- il y a 2 jours
The Human-AI Synergy Smart Devices for a Better Future
Catégorie
🤖
TechnologieTranscription
00:00Sous-titrage Société Radio-Canada
00:32Sous-titrage Société Radio-Canada
01:02Sous-titrage Société Radio-Canada
01:05Sous-titrage Société Radio-Canada
01:07Sous-titrage Société Radio-Canada
01:37Sous-titrage Société Radio-Canada
01:41Sous-titrage Société Radio-Canada
01:44Souris
02:15Sous-titrage Société Radio-Canada
02:47Sous-titrage Société Radio-Canada
02:51Sous-titrage Société Radio
03:21Sous-titrage Société Radio-Canada
03:54Sous-titrage Société Radio-Canada
03:56Sous-titrage Société Radio-Canada
03:59Sous-titrage Société Radio-Canada
04:00Sous-titrage Société Radio-Canada
04:02Sous-titrage Société Radio-Canada
04:07Sous-titrage Société Radio-Canada
04:11Sous-titrage Société Radio-Canada
04:23Sous-titrage Société Radio-Canada
04:26Sous-titrage Société Radio-Canada
04:31Sous-titrage Société Radio-Canada
04:33Sous-titrage Société Radio-Canada
04:45Sous-titrage Société Radio-Canada
04:55Sous-titrage Société Radio-Canada
04:58Sous-titrage Société Radio-Canada
05:02Sous-titrage Société Radio-Canada
05:07Sous-titrage Société Radio-Canada
05:11Sous-titrage Société Radio-Canada
05:14Sous-titrage Société Radio-Canada
05:16Sous-titrage Société Radio-Canada
05:19Sous-titrage Société Radio-Canada
05:22Sous-titrage Société Radio-Canada
05:22Sous-titrage Société Radio-Canada
05:24Sous-titrage Société Radio-Canada
05:26Sous-titrage Société Radio-Canada
05:26Sous-titrage Société Radio-Canada
05:26Sous-titrage Société Radio-Canada
05:27Sous-titrage Société Radio-Canada
05:28Ils sont vraiment capables de savoir ce que leur intente est leur intérêt.
05:32Ils peuvent supporter vous.
05:33La 3e est l'application.
05:36Je pense que avec la 1 et la 2e, l'AI capability,
05:40la 3e peut donner une meilleure AI expérience.
05:46Par exemple, l'AI, l'exploitation,
05:50et aussi l'application.
05:52Ce genre de application belongs à la 3e.
05:55Et si on parle de ChatGPT,
05:58je pense que c'est la 4e, l'AI application.
06:03Dans cette partie,
06:07votre téléphone peut permettre d'interfaire.
06:10Cette interface peut vous permettre
06:14de utiliser l'AI capability.
06:19Et à la même temps,
06:21ce qui peut aider à l'utiliser
06:25votre sécurité et votre privacité.
06:28Ce genre de données
06:29vous pouvez l'exchange avec la cloud AI.
06:33Et ce genre de services
06:35vous pouvez l'experience avec la cloud AI.
06:38Donc, avec cette 4e,
06:40je pense que dans le futur,
06:44votre téléphone peut empowerer et permettre
06:46d'autres personnes.
06:49Donc, par exemple,
06:50nous parlons de la 2e.
06:53Nous appelons, dans le futur,
06:55l'Operation System,
06:56c'est intent-based.
06:58Donc, ils peuvent,
06:59tous les jours,
07:00vous avez votre voix,
07:01votre glass,
07:02votre téléphone,
07:03votre PC, votre pad.
07:04Ils connaissent votre comportement.
07:06Ils connaissent votre comportement.
07:08Et puis,
07:10ils ont des jours,
07:10ils ont des questions.
07:11Qu'est-ce que votre
07:13potential requirement
07:14et votre intent ?
07:17Et puis,
07:18vous avez des choses
07:19par exemple,
07:21par exemple,
07:21vous avez un chat
07:22sur WhatsApp
07:23sur l'adresse
07:25de la restauration.
07:26Et puis,
07:27vous utilisez
07:288 d'apps,
07:29copy et paste,
07:31changez à un autre app
07:32pour naviguer.
