• 2 months ago
Variety returns in-person for its annual Entertainment & Technology Summit, presented by City National Bank. This one-day event will explore advancing technology’s impact on TV, film, gaming, music, digital media and consumer brands.

Category

People
Transcript
00:00So with AI, we are at an inflection point.
00:04And there's great optimism about the potential
00:07to innovate with AI, and there's also great unease
00:10in areas such as ethics.
00:12But one thing is certain, AI has arrived.
00:16If we could show the slide.
00:22Or without the slide.
00:24We have some recent data
00:25from Variety Intelligence Platform Survey,
00:29and you'll see that the majority
00:30of US media and entertainment companies
00:32are already using or exploring the use of AI.
00:38And in this session, we're going to explore
00:40AI's potential, looking at how it's being used today
00:44and what we can anticipate going forward.
00:48I'm gonna start by introducing our speakers
00:50and asking each of them to take a minute or two
00:53to talk about how you're using AI today.
00:57And in no particular order,
01:01I'm gonna keep the introductions brief,
01:03but please go online and learn more about our speakers.
01:08Seated next to me is Michael Wise,
01:10Chief Technology Officer
01:11at Universal Films Entertainment Group.
01:13Would you like to start?
01:15Sure, hi everybody.
01:16I guess the genie is a bit out of the bottle,
01:19last year and a half, it's been quite a year.
01:21Across the studio group at Universal,
01:24we've done, as everybody in our industry,
01:28an extraordinary amount of research, testing,
01:31understanding where generative AI can add value
01:33and where it really doesn't and where it's not ready.
01:36From our point of view, it's really all about enabling
01:40creators and the creative community
01:41that we work with and depend on to tell
01:44even more amazing and better stories
01:45in a way that really helps them tell stories,
01:48which is a big part of, an essential part of it,
01:50myself and my technology team is focused on.
01:53Areas where we found, I think, some near-term value
01:56is particularly in the area of business analytics,
01:58understanding things like box office performance,
02:02marketing effectiveness, that sort of thing.
02:05Helping our creatives with ideation,
02:08understanding the art of possible and storytelling.
02:11We've looked at examples of where we can use AI
02:14and some visual effects workflows,
02:16re-aging or de-aging is another
02:18really interesting one as well.
02:20And sort of, you know, the art of imagination
02:24and what's possible around AI and creativity
02:27and storytelling is what we're focused on.
02:30Great, and to elaborate more on creativity,
02:32we have Victoria Boussis,
02:35who's a filmmaker and creative technologist.
02:38Good morning, everyone.
02:40For me, I look back two years ago
02:42and two years ago when OpenAI first came out,
02:45a friend of mine told me to try it.
02:47Little did I know at that moment,
02:48my life as a creator would completely change.
02:51Now AI is at every critical phase of my creative process,
02:56all the way from research, story development,
02:59visual development, your chat GPTs, your Gemini's,
03:04all the way then to production, right?
03:06I do a lot of work in the 3D space,
03:08so now there's a lot of text and image to 3D model,
03:12editing, post-production, sound design, sound engineering,
03:17you name it, it has completely revolutionized
03:20how I create, how quickly I bring a product to people
03:25and how quickly I ideate.
03:26So it's been mind-blowing, but also equally thrilling.
03:34Next, we have Eli Collins,
03:35Vice President of DeepMind at Google.
03:39First of all, thank you so much for having me.
03:41We've pioneered a lot of the technologies in generative AI.
03:44I've been working really closely with creators,
03:46actually over a decade,
03:48starting with our music efforts back in 2016
03:50and kind of getting researchers in the studio
03:52and creators in our research lab.
03:56And we've just been really excited by all the recent,
03:58I think we're seeing what you all have seen
04:00in terms of the potential of the technology
04:02to enable and empower creators.
04:05That's always been our aim.
04:07And we're really excited to work with the broader community
04:09to see where it takes us.
04:13Do you wanna elaborate just a little bit more
04:15on some of your current initiatives?
04:17Sure.
04:18So we've pioneered models in pretty much every modality,
04:22whether it's text, image, audio, music, and video.
04:26And so we've got a pretty established work stream
04:29with the music industry.
04:30We've kind of showcased some of that work at IOH recently.
04:34More recently, we've been doing a lot of interesting work
04:36with filmmakers.
04:37So we showcased some of the work that we did
04:41with Donald Glover and his production company at IOH
04:43on kind of pushing the frontiers of storytelling
04:46with new aid models and figuring out
04:47how it can be integrated in the workflow.
