices for these chips move with the price of the currency. if you can make more money mining the currency, they could have more revenue, so they are in a perfect position and are trying to diversify. one aspect of that is open mining operations outside of china. at this point, most of bitcoin mining in terms of large-scale stuff is happening in china because of low energy. in rural china, you had historically coal-fired
3:17 pm
electricity for little money, and bitcoin is all about cheap power. that's where the action was. they have got projects in canada, switzerland, and a couple in the united states. we have relatively inexpensive electricity here. another part is the ai push, which is interesting on two levels. one is psychological and the -- technological and the other is regulatory. the chinese government has been pushing ai really hard. china wants to be a powerhouse in this world. and people are starting to say maybe we had a new cold war over ai. so they are going into this industry that is kind of in a gray area. you never know what the regulatory landscape is going to look like to one where the air is full throated support from beijing. jason: and when they are getting into the ai business, they're
3:18 pm
getting into a business populated by some big tech names. >> yeah, the chip they unveiled is named after a well-known chinese science fiction novel, i think all the three body problem. it is very similar to what google has done. google, last year, introduced a thing called a tpu, basically a chip that makes machine learning more efficient. you can only access it if you're a subscriber to google's cloud services. it means you can't just go out and buy one, but it also means you can't buy one in china, because google cloud is not available in mainland china. so bitmain has a niche products. jason: so you can't by the
3:19 pm
competing product. >> absolutely. to be clear, they think there is a market for this worldwide, this is not some kind of copycat, at least as far as they are saying. they were developing it long before google made their announcement. but does have the sort of built-in advantage in the home market. jason: this week, max takes on the topic of autonomous driving. carol: he takes us to this company which is very involved in the space, and how they could make self driving cars safer. >> mobile i is a jerusalem-based startup that does driver assistance systems. if you've been in a car over the past couple of years, these are systems where if you are about to hit something, it will slam on the brakes for you or handle adaptive cruse control, like cruise control, except it slows the car down if the car in front of you slows down they will also
3:20 pm
warn you if you're about to swerve out of your lane or falling asleep. it is a very large company that very few people have known about until now. there's something like 27 million of the systems on the road, all of the big automakers are using them. and last year, intel purchased the company for around $15 billion. jason: they are a part of a much broader trend that you get into in this story around autonomous driving, driverless vehicles. and yet there is a twist in terms of how they are approaching it. >> this is the driverless car company you have never heard of, but also the one in the lead. google, gm, and uber, have all attracted tons of press doing these very sophisticated systems that are driving around. and you know, we are talking
3:21 pm
that dozens of cars. whereas waymo has dozens of cars -- hundreds of cars, and their idea is to leapfrog these companies and release a design that can be used by other automakers. we are talking about bmw, fiat chrysler, a chinese company called neo-, that would be able to offer what waymo is doing. carol: different can be better, different can be less expensive, but is it better? you talk about that unfortunate accident that happened with the uber driverless car. >> this is one of the things that make driverless car scary, is that it is hard for anyone to know whether then one thing is better than the other. the uber incident, there was a
3:22 pm
woman in tempe, arizona walk across the street. she crossed a very busy divided road outside of an intersection and an uber driverless car which had a safety driver, in theory to stop the car, plowed right into her. it didn't even try to stop. in doing this story, one of the reasons i wanted to the story, i want to understand how a country figures out how to stop a car. carol: the founder said we could have caught that, because the way they are doing it is different. >> yeah, he took the video the police released, set it into the company system, which is similar to what a honda is doing with autonomous breaking. he told me that that fatality was avoidable, that this was a sign that the autonomous vehicle industry was getting it wrong.
