new toy or gift helps foster kids feel special, play and even learn. bring any new unwrapped toys and gifts to any mancini sleep world and make the holidays a little brighter. learn more online at sleep world.com. whether you've been hurt in a car, motorcycle or trucking accident, what you do next matters. make sure to call sweet james accident attorneys. every month, my firm and i help ndreds of people get the medical treatment they need and the financial compensation they deserve, all without paying a dime out of pocket. recently, five out of the top 50 settlements came from sweet james called the firm awarded best attorneys in america. call sweet james. raj mathai. moving the bay area forward. why is this woman wearing a mask? startling new research into the way machines see us or don't see us with doctor joy buolamwini of mit. plus factory in a box

franchising, manufacturing the way mcdonald's franchises hamburgers. helix ceo steve peg and it's an 11 ton time machine. new discoveries about the universe from a special telescope measuring dark energy. cosmologist alexei leto. that's this week on press here. good morning everyone. i'm scott mcgrew. there's been a lot of talk and worry in both movies and real life about how ai may someday hurt us someday in the future, but my first guest worries it's killing us now. just slowly. doctor joy bolognini, seen here giving a talk for tedx, has been called the conscience of the ai revolution by fortune magazine. inc. magazine says she's about as famous as you can be for academic work. doctor bolognini is part of mit's world famous media lab. she's advised the apec conference, including president biden on ai. she's a rhodes scholar and a fulbright

fellow, and she's the author of a new book, unmasking ai my mission to protect what is human in a world of machinenes. good morning to you. i want to start joy with the mask. we saw a brief clip of it in your ted talk. it's on the cover of your book. what is the significance of the mask? yes. so the mask is really basically part of my story of how i even started questioning ai systems when i was a graduate student at mit media lab. i was so excited. i was at the future factory, and i was building a cool class project, and this class project happened to use face tracking technology and unfortunately, the face tracking technology i was using literally did not detect my face until i put a white mask over my dark skin. and it was this moment of putting on a white mask to be made visible to a machine that i started asking, wait a second, are computers as neutral as i

assumed that they would be? so for me, this coding in a white mask incident is really what starts my research and eventually leads to the formation of the algorithmic justice league. and it's also this metaphor of the way in which we hide ourselves or conform ourselves to fit mechanized systems or systems that weren't always created with us in mind. i'll ask more about the justice league in a minute. the mask, though, is so incredibly symbolic, and then your research found that that that dark skinned, dark skinned females, for instance, were misidentified 34% of the time and light skinned males, if i have this number right, 0.08% of the time. decimal zero eight. no, the numbers are stark. and to clarify just a little bit, my research focused at the time on guessing the gender of a face. and so i tested gender

classification systems from microsoft, from ibm, from chinese company at the time, face++, and then later on amazon. so depending on the company, you got different results. some companies actually had perfect performance on lighter skin men. right. and the worst performance we saw. good. i'm glad. i'm glad the white man got to get even better than 0.08 because i was. yeah, that was even perfection in some cases. and with darker skinned women, when we actually went to the darkest skin based on the fitzpatrick classification, we got numbers closer to 40 plus error rates as well. and so the numbers you shared are more so averaging. but to give you a perspective and also it wasn't just darker skinned women. all of these systems perform better on male labeled faces than women's beled faces. in aggregate and in aggregate. they

all perform better on lighter faces than darker skinned faces. so i assume this is the sampling that the computer is drawing from that it just has a limited sample that it hasn't learned enough. that's a great point that you're bringing up. why is this happening in the first place? i was surprised i'm at mit. this is supposed to be the epicenter of innovation. get it together. what's going on? and so what i started doing was looking at data sets, because when we're looking at ai systems now, the ones that are really common, oftentimes we're looking at ai that's built on machine learning. so machine learning techniques work by like the name says, learning from data, learning from examples. so we saw that if training data wasn't very diverse, you get a very limited view of the world. so if you only saw faces that looked like people from nordic countries, you might think those are mainly what human faces look