04:49More recently, YouTube Madeon,
04:51we showcased a lot of the work that we've been doing
04:53with creators on that platform
04:55and kind of helping them integrate AI
04:59into their storytelling.
05:00And so we pretty much work across the spectrum
05:04from kind of individual artists
05:05to the Wycliffs of the world.
05:08And we're really just trying to enable people
05:12to push the frontiers of their medium
05:14and kind of learn from them and vice versa
05:16as we continue to push the technology.
05:20Great.
05:21I'd like to introduce Katja Reitmeyer,
05:23who's Director of Data Science and AI Media
05:25and Entertainment at NVIDIA.
05:27Good morning, everyone.
05:28Thanks for having me.
05:30So yeah, my perspective, I guess,
05:31is a little bit different.
05:32NVIDIA, we obviously produce GPUs.
05:34A lot of the AI in the world runs on our hardware,
05:37but we're also very much involved on the software side.
05:40In my area specifically,
05:42I'm working with a lot of the streaming companies,
05:44live media companies on data science and AI.
05:47And so that really includes everything
05:50around recommendation systems and data analytics.
05:53I think you were referring earlier,
05:54you get a lot of business data
05:56in terms of what you produce,
05:57what user data, what people are watching.
06:00So I'm really focused on, I would say,
06:02the non-sexy generative AI part
06:04because it's things that we've been doing for a long time,
06:06so this is not really new.
06:07But with the advent of AI and all these models,
06:10these LLMs, there's so much more that can be done
06:13that is much more user-friendly,
06:14that will gain us much better insights
06:16that we've been getting.
06:17And I think from the consumer perspective,
06:19at least in streaming,
06:20I think that means we're gonna get
06:22better, more tailored content,
06:24more things that we actually wanna watch
06:25and we can find it.
06:26Because I'm sure everybody's been
06:27on their favorite streaming site
06:29and re-watched a show that they've seen before
06:31because they couldn't find anything
06:32even though there's so much beautiful content out there.
06:35And so I'm really focused on figuring out
06:38how can we make that more efficient and better
06:41and obviously faster running on our hardware
06:42with our technology and with the help of our partners.
06:46And on the personal side, I mean, I'm thrilled like you,
06:49like in my daily life, I use AI constantly,
06:51whether that's through chat
06:53or I just use Google to create a podcast
06:57just with notes from my panel today
06:59where I'm just kind of hearing my own podcast of information
07:02it's just very useful.
07:04So I think it creates the possibility for people to learn
07:07and really democratize learning
07:08and anything you really wanna know these days,
07:11you can have an expert at hand.
07:13I think that's very exciting
07:14and that creates creators also in areas
07:17where we maybe didn't have them before
07:18because it's just accessible to so many more people.
07:22I mean, I think I have three AI apps on my phone
07:26and I mean, all the way from making
07:27your social media posts, literally,
07:29I'm like, please add emojis.
07:31There goes the emoji, please add a hashtag.
07:33There goes a hashtag.
07:34I'm like, this is brilliant.
07:35This is incredible, yeah.
07:38Okay, and our last speaker is Abhi Shauk
07:40who is Vice President Travel Incubator at Expedia Group.
07:45Well, hello everybody.
07:47Such a good crew, such a diverse crew.
07:49Pleasure to be here with you all.
07:51Look, at Expedia, we're storytellers too
07:56and the traveler is the hero in our story
08:02and the way we see it is a lot of what
08:06Gen AI and everything else that's powered through Gen AI,
08:09we see that as supporting cast and crew
08:12and with the traveler being the hero,
08:17our core goal is to ensure that we provide
08:23magical experiences right from discovery
08:26to trip building to shopping and to going through
08:30these beautiful experiences as they go through their travel
08:35and we come at it slightly differently.
08:39We don't wanna inject AI for the sake of doing it.
08:43We wanna just, we wanna reverse the order.
08:45We go applications first.
08:47What are those magical experiences
08:50and how can we leverage some of these generative models
08:55to peppered across these experiences
08:58as a collection of moments rather than
09:00just a bunch of chat bots opening and talking to you.
09:04It's kind of where most of our R&D is at
09:06and that's kind of where we see it
09:08and how we see it as powering the story.
09:13So I'd like to kick off by starting a discussion
09:16about creativity and Victoria, I'd love if you would start.