3:23 pm
jason: up next, tim cook on investing in the. -- in the united states. carol: and a long road to repairing u.s. china relations. jason: this is "bloomberg businessweek." ♪ in
3:24 pm
jason: welcome back. carol: you can also listen to us on radio. jason: in an interview for the david rubenstein show, apple ceo tim cook talked about injecting $350 billion into the u.s. economy. >> we are investing a time in this country. -- a ton in this country. and then yes, we are also going
3:25 pm
to buy some of our stock, because our stock is of good value. from a shareholder point of view, if we can buy stock from people who think it's worth less than we do, then that is good for the company, and it's good for the economy as well. as people sell stock, they sell taxes on their gains. carol: watch the new interview in the new season of the david rubenstein show. coming in june only on bloomberg television. jason: this week, we turned the entire economic section to take on a massive issue. carol: the relationship between the united states and china, we know it has been deteriorating and there is much at risk for both. >> chimerica is a combo of china and america and was coined by british historian neil ferguson
3:26 pm
to express the idea of this symbolizes between the two giant economies. we have china, being high savings and poor, and the u.s. being rich and low savings, and finding that they work together. china would send their cheap goods to america, pretty jobs -- creating jobs for millions of chinese, and while america got cheap goods, and was able to sell high-tech products into china. this is a very optimistic vision. jason: the were coining this is something that was good for the global economy. >> it was sort of necessary. but there were problems, even back then, there were problems. what it did was it ended up helping, among other things, cause the housing crisis, because it was a cheap debt that enables people to buy houses they couldn't afford which has
3:27 pm
global spillover effects. at the time, it was preceded as being a necessary binding, and there were hopes that things would get better. jason: here we are, and how are you feeling about chimerica? >> it is unraveling. the biggest thing was a cte, -- zte, where the company had illegally violated u.s. sanctions by transacting with iran and north korea, so a $1.2 billion fine. but the zte officials did not comply with the terms of the agreement. they secretly paid full bonuses to the people who had been involved in this sanctions busting. when they found out about this, they almost put a death sentence
3:28 pm
on zte, and cut it off from their u.s. suppliers for many years. it became apparent to everyone, including the chinese, that a national champion of china, that the country had been counting on, was extremely vulnerable to the winds, you might say, of the united states. this caused the chinese to think we cannot rely so much on the united states. we need to develop our own advanced technology. carol: but aren't they doing that? to some extent? we see more and more trade with europe and the rest of the world. >> they have been doing it. they have a program called made in china 2025, which involves massive subsidies for a whole range of high-tech, from robotics to electric vehicles.
3:29 pm
the united states has been saying, that seems like an unfair trade practice. you're subsidizing your industries, you did the same with solar panels, you basically stole an american invention, now you've got most of the world market. and so, a trade mission to china led by secretary steve mnuchin came with a message, you guys can't be subsidizing. and then this comes along with zte, and the chinese are saying that if they had any doubts before, they are going ahead. behind the tension -- it heightened attention. it has created more of this unraveling chimerica. jason: up next, the ai entrepreneur. carol: and a new era of the ocean research. jason: this is bloomberg businessweek.
3:30 pm
♪ retail.
3:31 pm
under pressure like never before. and it's connected technology that's moving companies forward fast. e-commerce. real time inventory. virtual changing rooms. that's why retailers rely on comcast business to deliver consistent network speed across multiple locations. every corporate office, warehouse and store near or far covered. leaving every competitor, threat and challenge outmaneuvered. comcast business outmaneuver.
3:32 pm
jason: welcome back to bloomberg business week. carol: still ahead in the sooner than you think issue got the little sail drone that could and a new hacking threat uncovered. first, we're going to go to the author of
3:33 pm
the definitive book on elon musk. we sent him around the world find the key players in artificial intelligence and he ended up in switzerland.'s listen to ashley. >> he's considered one of a.i. godfathers, one of the pioneers in the artificial intelligence field who's really pushed forward a lot of these all seep.s we've the trick is unlike most of the other people who are he'sdered godfathers not really a beloved figure. theseome up with all brilliant ideas, but he spends a lot of time accusing people of stealing or using them without credit and so you know, he's built up this of fearsome reputation in the a.i. world. ubiquitous at a lot of conferences and not just attendee working the rim. >> there's these big a.i. conferences where everybody famousp and he is for this thing, it's a term
3:34 pm
callednow, it's being schmidhubered. giving a presentation, showing off in front of all your peers and this, thisle of german voice appears and this very tall striking guy says i haved he a question for you and basically, he spends a few telling the presenter how they've stolen his ideas or they've kind of bastardized his work and haven't given him credit and ofally, there's a bit back and forth that goes on for a few minutes as he won't let go and he keeps the recordet straight at these conferences so this is why of badthis kind reputation over the years, story,h in my there's a guy who says for a young researcher now, it's of passage to be schmidhubered. >> it seems he's often or wholly right.
