Placeholder Image

字幕列表 影片播放

  • >> Welcome everyone.

  • Thank you for coming to the Intel AI lounge

  • and joining us here for this economist world event.

  • My name is Jack.

  • I'm the chief architect of our

  • autonomist driving solutions at Intel

  • and I'm very happy to be here and to be joined

  • by an esteemed panel of colleagues who are joining to,

  • I hope, engage you all in a frayed dialogue and discussion.

  • There will be time for questions as well,

  • so keep your questions in mind.

  • Jot them down so you ask them to us later.

  • So first, let me introduce the panel.

  • Next to me we have Michelle,

  • who's the co-founder and CEO of Fine Mind.

  • She just did an interview here shortly.

  • Fine Mind is a company that provides a technology platform

  • for retailers and brands that uses artificial intelligence

  • as the heart of the experiences

  • that her company's technology provides.

  • Joe from Intel is the head of partnerships and acquisitions

  • for artificial intelligence and software technologies.

  • He participated in the recent acquisition of Movidius,

  • a computer vision company that Intel recently acquired

  • and is involved in a lot of smart city activities as well.

  • And then finally, Sarush, who is data scientist by training,

  • but now has JDA labs,

  • which is researching emerging technologies

  • and their application in the supply chain worldwide.

  • So at the end of the day, the internet things

  • that artificial intelligence really promises

  • to improve our lives in quite incredible ways

  • and change the way that we live and work.

  • Often times the first thing that we think about

  • when we think about AI is Skynet,

  • but we at Intel believe in AI for good

  • and that there's a lot of things that can happen

  • to improve the way people live, work, and enjoy life.

  • So as things in the Internet, as things become connected,

  • smart, and automated, artificial intelligence

  • is really going to be at the heart of those new experiences.

  • So as I said my role is the architect

  • for autonomous driving.

  • It's a common place when people think

  • about artificial intelligence,

  • because what we're trying to do is

  • replace a human brain with a machine brain,

  • which means we need to endow that machine

  • with intelligent thoughts, contexts, experiences.

  • All of these things that sort of make us human.

  • So computer vision is the space, obviously,

  • with cameras in your car that people often think about,

  • but it's actually more complicated than that.

  • How many of us have been in a situation on a two lane road,

  • maybe there's a car coming towards us,

  • there's a road off to the right, and you sort of sense,

  • "You know what?

  • That car might turn in front of me."

  • There's no signal.

  • There's no real physical cue, but just something about

  • what that driver's doing where they're looking tells us.

  • So what do we do? We take our foot off the accelerator.

  • We maybe hover it over the brake, just in case, right?

  • But that's intelligence that we take for granted

  • through years and years and years of driving experience

  • that tells us something interesting is happening there.

  • And so that's the challenge that we face in terms

  • of how to bring that level of human intelligence

  • into machines to make our lives better and richer.

  • So enough about automated vehicles though,

  • let's talk to our panelists about some of the areas

  • in which they have expertise.

  • So first for Michelle, I'll ask...

  • Many of us probably buy stuff online everyday, every week,

  • every hour, hourly delivery now.

  • So a lot has been written about

  • the death of traditional retail experiences.

  • How will artificial intelligence and the technology

  • that your company has rejuvenate that retail experience,

  • whether it be online or in the traditional

  • brick and mortar store?

  • >> Yeah, excuse me.

  • So one of the things that I think is a common misconception.

  • You hear about the death of the brick and mortar store,

  • the growth of e-commerce.

  • It's really that e-commerce is beating brick and mortar

  • in growth only and there's still

  • over 90% of the world's commerce is done

  • in physical brick and mortar store.

  • So e-commerce, while it has the growth,

  • has a really long way to go and I think one of the things

  • that's going to be really hard to replace

  • is the very human element of interaction and connection

  • that you get by going to a store.

  • So just because a robot named Pepper comes up to you

  • and asks you some questions,

  • they might get you the answer you need faster

  • and maybe more efficiently, but I think as humans

  • we crave interaction and shopping

  • for certain products especially,

  • is an experience better enjoyed in person with other people,

  • whether that's an associate in the store

  • or people you come with to the store

  • to enjoy that experience with you.

  • So I think artificial intelligence can help it be

  • a more frictionless experience,

  • whether you're in store or online to get you from point A

  • to buying the thing you need faster,

  • but I don't think that it's going to ever completely replace

  • the joy that we get by physically going out into the world

  • and interacting with other people to buy products.

  • >> You said something really profound.

  • You said that the real revolution

  • for artificial intelligence in retail will be invisible.

  • What did you mean by that?

  • >> Yeah, so right now I think that

  • most of the artificial intelligence that's being applied

  • in the retail space is actually not something

  • that shoppers like you and I see when we're on a website

  • or when we're in the store.

  • It's actually happening behind the scenes.

  • It's happening to dynamically change the webpage

  • to show you different stuff.

  • It's happening further up the supply chain, right?

  • With how the products are getting manufactured,

  • put together, packaged, shipped, delivered to you,

  • and that efficiency is just helping retailers be smarter

  • and more effective with their budgets.

  • And so, as they can save money in the supply chain,

  • as they can sell more product with less work,

  • they can reinvest in experience,

  • they can reinvest in the brand,

  • they can reinvest in the quality of the products,

  • so we might start noticing those things change,

  • but you won't actually know that that has anything to do

  • with artificial intelligence, because not always in a robot

  • that's rolling up to you in an aisle.