09:19You talked in your introduction about
09:22affecting all aspects of filmmaking.
09:24Could you tell us a little bit about a recent project
09:26as an example and where were the biggest changes you saw,
09:31whether that be in pre-visualization
09:34or in the editorial department or elsewhere?
09:37Yeah, in terms of pre-vis
09:40and in terms of visual development,
09:42I mean using mid-journey is like an afterthought for me.
09:46It's just happened so naturally.
09:49Dolly, so I create my pitch deck documents.
09:51I create my movie posters.
09:52I create short form videos, scissor trailers
09:57and the various tool sets there.
09:59So your ability to actually showcase your work
10:01also is easy and to ideate and to allow people,
10:04especially in the C-suite, to really see your vision quicker
10:08and you're allowed to do it by yourself.
10:10I mean, of course, I have a concept team as well
10:12that works on top of those images and we get there faster,
10:15but it really allows me to translate the ideas in my mind.
10:19In terms of a current project that we're working on
10:21that I'm super excited about,
10:23I actually can't even hold my excitement in sometimes,
10:27is this incredible project we're working
10:28with one of the most iconic people of our time.
10:30And this project is going to be
10:32a location-based museum experience
10:34in addition to an in-headset experience
10:37and hopefully with the Apple Vision Pro as well.
10:39And in this one, what I wanted to do is challenge myself
10:42to be able to change the narrative
10:43based on various data points,
10:46using sentiment analysis and emotional analysis
10:48to add things from social media, various buzzwords.
10:52Are we in time of harmony?
10:53Are we in time of conflict?
10:57And adding elements from news reports and news clippings
10:59so that the visuals actually change dynamically
11:02as these inputs are taken in real time.
11:05Now that's incredible, right?
11:06Because it just brings out the beauty
11:09and it brings out depth to your narrative,
11:11which really connects personally to your audience
11:14at that moment that they're experiencing
11:16this thing in the world.
11:17What we're also doing is we're creating,
11:19it's either gonna be a 3D model of this person
11:21or possibly I'll talk to you about D-vine
11:25so that we create this ongoing conversation
11:28you can have with this person.
11:30And we're taking natural language models for that.
11:33Obviously, they're gonna be walled
11:34to protect the integrity of this person
11:36because that's one of the things that we think of
11:38in terms of ethics, human identity, and privacy.
11:44So we're working towards that,
11:45but how cool is it to be learning
11:47from one of your favorite people?
11:49So that's the way I'm using it now.
11:53And I'm actually really excited for it
11:56to come to audiences, so yeah.
12:00Runway just announced this morning
12:02that it would be providing up to $1 million grants
12:06to filmmakers.
12:07What does everybody think about that?
12:11Look, I think it's a great move.
12:15And what's actually happening is,
12:18if you think about it,
12:20creators are being democratized
12:23not just from the perspective of channel of distribution,
12:26but from the perspective of tools.
12:28You can be a one-person shop
12:31building everything from pre-production
12:33to creation to post-production
12:36at a scale that you haven't thought about in the past.
12:38And I think that's a big, big deal.
12:40And Runway's ability to do that,
12:42I think is pretty smart in terms of video creation.
12:46I think it's also very smart
12:47because at the end of the day,
12:48I mean, all of us react to what we see, right?
12:51We could be as geeky as we wanna be in the background,
12:55but you're gonna connect to the story
12:56that we're gonna tell.
12:58So Runway's expansive platform,
13:00and I think they just came out
13:01with their Gen 3 model as well,
13:03which is text to video, video to video,
13:05anywhere from 10 to 15 second video clips.
13:09I think that we can use all these things,
13:12but if we don't see it translate to touch your lives,
13:15then it doesn't really matter.
13:17Runway started by doing a Runway AI film festival as well.
13:22And so I think with funding creators,
13:24we're gonna get a lot more of that content out to people
13:28to also change the narrative of how AI is being perceived,
13:32to make it feel, as you said so correctly,
13:35this democratization of AI versus this fear towards it
13:40as ruining creativity
13:42and doing all these horrible, bad things.
13:45So I think it's brilliant.
13:47I'm excited for it,
13:48and hopefully more people will follow in their footsteps.
13:52Yeah, I mean, I think it's really important
13:53that the people building the technology
13:55work with the creative community.
13:56So I think it's another example of that,
13:59which I think is a really positive trend,
14:01and I think we'll see more of that.
14:03So I think it's kind of representative
14:06of something that's happening more broadly.