07:34Par exemple,
07:35vous savez,
07:35l'intent-based
07:36operating system,
07:38ils appellent
07:40et ils analysent
07:42le contenu.
07:44Ils analysent
07:46votre intent
07:47et matchs
07:48pour la meilleure
07:50applications
07:51et services.
07:52Récond,
07:53c'est l'intent-based.
07:54Ces jours,
07:56le phénomène
07:57le fonctionnal,
07:58le fonctionnal,
07:59le fonctionnal,
08:00le fonctionnal.
08:03Et aussi,
08:05il y a cette
08:06l'intent-bés.
08:08Je veux dire
08:09la COVID-19,
08:10le fonctionnal,
08:11le fonctionnal,
08:12le fonctionnal,
08:13pour les gens
08:14est différent. Parce que quand vous traînez avec votre propre data,
08:19vous avez, vous savez, votre phone, c'est, vous savez, le long,
08:24le long, le long model.
08:26Ce long model aussi studies vos habits.
08:30Donc, quand vous avez, par exemple, Gatshub en la matinée,
08:34ils vont, si vous avez un habit,
08:36quand vous avez des Gatshub en la matinée,
08:39vous voulez entendre la musique,
08:40ils vont vous montrer ça.
08:42Si vous avez des business travel,
08:45quand vous êtes close to l'airport,
08:47ils vont vous montrer votre boarding pass
08:50sur le display.
08:51Donc, ce serait très smart pour vous soutenir.
08:55Ce serait le futur,
08:56je veux dire,
08:57l'internet-based AI,
08:59l'AI-enable,
09:02la personnalisation,
09:03l'operative système.
09:05Et, of course,
09:08l'AI
09:08l'AI will help us
09:10in the data protection
09:12and the privacy.
09:14Et, of course,
09:15l'AI,
09:16l'AI,
09:16l'AI,
09:18l'AI,
09:20l'AI,
09:21l'AI,
09:21l'AI,
09:22l'AI,
09:22l'AI,
09:22l'AI,
09:23l'AI,
09:24l'AI,
09:25l'AI,
09:26l'AI,
09:28l'AI,
09:29l'AI,
09:29c'est
09:30l'AI,
09:37l'AI,
09:42l'AI,
09:43l'AI,
09:45l'AI.
09:47They will help you manage the data exchange with the cloud AI.
09:55Today, you know, many cloud AI, they always ask some, you know,
10:01the right for photos, contacts, lists, and also stories.
10:08Even for us, you know, I'm the expert on the smart device.
10:13Sometimes I feel some, you know, confused.
10:18I should be approved, allowed, or not.
10:22So in the future, the on-device AI will help you manage this.
10:29They will keep the minimum authorization for the right.
10:36And they will control the reasonable, the need, the data.
10:42To exchange with the cloud.
10:46So the data cannot be abused by cloud AI.
10:50So we use this way that use AI technology to build up protection for everybody.
10:57And of course, with AI capability, we also can do a lot of things, I mean, on-device AI.
11:05Can do things, for today, people cannot do.
11:09For example, in August, the Olympic game will start in Paris.
11:17So they have a lot of, you know, fast spot to compete by employer.
11:24For example, fencing.
11:26Yeah.
11:27Even for the professional photographer, it's very difficult to catch the best moment.
11:34But with the phone, with AI, you can do it very easily.
11:38For example, like this.
11:41They will record and analyze the eighth moment to help you select the best moment.
11:50For example, like this.
11:54You know, this fencing is so fast.
11:56Even we cannot identify who will be the winner.
12:01But, you know, with the AI technology, on-device AI, we can make it happen.
12:08Of course, cloud AI cannot.
12:11Because cloud AI needs more transportation, more throughput.
12:16It's a longer delay.
12:18But with, you know, the on-device AI, we can make that happen.
12:23I also want to say, many classic, you know, moments, also can be done by on-device AI.
12:35And this is, I think, maybe many people know this gentleman, Karl Laffield.
12:43It's a very, very famous fashion designer.
12:45This photo is shot by our cool studio in Paris.
12:53Our cool studio today, it has 90 years history.