like. and unlike humans, where we can look at 1 or 2 examples and get the overall idea, these approaches tend to require many different types of examples. so i kept running into what i affectionately call pale male data sets. i was just like, where's everybody else? like it wasn't. once you start pulling back and i go through in the book, right, i start going through all of these data sets and seeing the distribution. it's like, well, no wonder, because for our benchmarks, the way in which we say, how well are these systems working, the gold standard benchmarks, some of them had less than, let's say, 5% of women with dark skin. so you could still do really well overall. and all of these issues would not be seen. it's because i happen to be working on a random class project that i said, let's question the gold standard, which usually doesn't happen. google, rather famously overcompensated.hey had a text to photograph system in which

you could ask for things like, you know, show me the founding fathers writing the constitution or something. and it showed all black americans. so there's some effort being made. i suppose. i think it's the interesting thing with the gemini situation that you're mentioning, and gemini gate is google has run into issues with ai before and labeling before you had gorilla gate, where google photos were labeling black individuals as gorillas and their approach there was just to label nothing as gorillas, including actual gorillas. so i think we've seen google take extreme measures before. and in this particular case, i think what's really important is that addressing bias and representation is nuanced. maybe you're trying to be subversive, subversive in some way, or it's the cast of hamilton or bridgerton kind of thing where you're intentionally creating a different representation. the intent is

really important. what i also found fascinating with this particular instance is that when people who were used to being centered in representation then had that experience of being marginalized, there was an outcry, there was a pushback, and there was an immediate response. right. how quickly google took down gemini as the type of response we'd like to see when other types of problematic misrepresentations are put out there? i think the other thing i really want to caution people on is to not say, oh, we can't completely eliminate bias, so let's not try. right. i really think you have to approach things like algorithmic discrimination and ai bias as as problems that are chronic. and so this isn't this situation of some sort of infectious disease. you knock it out once and you're good. if you're thinking about this as a chronic problem, it means you have to manage it over time. and with the right structures in

place, you can keep it in check. but i think we have to come in with that humility that we're not going to get it all right from the outset. we have to make sure we have feedback loops. we need to be humble enough to admit when we're wrong. and we need to know that when it comes to algorithmic hygiene, this is continuous. so yeah, i think, you know, those of us who don't have phds, those of us who don't work at media lab, those sorts of things we need to be aware of these problems as well, because ai is influencing our lives, whether it is facial recognition or it is something we're doing at work. ai is magical, but we should be aware that that the back end of it still needs some work. absolutely. and i think it's important that we remind ourselves that when it comes to ai systems, we do have a voice and we do have a choice. oftentimes, companies will release products and make it seem like fomo. if you're not

using it, you're going to miss out and then everyone's using it. what are we doing? i don't even know. but it's ai and you might call things that aren't actually ai. ai to be part of the hype cycle and all of that. so i think my first kind of piece of advice for everyday people running into ai systems is you can take your time. it's not a situation where you have to jump in head first without knowing what's going on. the other thing i would encourage people to do is to think about asking questions about why we're using a particular system. so within my own research and my experience advising policy makers, decision makers, top, you know, companies, leaders of nations and so forth, what sometimes happens is there is this momentum around wanting to be at the cutting edge, wanting to be innovative without actually checking if the promise of what the technology might do

fits the reality of what it's offering. and so part of the work i'm doing is saying, hey, this is where we're falling short, this is how we're falling short. this is why we're falling short. and most importantly, this is what we can do. that way we can actually realize the promise of ai, but it doesn't make sense to bask in false promises that are hurting people when we can do so much better. and i assume that's the work, as i promised to ask about the algorithmic justice league. that's the work of the algorithmic algorithmic justice league. absolutely. so making sure people are aware of existing and emerging ai harms and risks. that way we can actually take action to address them. well, doctor joy buolamwini, i appreciate you being with us this morning. doctor joy buolamwini is the author of unmasking ai my mission to protect what is human in a world of machines. we're back in. i'm doug hopkins with