  • >> So you mentioned the supply chain.

  • That's something that we hear about a lot,

  • but frankly for most of us,

  • I think it's very hard to understand

  • what exactly that means,

  • so could you educate us a bit on what exactly

  • is the supply chain and how is artificial intelligence

  • being implied to improve it?

  • >> Sure, sure.

  • So for a lot of us, supply chain is the term

  • that we picked up when we went to school

  • or we read about it every so often,

  • but we're not that far away from it.

  • It is in fact a key part of what Michelle calls

  • the invisible part of one's experience.

  • So when you go to a store and you're buying a pair of shoes

  • or you're picking up a box of cereal,

  • how often do we think about,

  • "How did it ever make it's way here?"

  • We're the constituent components.

  • They probably came from multiple countries

  • and so they had to be manufactured.

  • They had to be assembled in these plants.

  • They had to then be moved,

  • either through an ocean vessel or through trucks.

  • They probably have gone through multiple warehouses

  • and distribution centers and then finally into the store.

  • And what do we see?

  • We want to make sure that when I go

  • to pick up my favorite brand of cereal, it better be there.

  • And so, one of the things where AI is going to help

  • and we're doing a lot of active work in this,

  • is in the notion of the self learning supply chain.

  • And what that means is really bringing

  • in these various assets and actors of the supply chain.

  • First of all, through IOT and others, generating the data,

  • obviously connecting them,

  • and through AI driving the intelligence,

  • so that I can dynamically figure out the fact

  • that the ocean vessel that left China

  • on it's way to Long Beach has been delayed by 24 hours.

  • What does that mean when you go to a Foot Locker

  • to buy your new pair of shoes?

  • Can I come up with alternate sourcing decisions,

  • so it's not just predicting.

  • It's prescribing and recommending as well.

  • So behind the scenes, bringing in a lot of the,

  • generating a lot of the data,

  • connecting a lot of these actors

  • and then really deriving the smarts.

  • That's what the self learning supply chain is all about.

  • >> Are supply chains always international

  • or can they be local as well?

  • >> Definitely local as well.

  • I think what we've seen over the last decades,

  • it's kind of gotten more and more global,

  • but a lot of the supply chain

  • can really just be within the store as well.

  • You'd be surprised at how often retailers

  • do not know where their product is.

  • Even is it in the front of the store?

  • Is it in the back of the store?

  • Is it in the fitting room?

  • Even that local information is not really available.

  • So to have sensors to discover where things are

  • and to really provide that efficiency,

  • which right now doesn't exist,

  • is a key part of what we're doing.

  • >> So Joe, as you look at companies out there

  • to partner or potentially acquire,

  • do you tend to see technologies

  • that are very domain specific for retail or supply chain

  • or do you see technologies that could bridge

  • multiple different domains

  • in terms of the experiences we could enjoy?

  • >> Yeah, definitely. So both.

  • A lot of infant technologies

  • start out in very niched use cases,

  • but then there are technologies that are pervasive

  • across multiple geographies and multiple markets.

  • So, smart cities is a good way to look at that.

  • So let's level set really quick on smart cities

  • and how we think about that.

  • I have a little sheet here to help me.

  • Alright, so, if anybody here played Sim City before,

  • you have your little city

  • that's a real world that sits here, okay?

  • So this is reality and you have little buildings and cars

  • and they all travel around

  • and you have people walking around with cell phones.

  • And what's happening is as we develop smart cities,

  • we're putting sensors everywhere.

  • We're putting them around utilities, energies, water.

  • They're in our phones.

  • We have cameras and we have audio sensors in our phones.

  • We're placing these on light poles,

  • which is existing sustaining power points around the city.

  • So we have all these different sensors

  • and they're not just cameras and microphones,

  • but they're particulate sensors.

  • They're able to do environmental monitoring

  • and things like that.

  • And so, what we have is we have this physical world

  • with all these sensors here.

  • And then what we have is we've created

  • basically this virtual world that has a great memory

  • because it has all the data from all the sensors

  • and those sensors really act as ties,

  • if you think of it like a quilt, trying a quilt together.

  • You bring it down together and everywhere you have a stitch,

  • you're stitching that virtual world

  • on top of the physical world

  • and that just enables incredible amounts of innovation

  • and creation for developers, for entrepreneurs,

  • to do whatever they want to do

  • to create and solve specific problems.

  • So what really makes that possible

  • is communications, connectivity.

  • So that's where 5G comes in.

  • So with 5G it's not just a faster form of connectivity.

  • It's new infrastructure.

  • It's new communication.

  • It includes multiple types of

  • communication and connectivity.

  • And what it allows it to do is all those little sensors

  • can talk to each other again.

  • So the camera on the light pole can talk

  • to the vehicle driving by or the sensor on the light pole.

  • And so you start to connect everything

  • and that's really where artificial intelligence

  • can now come in and sense what's going on.

  • It can then reason, which is neat,

  • to have computer or some sort of algorithm

  • that actually reasons based on a situation

  • that's happening real time.

  • And it acts on that, but then you can iterate on that

  • or you can adapt that in the future.

  • So if we think of an actual use case, we'll think of

  • a camera on a light post that observes an accident.

  • Well it's programmed to automatically notify

  • emergency services that there's been an accident.

  • But it knows the difference between a fender bender

  • and an actual major crash where we need

  • to send an ambulance or maybe multiple firetrucks.

  • And then you can create iterations

  • and that learns to become more smart.