3:35 pm
>> he's been at this for about 30 years and has come up with some of the biggest ideas in a.i. the whole reason a.i. memory, thisa kind of temporal ability to store information comes from and he's got countless other things so often yeah, he totally is right, and i think that's things that's most frustrating about it for people is that he's -- on only is he digging in these people, but he's correct. >> where does he go from here? as you say he's been at this quite some time. he's sitting there in switzerland, but he's created quite a little cottage industry around himself. what do we look for from him next? >> he's trying to create something that's known as an agi, artificial general intelligence. so the systems today are still very much dependent on do all this hand holding and instructions on how to work, whereas an agi an a.i. that you just kind of point it in a direction, say go find some x-ray or drive this car and it figures out
3:36 pm
how to do that on its own in ofelatively short amount time and ultimately if this ever happens, which is a question, it would be smarter than human beings by a dramatic margin and so he has this team at this company called nascence, researchers5 that he's brought together and that is their goal is to to make aagi and bunch of money from it and to jurgending take humanity to the next stage where we basically merge with machines. carol: we get more because this time around he's taking richard jenkins. jason: in the middle of the ocean. i love this story because i had thaten know we much to learn about the high seas. here he is. jenkins, i guess i would describe him as an adventurer of sorts and a sailor. he's from england, he spent his whole life building all a periodboats for of 10 years from 1999 to
3:37 pm
2009. to break the world's land sailing record, which is basically where you sailboat on wheels and you go across a desert go asu try to make it fast as you can and in 2009, he did break the record. this wild, crazy it into aturned company. today, he's the ceo of a company called saildrone has -- it's a robotic sailboat that is based on yacht designs and it goes out into the ocean of obtains all kinds scientific data. >> what is he finding that's so attractive to the world at large? surprisedind of during this story. we have very few relatively speaking research ships that theavailable to study ocean. the united states through noaa has 16 research vessels. these boats cost about $150 aboutn to build and $100,000 a day to run when
3:38 pm
they go out and do their research and so -- thereant to stop you because that is an extraordinary amount of money! how in the world does it cost that much to build a ship? >> well, you know, so these ships that are designed to hold dozens of researchers and they're filled with all kinds of scientific equipment. that part costs a lot. they have full data centers surprisedd i was by this, too, but it's just the logistics of running these entire boats and foring people out there four or five, six weeks. each saildrone you can rent for $2,500 a day and the idea is to have eventually saildronesese going out, all over the oceans, all over the world, backconstantly pulling all kinds of information. >> one of the things that people are very interested in, i learned this from your the habits of groups of sharks. why? >> there's about a dozen atearchers who are out this place called the white shark cafe, it's about hawaii andween
3:39 pm
diego. this one spot, every spring, nobody had any idea why they downis and they dive 1,500 feet to 3,000 feet and nobody knows what they were this there and so year, i went with richard to put a couple of saildrones alameda,ter from california, near san francisco. and we sent them off. it took them about three weeks to sail themselves to this white shark cafe, and then they found these transponders that about 37 of the sharks had on them to seen we were able that these sharks are doing these crazy dives every day and night and they're some sort of food supply in the deep ocean. >> and are you optimistic about what you think he'll find and what he'll uncover as this goes dare i say deeper and deeper?
3:40 pm
>> i am. raised $60t million in the last couple of months, which brings them up to $90 million so that's tong to be enough money build hundreds of these saildrones. the biggest thing people is going to be the insight they have around weather. today, our weather models from like a landfill of buoys that are just scattered around and from with theseand saildrones you can sail one directly into a hurricane, you can find out exactly the temperature of the water is in front of the hurricane, which governs its strength, you can figure out how fast its going, when and where it's going to last and globalld have this map of weather like we haven't had before and so the idea is you would sell this to shipping companies, logistics companies, energy companies, but it remains to companies these would actually pay for that kind of information. jason: up next, a whole new be worried about hackers. carol: and the social network used by the smartest kids, and now, it's being used by recruiters.
3:41 pm
jason: this is bloomberg business week.