brothers by homes. com and i want to buy your house. i'll make you an as is cash offer on your house within 24 hours. whether it's a total fixer upper or in perfect condition. brothers by homes.com is the easiest way to sell your house when you sell your house to brothers by homes.com. there's no fees, no commissions, no banks and no repairs. just go to brothers, buy homes. dot com for your free, no obligation cash offer 24 hours a day. that's brothers by homes.com space. the new frontier for one of my billboards. wherever there is or will be people, there's sure to be accidents and insurance companies willing to go to great lengths to deny claims. well, too bad, because i go to great lengths and i win big. so wherever you are or will be sweet james will be there to go to the moon and back for you. if you get injured in an accident,

wherever that may be. call sweet james been here before or haven't been here at all. your next favorite thing about this place is waiting to be discovered. did you know you can do this? pretty cool right? and you don't want to miss that. you can also ride this and then race on over to do this. and before you leave, you definitely want to see that. welcome back express. here there is a company in lonn that wants to build very small electric vehicles for asia. a modern update of the tuk tuk. those little taxis and cargo containers that got their name, perhaps from the sound their engine makes. the helix tuck can be configured as a passenger carrier for cargo or even as a little pickup truck. and that's cool. but the real reason that i wanted to talk to

steve pegg, the ceo and co-founder of the company, was because of the way he's thinking about manufacture, sort of a factory in a box, creating manufacturing in the countries in which they would sell these little vehicles by selling the physical factory and the knowledge to someone in that country. you know, steve, it reminds me a little bit like mcdonald's. i mean, mcdonald's knows how to make hamburgers. they know what equipment to use, and they it they educate a franchisee. is that a fair assessment? i think actually that's a really good analogy and one that a number of people have picked up on and i think we can talk about mcdonald's that you don't need to be a chef to be able to make burgers. you have a mcdonald's franchise in order to maybe add to your portfolio and to make money. helix is all about making sure that we've got good partners in the right places to service, really growing and challenging markets.

so providing all of that experience and all of those equipment to them makes absolute sense. now, to extend the mcdonald's metaphor a little bit, if i build a mcdonald's somewhere, i find or mcdonald's helps me find a local baker who's going to make the bread that is going to be the bun. are there going to be these sorts of suppliers in these various countries that you plan to sell these factories in a box to? sure. i mean, we can talk about supply chain. supply chain is the bloodline of any manufacturing and any company. really what we are doing at helix is making sure that we design the vehicle in such a way that we can go anywhere in the world, and procure, let's say, 60 to 80% of that supply chain. we call it a monolithic supply set. essentially, it's all in the way we've designed them. and how we bring on board potential suppliers and what expertise that they need, the least

expertise that they actually need, the larger access we have to local providers. now, why not just manufacture these in one central location? i mean, lots of companies do that. detroit might have a few car factories here and there, but if you want a ford, the chances are it came from one of just a couple of factories. that's a really good question, but we are now in the sensitive sustainability age where we've got to make sure that we have got a low carbon footprint with any product that we sell. and in the modern age, it does seem a bit crazy that we are shipping empty boxes around the world. so we're bringing factories into regions where they wouldn't normally have automotive factories. but also we are bringing enterprise into those regions for businesses that that you don't need to be a

chef to own a mcdonald's. you don't need to be a vehicle manufacturer or a vehicle manufacturer to build these vehicles. so it's making sure that we have all of the principle systems in place so as we don't have to centrally manufacture. and of course, we all know about iconic economies of scale. if we can make this system work with our supply chain and distribution network, then it doesn't need to be manufactured locally. well, and you're talking about putting up these factories within 180 days. i mean, that's that's incredibl. well, within helix we see a factory as part of our product offering. so normamally with a vehicle development, we would design the vehicle, we'd engineer, develop the vehicle, and then we would pass that over to the manufacturing for them to either build a factory or put

the vehicle through one of their existing factories. our factories are actually one of our products, so we work within our complex product lifecycle systems to make sure that that each factory is fully optimized and we aren't building a factory that has like, like an installed assembly line. i it's gone are e days of the henry ford line that our factories are more like dropship centers, where there's a lot of flexibility in terms of what what gets assembled, where and which stations. so if we can take out that fixed infrastructure, then that frees us up to be able to shortcut the factory deployment time. and lastly, i want to talk about the vehicle itself. i want to point out you are an expert in automotive manufacturing. you've worked at land rover and faraday. you've even helped build italian supercars. and now