  • Let's say there was a vehicle that was in the accident

  • that had a little yellow placard on it that said hazard.

  • You're going to want to send different types

  • of emergency services out there.

  • So you can iterate on what it actually does

  • and that's a fantastic world to be in

  • and that's where I see AI really playing.

  • >> That's a great example of what it's all about

  • in terms of making things smart, connective, and autonomous.

  • So Michelle as somebody who has founded the company

  • and the space with technology that's trying

  • to bring some of these experiences to market,

  • there may be folks in the audience

  • who have aspirations to do the same.

  • So what have you learned over the course

  • of starting your company and developing the technology

  • that you're now deploying to market?

  • >> Yeah, I think because AI is such a buzz word.

  • You can get a dot AI domain now,

  • doesn't mean that you should use it for everything.

  • Maybe 7, 10, 15 years ago...

  • These trends have happened before.

  • In the late 90s, it was technology

  • and there was technology companies

  • and they sat over here and there was everybody else.

  • Well that not true anymore.

  • Every company uses technology.

  • Then fast forward a little bit,

  • there was social media was a thing.

  • Social media was these companies over here

  • and then there was everybody else and now every company

  • needs to use social media or actually maybe not.

  • Maybe it's a really bad idea for you

  • to spend a ton of money on social media

  • and you have to make that choice for yourself.

  • So the same thing is true with artificial intelligence

  • and what I tell...

  • I did a panel on AI for Adventure Capitalists last week,

  • trying to help them figure out when to invest

  • and how to evaluate and all that kind of stuff.

  • And what I would tell other aspiring entrepreneurs is

  • "AI is means to an end.

  • "It's not an end in itself."

  • So unless you're a PH.D in machine learning

  • and you want to start an AI as a service business,

  • you're probably not going to start an AI only company.

  • You're going to start a company for a specific purpose,

  • to solve a problem, and you're going to use AI

  • as a means to an end, maybe, if it makes sense to get there,

  • to make it more efficient and all that stuff.

  • But if you wouldn't get up everyday for ten years

  • to do this business that's going to solve whatever problem

  • you're solving or if you wouldn't invest in it

  • if AI didn't exist, then adding dot AI

  • at the end of a domain is not going to work.

  • So don't think that that will help you

  • make a better business.

  • >> That's great advice.

  • Thank you.

  • Surash, as you talked about the automation then

  • of the supply chain, what about people?

  • What about the workers whose jobs may be lost

  • or displaced because of the introduction of this automation?

  • What's your perspective on that?

  • >> Well, that's a great question.

  • It's one that I'm asked quite a bit.

  • So if you think about the supply chain

  • with a lot of the manufacturing plants,

  • with a lot of the distribution centers,

  • a lot of the transportation,

  • not only are we talking about driverless cars

  • as in cars that you and I own,

  • but we're talking about driverless delivery vehicles.

  • We're talking about drones and all of these on the surface

  • appears like it's going to displace human beings.

  • What humans used to do,

  • now machines will do and potentially do better.

  • So what are the implications around human beings.

  • So I'm asked that question quite a bit,

  • especially from our customers and my general perception

  • on this is that I'm actually cautiously optimistic

  • that human beings will continue

  • to do things that are strategic.

  • Human beings will continue to do things

  • that are creative and human being will probably continue

  • to do things that are truly catastrophic,

  • that machines simply have not been able to learn

  • because it doesn't happen very often.

  • One thing that comes to mind is when ATM machines

  • came about several years ago before my time,

  • that displaced a lot of teller jobs in the banking industry,

  • but the banking industry did not go belly up.

  • They found other things to do.

  • If anything, they offered more services.

  • They were more branches that were closed

  • and if I were to ask any of you now if you would go back

  • and not have 24/7 access to cash,

  • you would probably laugh at me.

  • So the thing is, this is AI for good.

  • I think these things might have temporary impact

  • in terms of what it will do to labor and to human beings

  • but I think we as human beings will find

  • bigger, better, different things to do

  • and that's just in the nature of the human journey.

  • >> Yeah, there's definitely a social acceptance angle

  • to this technology, right?

  • Many of us technologists in the room,

  • it's easier for us to understand what the technology is,

  • how it works, how it was created,

  • but for many of our friends and family, they don't.

  • So there's a social acceptance angle to this.

  • So Michelle as you see this technology deployed

  • in retail environments, which is a space

  • where almost every person in every country goes,

  • how do you think about making it feel comfortable

  • for people to interact with this kind of technology

  • and not be afraid of the robots

  • or the machines behind the curtain.

  • >> Yeah, that's a great question.

  • I think that user experience always has to come first,

  • so if you're using AI for AI's sake or for the cool factor,

  • the wow factor, you're already doing it wrong.

  • Again, it needs to solve a problem

  • and what I tend to tell people who are like,

  • "Oh my God. AI sounds so scary.

  • "We can't let this happen."

  • I'm like, "It's already happening

  • "and you're already liking it.

  • "You just don't know

  • "because it's invisible in a lot of ways."

  • So if you can point of those scenarios

  • where AI has already benefited you and it wasn't scary

  • because it was a friendly kind of interaction,

  • you might not even have realized it was there

  • versus something that looks so different and...

  • Like panic driving.

  • I think that's why the driverless car thing

  • is a big deal because you're so used to seeing,

  • in America at least,

  • someone on the left side of the car in the front seat.

  • And not seeing that is like, woah, crazy.