3:42 pm
jason: welcome back. carol: you can also find us online at mobileand on our app. in the technology section this week, a different scarier side of the tech world. new threat, if you will, and possibly thousands of computers at risk. here's reporter jordan robertson. >> so we're writing this week about a researcher. a cybersecurity researcher from portland. he used to have a top job at security. hacking their
3:43 pm
microprocessors and looking for advanced threats there so what he has found is this. his research shows the meltdown and spectrum januaryilities from which affected basically all of the world's processors, he's able to go a level those using vulnerabilities and what that means is basically what he's saying is those vulnerabilities were really bad and it gets worse. your computer hardware is than you know. >> you're talking as jason said to the core. really going into the core a computer works, you know, enabling it to shut down. deep >> that's right. when we think of computers we tend to think of the stuff we interact with, the theating system, applications. that's not what he's concerned about here. about is concerned the firmware. most people won't know what firmware is, but it's the exists inside the chips on your computer. like the chip doesn't do anything without some code it and what he has discovered here is a way to hardware exploits to get inside the firmware
3:44 pm
of your computer, of your processors and that's where the secrets are stored. >> and you mentioned jordan that he worked at intel and said intel security, i want to make sure everybody understands we're talking about intel like ding ding ding, the company that has all of these chips inside so those areters and computers in personal computers, they're chips andde corporations they're chips inside the government. so how is the industry responding to this next level threat? whenre so in january, meltdown and specter were announced, the industry had a botched roll-out. a new type of threat that they weren't used to responding to so there were a lot of glitches. of problems lot with the industry's response. you know, so what's happening is you've got a lot of different players to fix their systems. it's not like a typical vulnerability where out aoft can push patch, you update your symptoms and you're in the clear. firmwareomes to vulnerabilities, you've got
3:45 pm
to have touches and dozens of companies that have on motherboards, they need to update their firmware before attacks like go away so you can just imagine the scale of the problem. this is not going away. carol: what's interesting, too, jason mention the government. intelligence services, hackersknown that have been able to access firmware for decades. >> this has been the open secret, yeah. to anybody from, you know, any of these intelligence services, they've been hacking forware for years, decades even so it's kind of amazing to me that the cybersecurity industry writ otherhas focused on stuff. they've focused on easier to detect attacks because companies have asked for. they say we don't want to get phished, we don't want spam e-mails so naturally the industry has gone there. how do we block spam, how do phishing. when you talk about hardware you're talking about very hard to detect attacks. companyere's the you probably don't know about, but you would if you're in college potentially. a smartnd you're college kid, you would know
3:46 pm
about piazza. a social network for homework help. that kidswebsite use to help study for their classes and it's computer ofence, but also sort other technical fields, math, science, engineering. studentsa way for to ask and answer questions sets.eir problem and it's usually done under the supervision of the professor --a if a professor realizes somebody is answering a question wrong, he'll steer them straight. that's a good question or a good answer. they claim that students are spending over three hours every night on this website. so think about it. goldmine because what do companies need now? they're desperate for technical talent and here's a website where they can -- it's like fishing in a
3:47 pm
lake and so they're using it now. got the the company idea of putting students who canin, you know, you opt out if you want, but if you opt in, the companies you andact interestingly enough the companies that are most contacting the students are the ones that are sort of like not top of mind for the yet they need technical talent. jason: so before we get to the companies, i want to go beginning and how this started. it seems like a no-brainer yet, and yet it didn't exist. woman was born in northeastpart of india, lived there until age 2, her father was a physicist, brought the canada and then the u.s. so she spent a lot of her formative years in the u.s. but then age 11 india, same
3:48 pm
poor part of northeast bihar. also dealing with the very sex discrimination and so on of india, it was hard to get yet she a woman and aced the exams to get into the india institute of technology, super bright, of and when she got there, she had a problem and were all studying with each other, but she had nobody to study with because were only like two other girls in her whole veryam and so she felt isolated and it was much harder for her because of that. it helps to study with other people. so she came to the u.s. and worked for a couple of big oraclees, including and facebook and went to stanford business school and there she said i really want to help girls like me, who have gone this.h what can i do? she created a website named piazza.