14:08The other thing is it also speaks to the new types
14:10of content that are getting created, right?
14:12They're not using the million dollars probably
14:15to do AI versions of what already exists.
14:19They're probably trying to inspire creators
14:21to do new types of things that are kind of enabled
14:24uniquely by the technology or would have happened otherwise,
14:26and it's important that we also kind of seed
14:29that sort of work as well.
14:31And so to me, it speaks to those two aspects
14:33of what's going on, which I think are.
14:34Yeah, I think you both have touched
14:36on something really, really interesting,
14:38and that is where is AI gonna lead to
14:40in terms of really the future of storytelling?
14:43I personally think less about will a technology
14:47be able to replace exactly what we do today
14:49and do it tomorrow in a different way.
14:51That's an interesting conversation,
14:52but I think a more interesting conversation for us
14:54is what does AI enable in terms of dynamic storytelling
14:58or new forms of storytelling that's more personalized,
15:01more emotive, and help really bring stories to life?
15:07As we think about the evolution of AI in our business,
15:12we really look at it through an ethical lens.
15:16What is responsible use of AI?
15:17How do we think about issues such as bias in audience,
15:20bias in storytelling, bias in story?
15:23What's an ethical use of these tools?
15:25Where did the IP come from?
15:26How is the IP used in the tool?
15:28How is the IP expressed in a way that's copyright-friendly,
15:30in a way that allows us to all sustain
15:32our business model and industry in a responsible way?
15:35Do these tools actually provide content that is,
15:38you know, again, is it safe?
15:40Does it cause harm?
15:42Does it engender, you know, violence
15:44or some unintended consequence
15:47that none of us, I think, would wanna see?
15:49So that's a really important thing for us, too.
15:52What are, could any of you give an example
15:55of an initiative right now that's taking place
15:57that you think will help address
15:59some of those ethical concerns?
16:02I mean, there's various industry groups
16:05really, really, really focused on this issue.
16:08In our studio, we have an active effort
16:12and conversation around these topics.
16:15It's important, it's a really moving and evolving space.
16:18I think also there's, I think ethics
16:21is a huge, huge piece of this, right?
16:23Because our liberties as human beings,
16:27our creativity, our copyrights, our privacy,
16:31the fairness of AI, the transparency of AI,
16:35I think is critical.
16:37And so right now, the European Commission,
16:40who's been a very proactive element
16:43in protecting ethics,
16:46released their Digital Transformation Law last September
16:49and recently now just released their AI Act,
16:52which creates this transparency, this adoption,
16:54and these legal criteria that companies have to follow
16:56to safeguard those liberties.
16:59The US is trailing a little behind.
17:02There is no federal law that protects things
17:06at the way that the EU is.
17:08But I believe that the state of California
17:10has been very proactive in starting legislation
17:14to start to think about these issues
17:16and be proactive versus reactive.
17:19But it's a shared responsibility that we all have
17:21as individuals, as organizations, and also as governments.
17:26I think it's, you know, obviously we have,
17:30we have to think about responsibility,
17:31everything from the creation and invention of the technology
17:34down through the deployment in our platform.
17:36So I think it's something, it's a big space.
17:38It's really important.
17:39We have to think about it everywhere.
17:42And it is something that we actually,
17:44I think people, we get so much attention
17:46on like the amazing models and the artifacts.
17:48We actually do a lot of research
17:49on the responsibility side as well.
17:51So we invented a technology called SynthID,
17:53which watermarks the content to basically,
17:57to let people know that it's been generated by AI
18:00in a way that's difficult and hard to remove.
18:02We watermark all of our generative media content,
18:04whether it's going through an enterprise setting
18:06or a consumer setting.
18:07And so I think it's kind of important across the board,
18:11both from the research side to the deployment side.
18:13And it's also, I think, another reason why it's so important
18:16that the technology companies,
18:18whether it's the people kind of building
18:19or deploying the models, work with the communities
18:22that are adopting and using those tools,
18:25whether it's an individual or whether it's a company.
18:28Because the things are changing so quickly,
18:30and we're still figuring out how,
18:32what the technology is best used for and how to deploy it.
18:34So I think this is gonna be a live conversation.
18:36It's not something that any one of us
18:37are gonna figure out in a vacuum.
18:38And that's why that collaboration,
18:40that's another reason why that collaboration
18:41is so important to us.
18:44Yeah, I think my perspective is a little different,
18:46as I said in the beginning.