  • So I think that it starts with the experience

  • and making it an acceptable kind of interface

  • or format that doesn't give you that,

  • "Oh my God. Something is wrong here," kind of feeling.

  • >> Yeah, that's a great answer.

  • In fact, it reminds me there was this really amazing study

  • by a Professor Nicholas Eppily

  • that was published in the journal of social psychology

  • and the name of this study was called A Mind In A Machine.

  • And what he did was he took subjects

  • and had a fully functional automated vehicle and then

  • a second identical fully functional automated vehicle,

  • but this one had a name and it had a voice

  • and it had sort of a personality.

  • So it had human anthropomorphics characteristics.

  • And he took people through these two different scenarios

  • and in both scenarios he's evil

  • and introduced a crash in the scenario

  • where it was unavoidable.

  • There was nothing going to happen.

  • You were going to get into an accident in these cars.

  • And then afterwards, he pulled the subjects and said,

  • "Well, what did you feel about that accident?

  • "First, what did you feel about the car?"

  • They were more comfortable in the one

  • that had anthropomorphic features.

  • They felt it was safer and they'd be more willing

  • to get into it, which is not terribly surprising,

  • but the kicker was the accident.

  • In the vehicle that had a voice and a name, they actually

  • didn't blame the self-driving car they were in.

  • They blamed the other car.

  • But in the car that didn't have anthropomorphic features,

  • they blamed the machine.

  • They said there's something wrong with that car.

  • So it's one of my favorite studies

  • because I think it does illustrate

  • that we have to remember the human element

  • to these experiences and as artificial intelligence

  • begins to replace humans, or some of us even,

  • we need to remember that we are still social beings

  • and how we interact with other things,

  • whether they be human or non-human, is important.

  • So, Joe, you talk about evaluating companies.

  • Michelle started a company.

  • She's gotten funding.

  • As you go out and look at new companies

  • that are starting up, there's just so much activity,

  • companies that just add dot AI to the name as Michelle said,

  • how do you cut through the noise

  • and try to get to the heart of is there any value

  • in a technology that a company's bringing or not?

  • >> Definitely. Well, each company

  • has it's unique, special sauce, right?

  • And so, just to reiterate what Michelle was talking about,

  • we look for companies that are really good

  • at doing what they do best, whatever that may be,

  • whatever that problem that they're solving

  • that a customer's willing to pay for,

  • we want to make sure that that company's doing that.

  • No one wants a company that just has AI in the name.

  • So we look for that number one and the other thing we do

  • is once we establish that we have a need

  • or we're looking at a company based on

  • either talent or intellectual property,

  • we'll go in and we'll have to do a vetting process

  • and it takes a whole.

  • It's a very long process and there's legal involved

  • but at the end of the day, the most important thing

  • for the start up to remember

  • is to continue doing what they do best and continue

  • to build upon their special sauce

  • and make sure that it's very valuable to their customer.

  • And if someone else wants to look at them

  • for acquisition so be it,

  • but you need to be meniacally focused on your own customer.

  • That's my two cents.

  • >> I'm thinking again about this concept

  • of embedding human intelligence,

  • but humans have biases right?

  • And sometimes those biases aren't always good.

  • So how do we as technologists in this industry

  • try to create AI for good and not unintentionally

  • put some of our own human biases into models

  • that we train about what's socially acceptable or not?

  • Anyone have any thoughts on that?

  • >> I actually think that the hype about AI taking over

  • and destroying humanity, it's possible and I don't want to

  • disagree with Steven Hawking as he's way smarter than I am.

  • But he kind of recognizes it could go both ways

  • and so right now, we're in a world

  • where we're still feeding the machine.

  • And so, there's a bunch of different issues

  • that came up with humans feeding the machine

  • with their foibles of racism and hatred and bias

  • and humans experience shame which causes them

  • to lash out and what to put somebody else down.

  • And so we saw that with Tay, the Microsoft chatbot.

  • We saw that with even Google's fake news.

  • They're like picking sources now to answer the question

  • in the top box that might be the wrong source.

  • Ads that Google serves often show men high paying jobs,

  • $200,000 a year jobs, and women don't get those same ones.

  • So if you trace that back, it's always coming back

  • to the inputs and the lens

  • that humans are coming at it from.

  • So I actually think that we could be in a way better place

  • after this singularity happens

  • and the machines are smarter than us

  • and they take over and they become our overlords.

  • Because when we think about the future,

  • it's a very common tendency for humans to fill in the blanks

  • of what you don't know in the future with what's true today.

  • And I was talking to you guys at lunch.

  • We were talking about this harbored psychology professor

  • who wrote a book and in the book

  • he was talking about how 1950s,

  • they were imagining the future

  • and all these scifi stories and they have flying cars

  • and hovercrafts and they're living in space,

  • but the woman still stays at home and everyone's white.

  • So they forgot to extrapolate the social things

  • to paint the picture in,

  • but I think when we're extrapolating into the future

  • where the computers are our overlords,

  • we're painting them with our current reality,

  • which is where humans are kind of terrible (laughs).

  • And maybe computers won't be

  • and they'll actually create this Utopia for us.

  • So it could be positive.

  • >> That's a very positive view.

  • >> Thanks. >> That's great.

  • So do we have this all figured out?

  • Are there any big challenges that remain in our industries?

  • >> I want to add a little bit more to the learning

  • because I'm a data scientist by training and a lot of times,

  • I run into folks who think

  • that everything's been figured out.