12:59This is the art of the light.
13:02And why and how they can shot these beautiful photos?
13:10Because they have a very complicated process
13:14to guarantee its photo is so classic.
13:20Yeah, many people I know want to have this photo, this portrait.
13:26And with on-device AI, we can recreate this best moment,
13:35this best portrait by today's our smartphone.
13:41You know, when we honor, cooperate with our cool,
13:45that we can see, I mean, the light, the beautiful,
13:50you know, the show on the devices.
13:53I mean, on-device AI, we can do a lot of things.
13:59And the first, here I want to, I'm really honored that, honorable,
14:04to claim their first smartphone to support our cool portrait.
14:11It's an honor to handle it.
14:14We are coming soon in June 12th in Paris.
14:19Thank you.
14:24And today we have smartphone.
14:27We also have a lot of other devices.
14:31How to integrate all the devices work together.
14:36Because we not only started from the phone application,
14:40the phone uses.
14:42We also have, you know, the house data for the smartwatch,
14:46the ring.
14:47Also that we have glasses.
14:49And the AI can study and manage the cross-device
14:54and the cross-operating system application.
14:58With this capability, you can see the service
15:02and the computing resource can be served among all the devices.
15:07So today your PC, your phone, your watch is not isolated.
15:12It can be supporting each other.
15:16You can see that today the photo or file, the service,
15:20and even for the camera can be served among all the devices.
15:26This is the AI-enabled cross-device and cross-OS application
15:30and the services.
15:34Honor is open to the whole industry collaboration.
15:39So to honor the future AI application,
15:43we'll cooperate with Microsoft,
15:47Curricom, Google,
15:50and this partner, you know,
15:52when all the partners work together,
15:55we do believe we can deliver a better data life.
15:59And of course, the AI capability should not only use
16:05in the flagship smartphone.
16:07they will be readily used in honor product.
16:11So we want more and more
16:14that users can enjoy the AI service and the capability.
16:20I do believe that with this AI capability,
16:23we can live better, live the way we want.
16:28Yeah, from this month,
16:30we'll start to upgrade our existing product for Magic V2,
16:38for honor netting.
16:40And in the future, that will support more,
16:43I mean, the AI capability in our smart devices.
16:48And I really appreciate the industry partner
16:52that collaboration with honor.
16:54we do believe.
16:55When we work together,
16:57we keep openness and collaborative
17:00and continue innovate in the AI smart device industry.
17:05We can make things better.
17:07Thank you very much.
17:08Thank you.
17:13Let's go to the, yeah.
17:16Invite my friends back to the States for next session.
17:20Okay, please.
17:26Well done.
17:27Come and join me for the panel.
17:29Justine.
17:37So, George, thank you very much for providing an overview
17:40of your expectations for honor.
17:43We're now joined by Professor Justine Cassell.
17:46For those of you who don't know Justine,
17:48she's a highly acclaimed expert in the AI world.
17:52We were reminiscing about early efforts back in the 1990s
17:57on products like the Nokia Communicator 9000,
17:59which I know will be a familiar product
18:01to a few of the older people in the audience.
18:04But Justine is a Dean's Professor at the School of Carnegie Mellon,
18:10but also currently on leave and based here in Paris,
18:14working for the Prairie AI Research Institute.
18:18So, we couldn't have someone more qualified
18:19to help us with this discussion.
18:22Thank you.
18:22But, George, let's start with the vision
18:24that you've got with your devices.
18:25And you talked about some of the new capabilities
18:28with current devices and future devices.
18:32and I wanted to ask you about Magic Portal
18:35because it feels to me, you know, I've used it.
18:37I had the software drop arrive on my phone this week.
18:40The ability to seamlessly take content
18:43and the phone almost knows what I want to do with it.
18:46I can drag it and drop it into another application very simply.
18:50What's your vision for that interaction with Magic Portal?
18:54So, the owner developed the Magic OS system.
18:58Our key driven and the motivation is how to make the smartphone really understand
19:03the people requirement and the intent.
19:07So, we use AI reconstruct operating system.
19:12And then, every day that we use the smartphone,
19:16you interact with your phone.
19:19you teach them what is your want.