18:48A lot of these conversations in the media
18:50and between people talking about AI
18:53are very much always focused a lot about generative AI
18:56that could be copyright infringement,
18:58that could be problematic, so to speak.
19:00And so I'm going back to the data science part of it,
19:04which there's so many things already done with AI,
19:07even in other areas, but in data science and data analytics
19:11that are, I would say, really just assistive,
19:15and they're really just helping us to do better work.
19:17And I brought up recommendation systems earlier.
19:21Those are very dear and near to my heart,
19:23where it's all about understanding you as a consumer
19:27and giving you the content that you're interested in
19:29and also creating content that maybe we haven't had before
19:32because we just didn't know that that was content
19:34that was interesting and we can analyze that better.
19:36And then on the worker side, too,
19:38is there's, I think, a lot of mundane tasks
19:41that are now being done by AI
19:42that nobody actually wants to do,
19:44but that are just kind of there,
19:45you have to turn through it, that can now be done by AI,
19:47and you can focus on work
19:49that's actually much more interesting.
19:50So we have something called RAG,
19:52which is Retrieval Augmented Generation,
19:54which enables you to actually have your own data sources
19:56connected to an LLM, for instance.
19:59And so now you can actually speak to your data set,
20:01so you have something that's very customized
20:03as a data scientist,
20:04and I know I'm getting a little bit into the detail,
20:06but you can get all the data you want
20:08by just talking to it.
20:09You don't have to go through Excel spreadsheets
20:11and all this churning through it,
20:13but you can get much more insight much more quicker.
20:16And that then kind of leads back
20:17into content production, too,
20:18where I think we're never gonna be there,
20:21well, I shouldn't never say never,
20:22but we're probably not gonna be there anytime soon
20:25where we know what a blockbuster is,
20:26where we know exactly what's gonna be the product
20:29that people wanna see,
20:30but I think we're gonna get much closer.
20:32And I think that's also for the creators very helpful
20:35to know like, okay, these are the areas
20:37that are interesting,
20:39and so when you wanna get projects greenlit,
20:41that might be helpful.
20:43So I think overall there's a lot of positive in it,
20:46and we're just very much always focusing
20:48on the things that are problematic,
20:50and I think there's areas that we're using AI,
20:54we've successfully used AI,
20:56and we're now being able to do a lot more
20:59and a lot more efficiently,
21:00and quite frankly a lot more interesting
21:03for the people that are working on it.
21:06Well, as we all know, it's evolving so quickly
21:08that we're seeing AI announcements in the news
21:10on a daily basis at this point.
21:13A notable one earlier this week was Stability AI,
21:16which is the company behind Stable Diffusion,
21:18announced that Jim Cameron had joined
21:20its board of directors,
21:21and in the announcement he said that,
21:24he described the intersection of generative AI and CGI
21:28as, quote, the next wave.
21:29What is everyone's thoughts on this?
21:33I think, look, we've heard of work from home.
21:40I think there comes a day where we have the phrase
21:44act from home, you know?
21:46So you can do your motions and your interactions
21:50in a lot of your filming from a confined space,
21:54and be able to transpose that into a different world.
21:58Gaming's doing that already a lot,
22:02and the ability, Moov AI is one such company
22:05that has built technologies that can support
22:09that kind of behavior.
22:10You have about four iPhones all around you,
22:14and you can act out and you can change the background
22:17to fit the theme of what it is,
22:19and so you can clearly see how this is going into a space
22:22where the next wave of TikTok videos
22:26are just gonna be a step above
22:28in terms of quality of creation,
22:30and I'm very excited about that.
22:32And the intersection of CGI, AI,
22:37and Gaussian splatting is gonna bring a wave
22:41of new opportunities for creators
22:43that I think we can only have dreamed of in the past.
22:47I'm just gonna say it's another good example of industry
22:51partnering together both on the kind of creative side
22:55as well as the technology side,
22:56and I think that's, like I said before,
22:58I think that that's a good thing.
23:00And I also think it's just kind of another evolution
23:03of technology in film.
23:06I mean, when I was an undergrad film major,
23:08we cut on physical film, 16 millimeter,
23:10and then I saw, you know, then NYU brought in Avids,
23:13and then I saw it kind of get democratized with Final Cut,
23:16and now obviously, you know,
23:17there's all of the kind of app-based tools today,
23:19and I think, you know, AI is another step
23:22in that kind of journey for how technology
23:24has been kind of integrated productively in production,
23:28and I think that's, you know,
23:29to me it's another point on that curve.