3:49 pm
>> and it's not just a small little website. it's free. you're talking about two and a half million users, eyeballs, that stay there several hours a day. >> it started with just, you few professors that she tried it out on, but in 2011 she opened it up to all comers and it just took off. like 98% they say of the computer science students of american0 universities are on this website. jason: that's unbelievable. go back to the companies, how do they get in? access pay a fee, fee. and that -- again for students who opt in, they can find out well, what courses they've taken, what specialties -- for example, machine learning, i need someone in machine learning. graduating 2018, yes. boxes.hose carol: up next, the data scientist helping create robots. jason: this is bloomberg
3:50 pm
3:51 pm
carol: welcome back to "bloomberg businessweek." jason: you can also listen siriusn the radio on boston,d on 161 in 991fm in washington, d.c. area.960 in the bay carol: and in london on dab mux3 and in asia on the app. time to check in with our bloomberg editor in chief to what -- stood out for him. >> we also talked about and a little bit about tom wolf. here's what he had to say. thisen we put all together this time we noticed one theme that we andly wanted to capture own on the issue it was artificial intelligence. it's changing everything. money, driving cars, just upending business as a
3:52 pm
whole, and it's freaking people out, which is one of the things we wanted to have on the cover. jason: and you really tapped some amazing writers to bring this to us. ashlee vance, his work was amazing. outsourced it to ashlee and he's such a really great tech voice and was able to go places that people don't really know about, which one of the things that we put in an was the oral history of artificial intelligence, which lo and behold canada this major driving force behind a.i. for years. justin trudeau is talking about did you know mecca forthis a.i.? it's incredible. jason: so going beyond sooner than you think, what else really jumped out to you in the magazine this week? >> we've had this ongoing theme of trying to talk to leaders and yet again, this week we got another one which is from turkey and everyone's eyes because erdogan who's been decadeor more than a now has really strengthened his grip on power and we had
3:53 pm
exclusive interviews with him in london this week and it framed everything because the capital markets are taking a look at what's in turkey and they're not liking the influence that he's exerting government, and i think markets are going to be really attuned to this in the coming weeks. carol: everybody should go to the website and check out interview because was the really worth noting. we also can't forget mr. wolf. >> so ripe tom wolfe. he's one of the reasons that i got into journalism. game changer for journalism going back to the '60s and '70s and into '80s, creating this thing called new journalism so he was really like a north star for me personally reasons i got here, but, you know, the reason that wall street remembers him is this seminal work that he did in the '80s and it was really a zeitgeist of the '80s. in our finance section we put a little white hat. in white. man check in on we
3:54 pm
token valley related harassment and diversity. interestingly she looks at a.i. from a different perspective. a.i.: taking a look at and bias and the criminal justice system. it's a really interesting story. ellen.ot more from >> she's been pretty well known among people who are fairnesshampioning in machine learning and a.i., but she became kind more wellow, known last december when she abouta medium post some of the harassment that she had experienced in her career in academia. blush people might think well it's a machine, how can it be biased? goes back to the data sets that humans these models and predictive models that they then apply to real life settings. so, for example, she and her coworkers a few years ago published by a predictive policing company. it's been around for a few i've covered them in the past. and this company makes a
3:55 pm
helps police who are out on the beat figure out which areas in a are most likely to have crime so it's not people, it's not quite like minority similar,ut it's you know. it will tell you maybe you should go check out the intersection of first street a. avenue at this time of day and see what might be so a lot ofere that output is based off of police records that maybe the police department has been gathering for a long time and so she and her coworkers looked at two things, you know. they looked at arrest to drugrelated crimes in oakland and compared them to public health data about who's most be actually using drugs in a similar area and just doing that comparison they found already police enforcement is disappropriationly focused on communities of color. you see there's more criminal activity there the usage. and then when they ran their
3:56 pm
was supposed to do this predictive output, was found that it telling cops also to go places where they would only be increasing and sort of amplifying the bias that was already shown in police records. see that even though people tend to have this idea of computers as fair, swayed by the same things that humans are, the reality is when you use from humanomes activity, such as policing, it's really hard to get past thathings that made data maybe askew in the first place, especially if you don't have a complete set, like if what you're working off of is data made and human >> so what are her recommendations to fix it? said and i had never thought about this before, what's actually hard is not thatncing people algorithms could be biased, but figuring out what people agree on is fair because this is a question in philosophy, as well, what is
3:57 pm
fairness? people tend to disagree depending on the context. it becomes even more have toted when you be as specific as saying we're making this algorithm and the output is going to be capital f. fair. people want the same things, but they disagree on how to get there. carol: "bloomberg businessweek"e on newsstands now. aton: and what was your must read? carol: you've got to read the whole issue if you want know anything that's going on in the world of technology and what's to come. outside of that, really enjoyed talking to peter coy about his story about chimerica, the interdependent relationship between the united states and china. it's a flawed relationship risk.ere's a lot at jason: my must-read, i can't get enough of ashlee vance. character, he wrote the book on elon musk. he found a guy in switzerland of all places, introduced a new verb into my vocabulary, getting schmidhubered. carol: it's a great story.
3:58 pm
jason: more bloomberg television starts right now. mom you called?
3:59 pm
oh hi sweetie, i just want to show you something. xfinity mobile: find my phone. [ phone rings ] look at you. this tech stuff is easy. [ whirring sound ] you want a cookie? it's a drone! i know. find your phone easily with the xfinity voice remote. one more way comcast is working to fit into your life, not the other way around.
4:00 pm
announcer: the following is a qs >>.rogram for moneyfa welcome to a brand-new edition of money. more and more, what would that be? >> i think the economy is doing well>>, and it has good financials but the one word i would use his uncertainty. >> i am worried about my 401(k). we got to stop the spending. >> everything is a big


info Stream Only

Uploaded by TV Archive on