  • Everything is done. This is so cool.

  • We're good to go and one of the things

  • that I share with them is something

  • that I'm sure everyone here can relate to.

  • So if a kindergartner goes to school

  • and starts to spell profanity,

  • that's not because the kid knows anything good or bad.

  • That is what the kid has learned at home.

  • Likewise, if we don't train machines well,

  • it's training will in fact be biased to your point.

  • So one of the things that we have to kep in mind

  • when we talk about this is we have to be careful as well

  • because we're the ones doing the training.

  • It doesn't automatically know what is good or bad

  • unless that set of data is also fed to it.

  • So I just wanted to kind of add to your...

  • >> Good. Thank you.

  • So why don't we open it up a little bit for questions.

  • Any questions in the audience for our panelists?

  • There's one there looks like (laughs).

  • Emily, we'll get to you soon.

  • >> I had a question for Sarush

  • based on what you just said about us training

  • or you all training these models and teaching them things.

  • So when you deploy these models to the public

  • with them being machine learning and AI based,

  • is it possible for us to retrain them

  • and how do you build in redundancies for the public

  • like throwing off your model and things like that?

  • What are some of the considerations that go into that?

  • >> Well, one thing for sure is training is continuous.

  • So no system should be trained once,

  • deployed, and then forgotten.

  • So that is something that we as AI professionals

  • need to absolutely, because...

  • Trends change as well.

  • What was optimal two years ago is no longer optimal.

  • So that part needs to continue to happen

  • and we're the where the whole IOT space is so important

  • is it will continue to generate relevant consumable data

  • that these machines can continuously learn.

  • >> So how do you decide what data though, is good or bad,

  • as you retrain and evolve that data over time?

  • As a data scientist, how do you do selection on data?

  • >> So, and I want to piggyback on what Michelle said

  • because she's spot on.

  • What is the problem that you're trying to solve?

  • It always starts from there

  • because we have folks who come in to CIOs,

  • "Oh look.

  • "When big data was hot, we started to collect

  • "a lot of the data, but nothing has happened."

  • But data by itself doesn't automatically do magic for you,

  • so we ask, "What kind of problem are you trying to solve?

  • "Are you trying to figure out

  • "what kinds of products to sell?

  • "Are you trying to figure out

  • "the optimal assortment mix for you?

  • "Are you trying to find the shortest path

  • "in order to get to your stores?"

  • And then the question is, "Do you now have the right data

  • "to solve that problem?"

  • A lot of times we put the science

  • and I'm a data scientist by training.

  • I would love to talk about the science,

  • but really, it's the problem first.

  • The data and the science, they come after.

  • >> Thanks, good advice.

  • Any other questions in the audience?

  • Yes, one right up here.

  • (laughing)

  • >> Test, test. Can you hear me?

  • >> Yep.

  • >> So with AI machinery becoming more commonplace

  • and becoming more accessible to developers and visionaries

  • and thinkers alike rather than being just a giant warehouse

  • of a ton of machines and you get one tiny machine learning,

  • do you foresee more governance coming into play

  • in terms of what AI is allowed to do

  • and the decisions of what training data is allowed

  • to be fed to Ais in terms of influence?

  • You talk about data determining

  • if AI will become good or bad,

  • but humans being the ones responsible

  • for the training in the first place, obviously,

  • they can use that data to influence as they,

  • just the governance and the influence.

  • >> Jack: Who wants to take that one?

  • >> I'll take a quick stab at it.

  • So, yes, it's going to be an open discussion.

  • It's going to have to take place,

  • because really, they're just machines.

  • It's machine learning.

  • We teach it.

  • We teach it what to do, how to act.

  • It's just an extension of us and in fact,

  • I think you had a really great conversation

  • or a statement at lunch where you talked about

  • your product being an extension of a designer because,

  • and we can get into that a little bit, but really,

  • it's just going to do what we tell it to do.

  • So there's definitely going to have to be discussions

  • about what type of data we feed.

  • It's all going to be centered around the use case

  • and what that solves the use case.

  • But I imagine that that will be a topic of discussion

  • for a long time about what we're going to decide to do.

  • >> Jack: Michelle do you want to comment on this thought

  • of taking a designer's brain

  • and putting it into a model somehow?

  • >> Well, actually, what I wanted to say was that

  • I think that the regulation and the governance

  • around it is going to be self imposed

  • by the the developer and data science community first,

  • because I feel like even experts who have been doing this

  • for a long time don't rally have their arms

  • fully around what we're dealing with here.

  • And so to expect our senators, our congressmen, women,

  • to actually make regulation around it is a lot,

  • because they're not technologists by training.

  • They have a lot of other stuff going on.

  • If the community that's already doing the work

  • doesn't quite know what we're dealing with,

  • then how can we expect them to get there?

  • So I feel like that's going to be a long way off,

  • but I think that the people who touch and feel

  • and deal with models and with data sets

  • and stuff everyday are the kind of people

  • who are going to get together and self-regulate for a while,

  • if they're good hearted people.

  • And we talk about AI for good.

  • Some people are bad.

  • Those people won't respect those convenance

  • that we come up with, but I think

  • that's the place we have to start.

  • >> So really you're saying, I think,

  • for data scientists and those of us working in this space,

  • we have a social, ethical, or moral obligation

  • to humanity to ensure that our work is used for good.

  • >> Michelle: No pressure.

  • (laughing)

  • >> None taken.

  • Any other questions?