23:32Yeah, I think that's right.
23:34I think I would add that there's still a lot of work
23:37to do with diffusion models and AI in general
23:41to really make them useful from a creative perspective.
23:43I mean, they'll say, you know,
23:45a picture tells 1,000 words,
23:47but I posit it takes more than 1,000 words
23:50in a prompt to generate a picture
23:51that actually looks any good.
23:53It actually doesn't work in reverse.
23:55And, you know, in our experience,
23:56creatives think in terms of composition with a camera,
24:01playing music with an instrument,
24:02not typing words into a computer.
24:04Not every creative is a prompt engineer,
24:06as I'm trying to say,
24:07and there's so much variation and so much randomness
24:10to these images that are created.
24:12There's continuity issues.
24:13A character might look dramatically different
24:15one 15-second clip to another.
24:18So full narrative storytelling in these technologies
24:21I think is a very long way off,
24:22again, until they can have the sort of control
24:25that creatives are familiar with
24:27and can work with and create with,
24:29create story with,
24:30and the outputs are predictable, they're useful,
24:33and at the end of the day
24:34actually help generate emotion in human beings.
24:36So if you think about visual effects,
24:38for that matter, animation workflows,
24:40it's really about AI tools and technology
24:42to help speed the ideation process.
24:44You might have a tool, for example,
24:45that might generate what a generic canonical microphone
24:50will look like, right?
24:50But it might not have the variety logo
24:52or it might not have the mouthpiece or whatever.
24:55But the head start of having a 3D model as a microphone
24:58is actually very helpful for speeding things up.
25:00But sort of whole cloth creating images and video
25:03I think is a long ways off,
25:05at least in my part of creating, you know,
25:07motion pictures and television shows.
25:09Yeah, I think that's right.
25:10I mean, I think we, the story thus far has been,
25:13you know, it's amazing what computers can do.
25:15Like you put in this prompt and this video existed
25:17and it's kind of been the, like what, you know,
25:19wow, how high quality, how creative,
25:20like this didn't, you know,
25:21my idea just kind of got visualized.
25:24And that's been exciting and that's been impressive.
25:26But as anyone who's gonna work closely with creators,
25:28like that's not what the, you know,
25:30that's not the workflow.
25:32It's not just I have one prompt
25:33and I get a video and I'm done, right?
25:34It's a highly iterative workflow
25:37where like controllability and editing
25:39and some of the things that Michael Victoria talked about,
25:41like those are first class citizens in a workflow.
25:44It's not, you know, just, you know, prompt in, output out.
25:47And I think that's, that sometimes gets lost
25:49as we're kind of, you know,
25:50correctly impressed by the technology.
25:52But like if you really get in to understand the workflows,
25:55it's, you know, obviously the technology
25:57has to serve the workflow and we're not there yet.
25:59Yeah, I think at the end, you know, creativity wins.
26:02I always give this example, like I'm a bit of a nerd.
26:05I'm not super creative.
26:06I have all these tools at my hand.
26:07Could I create a good concept for a movie or a clip,
26:11anything that people want to see?
26:12Probably not because I just don't have the ideas for that.
26:15And then, you know, you have somebody,
26:16a friend of mine who's an AI director.
26:18She creates these amazing 30 second clips
26:21just out of thin air and they're creative
26:23and they're wonderful and it's all done with AI.
26:25And it's, we talked about this earlier in the green room,
26:27it's about democratization.
26:29Like she doesn't need much more
26:30than a couple of subscriptions
26:31and her brilliant mind of being creative.
26:33And I think that's what gets lost a lot of times
26:36that it's all about the tools.
26:38It's not, it's still all about the human
26:39and the creativeness of the human.
26:42I love my panelists.
26:43They all love creatives and are all supporting us.
26:45Yay, big win for us, yay.
26:47But at the end of the day, it is about story, right?
26:49And telling that effective story.
26:52And for me, the biggest issue that I've had
26:54is that control element, right?
26:57You can put a bunch of prompts, you can use the seed,
27:00you can use the weights and the various criteria you can
27:02to like try to create that continuity,
27:04but at the end you don't.
27:06And of course I can put camera moves
27:07because you have to be the storyteller technologist now
27:10to navigate this kind of brave new world.
27:13And it becomes very, very challenging.
27:17It becomes very, very challenging to tell that story.
27:19And what ends up happening is
27:21you're almost writing the prompt.