you're building these tiny little delivery vehicles. but that last mile, you know, from the bakery to the store with the bread truck, a particularly in asia that is done by tiny, tiny little vehicles. it is. so when you talk about last mile, you're talking about the last mile in a really long complex. but when you use the analogy of the baker, you're talking about only mile. and i think the challenge is within dense cities. let's talk about asia. you know, there are so many old vehicles out there using internal combustion engines that are addingo much pollution. so this is where we started with the helix is this type of vehicle seems to be the perfect product market fit for those regions. well, steve, i appreciate your time and i wish you just the best of luck with your with your new vehicle. steve pegg is the ceo and

co-founder at helix, and we'll be back in just a minute. if you get sick with covid 19, someone will feel all alone. get an updated vaccine. this season so you don't let down the ones who matter most. last fall, more people were hospitalized from covid than the flu. covid is always changing. protect yourself this season with an updated vaccine. do it for you and them at pg&e. we're adding more safe and affordable renewable energy every day. we're deploying new technologies that will allow electric vehicles to power homes, which helps customers save more money by capturing low cost energy. and we're working on new ways to store power so that this safe, clean and low cost energy is available around the clock. in fact, we've got one of the world's biggest battery storage plants right here in our backyard. explore the future with us at pg&e .com/ innovatio.

accidents happen and when they do, every second counts. whether you've been involved in a car, truck or motorcycle accident waiting to reach out to my winning team of attorneys could be a costly decision. evidence fades, memories dim and vital details slip away. your case deserves immediate attention. don't navigate the legal maze alone. at sweet james, we're here for you. let us fight for the justice you deserve. time is ticking and so is the strength of your case. call now for a free c consultatn. your future depends on it. welcome back to. for us here, cosmologists. a few days ago announced they had completed a 3d map of the universe. more than 30 million galaxies with a precision of. get ready for this 1%, which i think gives you some idea as to how difficult it is to map the universe. i've asked cosmologist alexei leto from the university of california at santa cruz, who

works on the mapping project and soon will be promoted to lead organizer to help me understand what all this means. and she's promised to be patient with me. good morning, doctor leto. let's start with the super easy one. and that's what is the difference between a cosmologist and an astronomer. that's a great question, scott. astronomy typically studies the night sky with telescopes. astronomy then morphed into astrophysics, which brings in physics to understand the cosmos. and then from astrophysics was born cosmology, which is a subfield in astrophysics which focuses especially on the understanding, the origin and the evolution of the universe. and when we talk about the universe just to remindnd us of o elementary level astronomy, you know, we live in a solar system which has 8 or 9 planets, depending on when you went to elementary school, which is in a galaxy, the milky way galaxy, which is in the universe, which is when

we talk about the cosmos, we're and what you mapped, we're talking about all the known places that there are. well, so the universe is filled with billions and billions of galaxies all t the way bk to sort of the beginning of time, 13 billion years ago. desi has created the largest map of galaxies currently mapping about 11 million galaxies. but that's far from being all of the galaxies in the universe. so it's still an incomplete picture, but it is still the largest. it's the neighborhood, right? i mean, it's the local neighborhood. so actually, no, desi doesn't necessarily focus on the localal neighborhoods so much as it focuses on distant galaxies. there are other programs. there's a part of desi that does do the local neighborhood and other programs that do the local neighborhood. but the really the key goal of desi is to map the biggest and brightest galaxies as a function of time all the way back to sort

of the early ages of the universe. and when i say 1% accuracy, what what are you referring to? what what's 1%? yeah, yeah, that's a great question because it's terrible accuracy. so let's maybe let's just say what the goal of desi is first. and i can tell you what the 1% means. so the key goal of this experiment is to understand dark energy. so dark energy permeates space. and time. it's one of the fundamental mysteries of astrophysics. in many ways, it's more fundamental than dark matter. and dark matter is very different than dark energy. we could go on that tangent if you like. and so because dark energy is so mysterious, we're trying to map out its properties and provide information to the theorists so they can try to figure out what it is. and so by mapping out that, we call it the equation of state. it's a number that we call w. and so what desi