  • Anything else? >> I just wanted to

  • talk about the second part of what she said.

  • We've been working with a company that builds robots

  • for the store, a store associate if you will.

  • And one of their very interesting findings

  • was that the greatest acceptance of it right now

  • has been at car dealerships

  • because when someone goes to the car dealer

  • and we all have had terrible experiences doing that.

  • That's why we try to buy it online,

  • but just this perception that a robot would be unbiased,

  • that it will give you the information

  • without trying to push me

  • one way or the other. >> The hard sell.

  • >> So there's that perception side of it too that,

  • it isn't that the governance part of your question,

  • but more the biased perception side of what you said.

  • I think it's fascinating how we're already trained

  • to think that this is going to have an unbiased opinion,

  • whether or not that true.

  • >> That's fascinating.

  • Very cool.

  • Thank you Sarush. Any other questions in the audience?

  • No, okay.

  • Michelle, could I ask, you've got a station over there

  • that talks a little bit more about your company,

  • but for those that haven't seen it yet,

  • could you tell us a little bit about

  • what is the experience like

  • or how is the shopping experience different

  • for someone that's using your company's technology

  • than what it was before?

  • >> Oh, free advertising.

  • I would love to.

  • No, but actually, I started this company

  • because as a consumer I found myself

  • going back to the user experience piece,

  • just constantly frustrated with the user experience

  • of buying products one at a time and then getting zero help.

  • And then here I am having to google

  • how to wear a white blazer to not look like an idiot

  • in the morning when I get dressed with my white blazer

  • that I just bought and I was excited about.

  • And it's a really simple thing,

  • which is how do I use the product that I'm buying

  • and that really simple thing has been just

  • abysmally handled in the retail industry,

  • because the only tool that the retailers

  • have right now are manual.

  • So in fashion, some of our fashion customers

  • like John Varvatos is an example we have over there,

  • it's like a designer for high-end men's clothing,

  • and John Varvatos is a person,

  • it's not just the name of the company.

  • He's an actual person and he has a vision

  • for what he wants his products to look like

  • and the aesthetic and the style and there's a rockstar vibe

  • and to get that information into the organization,

  • he would share it verbally with PDFs, thing like that.

  • And then his team of merchandisers

  • would literally go manually and make outfits on one page

  • and then go make an outfit on another page

  • with the same exact items and then products

  • would go out of stock and they'd go around in circles

  • and that's a terrible, terrible job.

  • So to the conversation earlier about people losing jobs

  • because of artificial intelligence.

  • I hope people do lose jobs

  • and I hope they're the terrible jobs

  • that no one wanted to do in the first place,

  • because the merchandisers that we help,

  • like the one form John Varvatos,

  • literally said she was weeks away from quitting

  • and she got a new boss and said,

  • "If you don't ix this part of my job, I'm out of here."

  • And he had heard about us.

  • He knew about us and so he brought us in

  • to solve that problem.

  • So I don't think it's always a bad thing,

  • because if we can take that route, boring,

  • repetitive task off of human's plates,

  • what more amazing things can we do with our brain

  • that is only human and very unique to us

  • and how much more can we advance ourselves

  • and our society by giving the boring work

  • to a robot or a machine.

  • >> Well, that's fantastic.

  • So Joe, when you talk about Smart Cities,

  • it seems like people have been talking

  • about Smart Cities for decades

  • and often people cite funding issues,

  • regulatory environment or a host of other reasons

  • why these things haven't happened.

  • Do you think we're on the cusp of breaking through there

  • or what challenges still remain for fulfilling

  • that vision of a smart city?

  • >> I do, I do think we're on the cusp.

  • I think a lot of it has to do, largely actually,

  • with 5G and connectivity, the ability to process

  • and send all this data that needs to be shared

  • across the system.

  • I also think that we're getting closer

  • and more conscientious about security,

  • which is a major issue with IOT,

  • making sure that our in devices or our edge devices,

  • those things out there sensing, are secure.

  • And I think interocular ability

  • is something that we need to champion as well

  • and make sure that we basically work together

  • to enable these systems.

  • So very, very difficult to create

  • little, tiny walled gardens of solutions in a smart city.

  • You may corner a certain part of the market,

  • but you're definitely not going to have that ubiquitous benefit

  • to society if you establish those little walled gardens,

  • so those are the areas I think we need to focus on

  • and I think we are making serious progress in all of them.

  • >> Very good.

  • Michelle, you mentioned earlier

  • that artificial intelligence was all around us

  • in lots of places and things that we do on a daily basis,

  • but we probably don't realize it.

  • Could you share a couple examples?

  • >> Yeah, so I think everything you do online

  • for the most part, literally anything you might do,

  • whether that's googling something or you go to some article,

  • the ads might be dynamically picked

  • for you using machine learning models

  • that have decided what is appropriate based on you

  • and your treasure trove of data that you have out there

  • that you're giving up all the time

  • and not really understanding

  • you're giving up >> The shoes that follow you

  • around the internet right? >> Yeah, exactly.

  • So that's basically anything online.

  • I'm trying to give in the real-world.

  • I think that, to your point earlier about he supply chain,

  • just picking a box of cereal off the shelf

  • and taking it home, there's not artificial intelligence

  • in that at all, but the supply chain behind it.

  • So the supply chain behind pretty much

  • everything we do even in television,

  • like how media gets to us and get consumed.

  • At some point in the supply chain,

  • there's artificial intelligence playing in there as well.