27:23So my first prompt will start with camera pans.
27:26And this is what happens.
27:28The subject is this, and then this is the action.
27:31So I start with that and then I send it,
27:33I press enter, and then it gives me back my feedback.
27:36And by looking at what it gave me,
27:38you actually have to go back and deconstruct
27:40the order of your words to see,
27:42oh, how is this algorithm working?
27:44And once you do that enough times,
27:46you start kind of creating this relationship
27:48where you're co-creating with a computer.
27:51But that's essentially what it is.
27:52So it very much is a two-way road to try to understand
27:55how it works so that you can actually get what you need.
27:59We're running a little tight on time,
28:00so I'm gonna move this along.
28:01So we have a few more topics
28:03that we want to go through.
28:04But one of them is, it goes without saying,
28:06there is a lot of unease in the community
28:09about job security.
28:10What would you tell artists?
28:14If you weren't threatened by Photoshop,
28:19you shouldn't be threatened by AI, right?
28:25The whole suite of tools that I see coming out
28:28is meant to augment your potential.
28:33It certainly is nowhere close to replacing your potential.
28:37And to your point, I'll give you an example.
28:40Last week, I was trying to create my own little video
28:43of a beautiful beach and a sunset.
28:48And it was a five-second clip.
28:50And two seconds in, dolphins started coming in from nowhere.
28:56And all of a sudden, the sun morphed into something
28:58that's, I don't know.
29:00But at the end of the day,
29:02there is this interaction model that you brought up,
29:05which I think is very special.
29:08But at the end of the day,
29:08that interaction model is helpful to augment.
29:14It's its own science in working with the machine
29:19in a way that brings a new level of interactions,
29:22which is not just a bunch of buttons or toggles
29:25or dropdowns.
29:27But it's a pattern of conversation.
29:31And that conversation is through prompts today.
29:34It could evolve.
29:35But the bottom line is, you are always in control.
29:37So if you're not afraid of Photoshop,
29:40don't be afraid of AI.
29:42Any additional thoughts, Katya?
29:44Yeah, I mean, it's kind of going back
29:45to also what I said earlier, right?
29:47The creativity part wins.
29:48And at the end of the day,
29:50if you, we are in an industrial revolution in another one.
29:55I'm sure of it.
29:56And so I think there's also gonna be
29:59a lot of opportunities and jobs
30:00that we don't even know what they're gonna be in five years.
30:03I mean, we already have something called prompt engineer,
30:05which if you told somebody five years ago, that's a job,
30:08what?
30:09We don't know what that means.
30:10And so I think a lot of the things
30:12that we're gonna see over the next few years,
30:14I think the World Economic Forum,
30:15I read a statistic,
30:16they actually expect net positive job growth with this area.
30:22And so I think the job might be different.
30:25We're working with somebody,
30:26we're having, we're not working just with humans,
30:28we're working with humans and maybe AIs,
30:30different kinds of AIs.
30:32But yeah, I agree with the fear of not afraid of Photoshop,
30:35should be afraid of this either.
30:38We only have a few minutes left.
30:39So what I'd like to do is start at the end,
30:41we'll run down and answer,
30:43each of you have a chance to answer this question.
30:45But I'd like to end with each of you
30:47looking into your crystal ball.
30:49And for the next year or 18 months,
30:52what is the biggest challenge or opportunity
30:57or if there's a call to action,
30:59but what would you emphasize for where AI is heading?
31:06Look, I think the time has come for research
31:09to start going deeper into certain verticals, right?
31:14I'll bring up the concept of hallucinations.
31:16I'm sure you guys have heard of hallucinations
31:19as being a problem statement when it comes to AI.
31:22And that is so true if you look at it
31:23from the verticals of legal, healthcare,
31:28logic, rational decision-making, fair.
31:31But as a creator, I wanna riff.
31:35I wanna have crazy thoughts.
31:37I wanna have crazy back and forths and optionalities
31:40for conversations to take off
31:42in multiple different patterns.
31:44So should hallucinations be considered a wrong thing
31:48or is there a variant that research can bring
31:52that can provide food for thought for the right brain
31:54as thought starters for us, for creators to extend out of?
31:58And I think research in that space
32:00would certainly augment the potential capabilities.
32:04And going back to the interaction model,
32:05I think the interaction model itself would set the pattern
32:09for what and how we as creators could riff off ideas
32:14with our helpful AI partners.
32:17That's where I see the opportunity at
32:19as one of the many opportunities.