has done is measured that number w to 1%. fair enough. now dark matter dark energy not the same thing, but one of them is actually contributing to the expansion of the universe. have i got that right? yes. correct. so dark matter is probably some kind of particle. it was discovered in the 1920s. well, its existence was discovered. we don't know what kind of particle it is, but we have many different candidates. so in many ways, it's maybe not as much of a mystery. dark energy was discovered later in the 1990s and is a much larger mystery because we really have no clue what it is now. dark energy causes the universe to accelerate in its expansion, whereas dark matter would have the opposite tendency of helping the universe kind of contract more so, they have opposing effects. that makes sense. it is exciting that we think we know so much, and then we discover

there are tngs we can't even begin to describe. oh, absolutely. this is a amazing career to have because we discover things every year. every year i teach a 101 class in astronomy, and every year i have new things to say and new results to show. and i'm always surprised updating mydes like, wow, a lot of cool things has happened this year in in astrophysics. so it's very cool. ot you interested in being a cosmologist? right? i, i really have to credit carl saga. sure. and the cosmos, which i think probably 80% of my colleagues would do the same. and, you know, i read his book pale blue dot as as a teenager. and it was just fascinating to me. and i was just really always interested in the big picture. big picture question of the universe. and, you know, where it came from and where we're going. well, thank you for being so patient with me. doctor alexei leto is from the

university of california at santa cruz. we'll be back in just a minute. hey, faithful. you're invited to come watch the game with us for sunday night football in the city. join us tonight at thrive city in san francisco from 5 to 8:30 p.m. rsvp at nbcbayarea.com. slash football night we keep seeing every night the big ticker going up of all the money you save, whether it's work, family, or finances. kris has your back. that's when i said this is not resolved. i'm going to call kris. you responded so quickly. nbc bay area responds with chris chmura gia diamonds lowest prices ever. the jewelry exchange has natural one carat gia just 1990 custom mounts are just 4.99 carat, 50 2992 carat, 5990. thousands of diamonds guaranteed the lowest price. the jewelry exchange redwood city. when my grandfather was diagnosed with mesothelioma, he didn't know which law firm to hire. his doctor gave him one name. his old union buddy gave

another. but my grandfather. he did his homework and found whites in luxembourg. it was hard to find anyone who could match their experience, so the choice was easy. call whites in luxembourg today at 800 885 miso. that's 800 885 miso. set sail for olo hawaiian barbecue and discover the thrill of the uno keiki meal with fresh made to order food and quality ingredients that bring the flavors of the island to your family. disney's moana two, now playing only in theaters. rated pg. that's our show for this week. my thanks to my guests, and thank you for making us part of your sunday morning. tis the season to shop at sebastopol hardware. make your list and check it twice. there's gifts for the not so naughty and nice. keep your cocoa hot in cool yeti tumblers. stay in style all season long with top clothing

brands and enjoy home cooked meals in high quality cookware. sebastopol hardware santas favorite hardware store select sebastopol hardware is yr store at ace hardware. com. been here before or haven't been here at all? your next favorite thing about this place is waiting to be discovered. did you know you can do thi? pretty cool right? and you don't want to miss that. you can also ride this and then race on over to do this. and before you leave, you definitely want to see that. you know what makes my firm different? we focus on you. weocus on telling your story. how this accident has impacted your life, your family, and your future. and when we tell your story to the insurance companies, we tell a story that demands action. when you're more than just a case number, that's when real results happen. results like

over a billion dollars recovered for our clients and a 90% winning record. at sweet james, we deliver results that matter because your story is our priority. so if you or a loved one have been hurt in an accident, call now for a free consultation. if there's one word that describes shag carpet, it'd be selection or showroom because no one has larger showrooms than s and g, or a fleet of mobile showrooms filled with samples. so you can shop at home. ou estimates in-home estimates are always free at s and g or maybe warranty because s and g offers a lifetime installation warranty. no, there's no one word for s and g, which isn't even really a word is it's and g more than carpet where quality is