  • >> So to start us in the supply chain where we can get

  • the same day even within the hour delivery.

  • How do you get better than that?

  • What's coming that's innovative in the supply chain

  • that will be new in the future?

  • >> Well, so that is one example of it,

  • but you'd be surprised at how inefficient

  • the supply chain is, even with all the advances

  • that have already gone in,

  • whether it's physical advances around

  • building modern warehouses and modern manufacturing plants,

  • whether it's through software and others

  • that really help schedule things and optimize things.

  • What has happened in the supply chain

  • just given how they've evolved is they're very siloed,

  • so a lot of times the manufacturing plant

  • does things that the distribution folks do not know.

  • The distribution folks do things

  • that the transportation folks don't know

  • and then the store folks know nothing

  • other than when the trucks pulls up,

  • that's the first time they find out about things.

  • So where the great opportunity in my mind is,

  • in the space that I'm in, is really the generation of data,

  • the connection of data, and finally,

  • deriving the smarts that really help us improve efficiency.

  • There's huge opportunity there.

  • And again, we don't know it

  • because it's all invisible to us.

  • >> Good. Let me pause and see

  • if there's any questions in the audience.

  • There, we got one there.

  • >> Thank you. Hi guys, you alright?

  • I just had a question about ethics

  • and the teaching of ethics.

  • As you were saying, we feed the artificial intelligence,

  • whereas in a scenario which is probably a little bit

  • more attuned to automated driving,

  • in a car crash scenario between

  • do we crash these two people or three people?

  • I would be choosing two, whereas the scenario

  • may be it's actually better

  • to just crash the car and kill myself.

  • That thought would never go through my mind,

  • because I'm human.

  • My rule number one is self preservation.

  • So how do we teach the computer this sort of side of it?

  • Is there actually the AI ethic going to be

  • better than our own ethics?

  • How do we start?

  • >> Yeah, that's a great question.

  • I think the opportunity is there as Michelle

  • was talking earlier about maybe when you cross that chasm

  • and you get this new singularity,

  • maybe the AI ethics will be better than human ethics

  • because the machine will be able to think about

  • greater concerns perhaps other than ourselves.

  • But I think just from my point of view,

  • working in the space of automated vehicles,

  • I think it is going to have to be something that the industry,

  • and societies are different,

  • different geographies, and different countries.

  • We have different ways of looking at the world.

  • Cultures value different things and so I think technologists

  • in those spaces are going to have to get together

  • and agree amongst the community

  • from a social contract theory standpoint perhaps in a way

  • that's going to be acceptable to everyone

  • who lives in that environment.

  • I don't think we can come up with a uniform model

  • that would apply to all spaces,

  • but it's got to be something though that we all,

  • as members of a community, can accept.

  • And so yeah, that would be the right thing to do

  • in that situation and that's not going to be

  • an easy task by any means, which is, I think,

  • one of the reasons why you'll continue to see humans

  • have an important role to play in automated vehicles

  • so that the human could take over

  • in exactly that kind of scenario,

  • because the machines perhaps

  • aren't quite smart enough to do it

  • or maybe it's not the smarts or the processing capability.

  • It's maybe that we haven't as technologists and ethicists

  • gotten together long enough to figure out

  • what are those moral and ethical frameworks

  • that we could use to apply to those situations.

  • Any other thoughts?

  • >> Yeah, I wanted to jump in there real quick.

  • Absolutely questions that need to be answered,

  • but let's come together and make a solution

  • that needs to have those questions answered.

  • So let's come together first and fix the problems

  • that need to be fixed now so that

  • we can build out those types of scenarios.

  • We can now put our brainpower to work

  • to decide what to do next.

  • There was a quote I believe by Andrew Ningh Bidou

  • and he was saying in concerning deep questions

  • about what's going to happen in the future with AI.

  • Are we going to have AI overlords or anything like that?

  • And it's kind of like worrying

  • about overpopulation at the point of Mars.

  • Because maybe we're going to get there someday

  • and maybe we're going to send people there

  • and maybe we're going to establish a human population on Mars

  • and then maybe it will get too big

  • and then maybe we'll have problems on Mars,

  • but right now we haven't landed on the planet

  • and I thought that really does a good job of putting

  • in perspective that that overall concern

  • about AI taking over.

  • >> So when you think about AI being applied for good

  • and Michelle you talked about

  • don't do AI just for AI's sake, have a problem to solve,

  • I'll open it up to any of the three of you,

  • what's a problem in your life

  • or in your work experience that you'd love somebody

  • out here would go solve with AI?

  • >> I have one.

  • Sorry, I wanted to do this real quick.

  • There's roads blocked off and it's raining

  • and I have to walk a mile to find a taxi

  • in the rain right now after this to go home.

  • I would love for us to have some sort of ability

  • to manage parking spaces and determine when

  • and who can come in to which parts of the city

  • and when there's a spot downtown,

  • I want my autonomous vehicle to know which one's available

  • and go directly to that spot and I want it to be cued

  • in a certain manner to where I'm next in line and I know.

  • And so I would love for someone to go solve that problem.

  • There's been some development on the infrastructure side

  • for that kind of solution.

  • We have a partnership Intel does with GE

  • and we're putting sensors that have,

  • it's an IOT sensor basically.

  • It's called City IQ.

  • It has environmental monitoring, audio, visual sensors

  • and it allows this type of use case to take place.

  • So I would love to see iterations on that.

  • I would love to see, sorry there's another one

  • that I'm particular about.