19:23And based on this, you know, the idea and the philosophy,
19:26that we develop a Magic Portal.
19:29Magic Portal, you know, the first,
19:31when you that focus some message or a photo, a picture,
19:36and a multimode, I mean the information,
19:38they will analyze what is the key message for this, you know,
19:42for example, for this message or this photo.
19:45And then, every day when you, you already told your phone something, right?
19:50For example, that your friend, you go to restaurant,
19:56you have it to go shopping or someplace.
19:59So, when you post the message,
20:02and then they potentially the magic line with the AI engine is working.
20:07They will study that, what is your intent.
20:10And then, they prioritize the best choice for you.
20:15They'll help you to find the best option.
20:18And then, the one step that go to that, what you want.
20:22So, you're baking AI into your Magic OS,
20:26into the framework of the device from the bottom up.
20:29Sure.
20:29And how's that journey happen?
20:31So, you obviously have engineered this for quite some time,
20:35and now we're starting to see it.
20:36So, the philosophy was what? To make it easier to use the phone?
20:40In fact, we've worked on this, you know,
20:43in the past four to five years.
20:46Right.
20:46I've honored to work on this direction strategy.
20:50We do believe that AI should empower the people.
20:54So, you know, when we travel, you know,
20:57I walk along the, you know, Santa's River,
21:01only the phone with me.
21:02So, when we rely on my phone,
21:04that should be the, give me the best support.
21:07So, when we did study and support me,
21:10there should be the, you know,
21:12really understand what is my needs, my requirement.
21:15So, this magic portal today,
21:17we already support globally top 100 apps.
21:21Can easily, I mean, prioritize what is your requirement.
21:25So, almost the phone starts to know what you want before you kind of know.
21:30Yeah.
21:30Let's turn to you, Justine,
21:31because, of course, this concept of multimodality
21:34is something that you've studied for, you know, over a decade.
21:37Yeah.
21:37So, maybe you could share some thoughts, you know,
21:40with the honor solution in mind,
21:42the broader context of how humans interact with their devices.
21:45So, George has talked about intention recognition.
21:49And, of course, intention recognition is a very old concept
21:53in natural language processing.
21:55But, we don't just use the words that we speak
22:00in order to understand other people.
22:03We are multimodal creatures.
22:05Our bodies move.
22:07Our voices have intonation and loudness.
22:10And, all of these modalities play a role in communicating to others
22:16and in understanding what others are communicating to us.
22:20And so, when George talks about the magic portal,
22:23I think what he's talking about is using all of the modalities
22:27that are available to us from the variety of apps that we use
22:32on a phone to better understand what people are doing
22:37and better understand what they're trying to do,
22:40what their intentions are.
22:43And, Justine, I mean, thinking about the multiple ways
22:45in which we interact with devices, it's not just our hands,
22:48it's our voice, it's our eyes.
22:50You've thought a lot about that.
22:52And, you know, on a recently at the Mobile World Congress event,
22:54we're doing some, you know, kind of breakthrough research
22:58around the use of eyes for user interface.
23:00But, maybe you could talk about that multimodality with regards
23:03to the different ways in which we work with devices
23:06and how that will evolve.
23:07Eye gaze is a very powerful cue.
23:10For example, I'm looking at you, I'm pointing my eyes at you
23:14to indicate that I'm responding to something that you said.
23:18And, that use of eye gaze can allow us to know what people are concentrating on
23:23in a scene.
23:25For example, you can imagine that when you watch a video that's just so tiny
23:29on your phone, if you focus on a particular part of that video,
23:34more pixels can be sent to that part of the video
23:38to increase the size and increase the ability to see what's going on.
23:44But, we also have to keep in mind that there are multiply abled people in the world.
23:50Some of us can use our bodies and our hands and our eyes and our voice.
23:55But, some people are locked into their bodies.
23:58They can't use their voice to control a phone.
24:01And, for them, the use of eye gaze is extremely powerful in conveying what they want to do.
24:10So, the more modalities that we can engage into an AI system,
24:16the more different kinds of people we can help with that AI system.
24:22So, George, let's talk about another one of the products you talked about.