32:22I'd love to hear all of your thoughts.
32:24I'll try to keep it short.
32:25Well, I'm gonna plug a little bit my company
32:28in the last 30 seconds.
32:30What I'm actually truly excited about is,
32:32NVIDIA, we're so much known for hardware
32:34and we're so much more in the area I'm working on.
32:37Also, I was just on the phone with a customer yesterday
32:40and they talked about how data science
32:42is sort of the little evil stepchild
32:45that nobody cares about.
32:46There's no budgets for it, but it is so important.
32:49And so I'm pounding that drum
32:51that I'm excited about talking about this a lot more
32:54because it is important.
32:56There's a lot of jobs in that area,
32:57people doing fantastic work.
32:59And we're very much investing in it.
33:01And we see ourselves as supporting those efforts
33:04and helping with our technology stack,
33:06with our partner ecosystem, and with our hardware.
33:08And so that's my plug at the end of the session.
33:13I mean, first of all, I think we're gonna see
33:16continued progress in all the models,
33:16like the amazing progress we've seen
33:18over the last several years.
33:19I think that's gonna continue.
33:21So that's kind of the first part of the crystal ball.
33:23The second is we're just entering
33:24in this really exciting deployment phase.
33:26I mean, she talked about taking her notes and docs
33:29for this prep session and automatically
33:31creating an amazing podcast.
33:32That's something we launched like a week and a half ago.
33:35She's already, it's incredible work.
33:38Through a product called Notebook LM,
33:39which we've now brought to everyone.
33:43We've now got millions of people
33:44using our generative media models every day.
33:46I think that's gonna get,
33:48I think that number is gonna get into the billions
33:50over the next several years.
33:51So I think we're gonna kind of continue
33:53to see an amazing progress.
33:55But also, I think from a product perspective,
33:57what's so exciting is we're gonna start,
33:58we're now starting to see like the deployment phase of that.
34:01And I think that's gonna continue.
34:03We're just in the very early innings of it.
34:05As Michael said, a lot of the technology
34:06is still being developed.
34:09There's a ton of work to be done on the research side.
34:11And so this isn't gonna be solved in a year.
34:14But I think it's gonna get,
34:15it's getting a lot of momentum.
34:18I think for me, like AI is like a superpower.
34:20I feel like I put on my Iron Woman suit on
34:22and I'm just like unleashed.
34:24And I feel very limitless.
34:25And I think the future of AI
34:26will create that limitless even more,
34:28especially with graphics cards and CPU power
34:31and data centers and cloud computing.
34:33It's going to enable these narratives to become dynamic,
34:36changing on our preferences, changing on our choices.
34:38And no story is gonna be the same.
34:40Stories will completely become lived experiences
34:43as we add scent as well as we talked outside earlier.
34:47So I think there's a lot of hope and promise in AI.
34:50And anytime you may feel a moment
34:52where you're just kind of pausing
34:53because you're wondering, oh, wow,
34:54is this getting too creepy?
34:55Just remember back sometimes in history,
34:57every time we've taken a giant leap as humankind,
35:00as a society, we've taken these giant leaps
35:02from the agricultural revolution
35:03to the printing revolution, to the industrial revolution,
35:07the digital revolution with computers being accessible
35:09to everybody, internet, how we communicate.
35:11And this is just one more iteration
35:13of the power of the human mind to innovate, create.
35:17So don't be scared of it.
35:20Jump into it and remain relevant.
35:23And Michael.
35:24Yeah, I think that's right.
35:27I'm old enough with enough gray hair
35:29to live through a lot of technology disruption
35:32in the media industry.
35:34I think we're at a moment in time,
35:36I don't know how long this moment in time is.
35:37I could wake up tomorrow and read yet another announcement
35:40that blows my mind.
35:41But I think it's important in this moment
35:43to get really the foundational things right.
35:46We've talked a little bit about labor.
35:47We have to get the labor piece of it right.
35:49We have to get the IP and copyright of it right,
35:52or we're not gonna have an industry
35:53if copyrights aren't effective.
35:55We have to get the ethics of it right.
35:58At the same time, we do have to experiment,
36:00understand what this technology can do
36:01and how it can empower storytellers
36:03and a whole new generation of entertainment
36:06that I don't think any of this room can even imagine.
36:09And we need to get that right too.
36:10So, it's a fun time to be here.
36:13Okay well thank you everybody.
36:14Please join me in thanking our speakers.

Recommended