  • Growing up I lived in Southern California

  • right against the hills, a housing development,

  • because the hills and there was not a factory,

  • but a bunch of oil derricks back there.

  • I would love to have sensor that senses the particulate

  • in the air to see if there was too many fumes coming

  • from that oil field into my yard growing up as a little kid.

  • I would love for us to solve problems like that,

  • so that's the type of thing that we'll be able to solve.

  • Those are the types of innovations that will be able

  • to take place once we have these sensors in place,

  • so I'm going to sit down on that one

  • and let someone else take over.

  • >> I'm really glad you said the second one

  • because I was thinking,

  • "What I'm about to say is totally going to

  • "trivialize Joe's pain and I don't want to do that."

  • But cancer is my answer, because there's so much data

  • in health and all these patterns

  • are there waiting to be recognized.

  • There's so many things you don't know about cancer

  • and so many indicators that we could capture

  • if we just were able to unmask the data and take a look,

  • but I knew a brilliant company

  • that was using artificial intelligence specifically

  • around image processing to look at CAT scans

  • and figure out what the leading indicators

  • might be in a cancerous scenario.

  • And they pivoted to some way more trivial problem

  • which is still a problem

  • and not to trivialize parking an whatnot,

  • but it's not cancer.

  • And they pivoted away from this amazing opportunity

  • because of the privacy and the issues

  • with HIPPA around health data.

  • And I understand there's a ton of concern with it getting

  • into the wrong hands and hacking and all of this stuff.

  • I get that, but the opportunity in my mind

  • far outweighs the risk and the fact that they had to change

  • their business model and change their company essentially

  • broke my heart because they were really onto something.

  • >> Yeah that's a shame and it's funny you mention that.

  • Intel has an effort that we're calling the cancer cloud

  • and what we're trying to do is provide some infrastructure

  • to help with that problem and the way cancer treatments

  • work today is if you go to a university hospital

  • let's say here in Texas, how you interpret that scan

  • and how you respond and apply treatment,

  • that knowledge is basically just kept

  • within that hospital and within that staff.

  • And so on the other side of the country,

  • somebody could go in and get a scan

  • and maybe that scan brand new to that facility

  • and so they don't know how to treat it,

  • but if you had an opportunity with machine learning

  • to be able to compare scans from people,

  • not only just in this country,

  • but around the world and understand globally,

  • all of the hundreds of different treatment pads

  • that were applied to that particular kind of cancer,

  • think how many lives could be saved,

  • because then you're sharing knowledge

  • with what courses of treatment worked.

  • But it's one of those things like you say,

  • sometimes it's the regulatory environment

  • or it's other factors that hold us back

  • from applying this technology to do some really good things,

  • so it's a great example.

  • Okay, any other questions in the audience?

  • >> I have one. >> Good Emily.

  • >> So this goes off of the HIPPA question, which is,

  • and you were talking about

  • just dynamically displaying ads earlier.

  • What does privacy look like in a fully autonomous world?

  • Anybody can answer that one.

  • Are we still private citizens?

  • What does it look like? >> How about from a

  • supply chain standpoint?

  • You can learn a lot about somebody in terms

  • of the products that they buy and I think to all of us,

  • we sort of know maybe somebody's tracking

  • what we're buying but it's still creepy

  • when we think about how people

  • could potentially use that against us.

  • So, how do you from a supply chain

  • standpoint approach that problem?

  • >> Yeah and it's something that comes up in my life

  • almost every day because one of the thing's

  • we'd like to do is to understand consumer behavior.

  • How often am I buying?

  • What kinds of products am I buying?

  • What am I returning?

  • And so for that you need transactional data.

  • You really get to understand the individual.

  • That then starts to get into this area of privacy.

  • Do you know too much about me?

  • And so a lot of times what we do is data

  • is clearly anonymized so all we know is customer A

  • has this tendency, customer B has this tendency.

  • And that then helps the retailers

  • offer the right products to these customers,

  • but to your point, there are those privacy concerns

  • and I think issues around governance, issues around ethics,

  • issues around privacy, these will continue to be ironed out.

  • I don't think there's a solid answer

  • for any of these just yet.

  • >> And it's largely a reflection of society.

  • How comfortable are we with how much privacy?

  • Right now I believe we put the individual in control

  • of as much information as possible

  • that they are able to release or not.

  • And so a lot of what you said,

  • everyone's anonymizing everything at the moment,

  • but that may change as society's values change slightly

  • and we'll be able to adapt to what's necessary.

  • >> Why don't we try to stump the panel.

  • Anyone have any ideas on things in your life

  • you'd like to be solved with AI for good?

  • Any suggestions out there that we could then hear

  • from our data scientist and technologist and folks here?

  • Any ideas?

  • No?

  • Alright good.

  • Alright, well, thank you everyone.

  • Really appreciate your time.

  • Thank you for joining Intel

  • here at the AI lounge at Autonomous World.

  • We hope you've enjoyed the panel and we wish you

  • a great rest of your event here at South by Southwest.

  • (audience clapping)

  • (bright music)

>> Welcome everyone.

字幕與單字

單字即點即查 點擊單字可以查詢單字解釋

B1 中級 美國腔

AI for Good Panel - Autonomous World - SXSW 2017 - #IntelAI - #theCUBE (AI for Good Panel - Autonomous World - SXSW 2017 - #IntelAI - #theCUBE)

  • 188 18
    alex 發佈於 2021 年 01 月 14 日
影片單字