24:25You showed a stack of AI-capable elements in the kind of Honor profile.
24:32And, you talked about the magic ring concept,
24:35where you're bringing together an ecosystem of devices,
24:39where it's not just about empowering the phone, for example,
24:43but allowing that device to work with a multitude of other things.
24:46I saw you highlighted the watch. You had the PC.
24:50How do you see your magic ring concept helping differentiate
24:54Honor from its rivals?
24:57Yes, sir.
24:58You know that today, we use a lot of different devices.
25:02You mentioned the phone, tablet, the PC, the watch.
25:06In fact, when we, for example, do sports jogging,
25:10that is just a weird watch.
25:11But, you know, we know the watch can, you know,
25:15the monitoring, the heartbeat, and also something.
25:18But, I mean, the computing capability weak.
25:24So, with the magic ring, that all devices work together,
25:30can be unified and managed by the AI engine.
25:35So, when you use the watch service,
25:38you have the support backside with your computer,
25:42your PC and your phone, your tablet.
25:44So, I mean, in the honor of the design philosophy,
25:50the magic OS, magic ring, is human-centric.
25:54Human is the center of all the devices.
25:58When you use the watch, the watch is the central device.
26:02Right.
26:02The phone, the pad, the PC is the supporting device for your watch.
26:07The magic ring makes this happen.
26:10So, this human-centric concept, Justine,
26:13but also maybe this idea of this distributed architecture of devices.
26:17I know that was something that caught your attention.
26:19How does that fit into the world that you see in the future?
26:22So, I have to say that I laughed when I heard George talk about the magic ring.
26:27Because, for decades, I've said, I have fat little fingers.
26:32I'm really not very good at typing on my phone.
26:35When I come home, I simply want the information that I'm working on
26:39to fly to a screen on the wall or onto my laptop.
26:44So, I don't have to keep typing on that tiny little phone.
26:48Likewise, the other day I needed to send a photograph of a document to a colleague.
26:53So, I took a picture of the document on my phone.
26:57I saved it to the phone.
27:00I attached it to an email on my phone.
27:04I sent it to myself.
27:07I went to my PC.
27:08I retrieved it.
27:10I saved it to the desktop of my laptop.
27:13And then finally, I was able to send the darn photo to the colleague.
27:18It had taken me 20 minutes.
27:20And when we talk about this kind of ecosystem of devices,
27:24it seems to me that this is an idea that's been a long time in coming.
27:29For two reasons.
27:30One, so that we can drag and drop both the power of phones
27:33and the content of phones to other devices.
27:37But also because, as I said earlier, we're multimodal.
27:41Increasingly, we're seeing not just phones, not just watches,
27:45but also rings, and more recently, pins.
27:49And all of that information should be able to work together to convey what we're doing
27:55and to understand what we want to do.
27:57So, we're very tight on time.
27:59So, we're going to have to get this last question really tight.
28:01But one of the other things you talked about was on-device AI,
28:05and the importance of privacy and security for people.
28:08George, just a few comments on your philosophy around using on-device AI
28:13to ensure that people's personal information is ring-fenced.
28:16How's that being designed into your products?
28:18So, the other philosophy is that the end consumer data and privacy
28:24should be fully protected by the, I mean, your product, by the innovation.
28:31So, the AI in time, this is the most important thing for other.
28:34This is our design philosophy.
28:36Okay.
28:37And Justine, anything on privacy?
28:39Well, certainly, if we're using all of the devices that are available on our phone
28:44and all of the apps that are available on our phone,
28:46we don't necessarily want all that information to be sent to the cloud.
28:50And I think everybody agrees that on-device AI is the future.
28:54Of course, it's compute-intensive,
28:56and we need to understand how to use large language models,
29:00how to reduce the energy they require
29:02so that our phones last more than 10 minutes.
29:05But we also need to focus on security on the phone
29:09so that less information about us is shared throughout the world.
29:13Well, I'm very sorry. We've run out of time.
29:15There's so much we could have discussed.
29:17But George, Justine, thank you so much for sharing your thoughts.
29:21Please join me in thanking the panel for their contributions.
29:24Thank you.
29:25Thank you.
Commentaires