- il y a 2 jours
Creativity & AI How does IP Survive in The World of AI
Catégorie
🤖
TechnologieTranscription
00:00Hello, good afternoon, my name is Leslie Silverman, and I run New Tech at United Talent Agency in Los Angeles.
00:09As you can see, all across the floor, AI is one of the three major thematic focus areas for Viva
00:16Tech.
00:17I used an AI search engine this morning, perplexity, to come up with some key facts about Viva Tech this
00:24year.
00:25Nearly half of the exhibiting partners at Viva Tech are focused on AI solutions.
00:30So how do marketers best take advantage of these solutions within their organizations?
00:35One thing I want to say at the top before our esteemed panelists introduce themselves is that we are not
00:42going to delve into the complex legal framework around the protectability of AI outputs this afternoon,
00:50simply because there is no global legal standard.
00:54So we are going to focus on that intersection of creativity and AI, and so I will, without further ado,
01:04allow my esteemed panelists to introduce themselves.
01:06So we'll start with you, Jace.
01:07Sure, thank you, Leslie.
01:08Jace Johnson, and I run Global Public Policy and Ethical Innovation at Adobe, and I'm based in Washington, D.C.
01:16Hi, I'm Johnny Geller. I'm a literary agent and CEO of Curtis Brown, which is the UK's largest literary and
01:25talent agency.
01:26Hi, I'm Carla Gazzetti. I'm Senior Vice President of Product Experience and Innovation for Extreme Networks.
01:32We bring connectivity safely throughout the entire globe and actually work with quite a few governments here in France, so
01:39I'm very excited to be here.
01:41And I am Alberto Prado, Vice President and Head of Digital and Partnerships for Unilever R&D, and in that
01:47capacity, I'm leading the digital transformation of Unilever's R&D organization, which is about 5,000 people distributed globally.
01:54Such a fantastic diversity of perspectives on this panel, and I'm looking forward to asking you my deep questions on
02:02this topic.
02:03So, according to the VivaTech Barometer, AI is moving from an innovative technology to an indispensable tool.
02:1288% of executives are looking to invest in some way in 2024.
02:16So, what are the investments and new applications that you each are excited about, and maybe what are the obstacles
02:24that are keeping you up at night?
02:25And I'll start with you, Jace.
02:27Sure. Excellent question about those opportunities.
02:31There are a lot of opportunities.
02:33Our Chief Technology Officer at Adobe spoke on this stage yesterday, and he mentioned something about 100 times more opportunities
02:44to create things,
02:45and 100 times more for people to create them.
02:48And I think that organizations are now starting to realize that lowering the bar for creativity and allowing more people
02:55in is generating this huge opportunity inside of every business for people to get creative.
03:02We used to talk about productivity.
03:04Now we talk a lot more about creativity.
03:07Fantastic.
03:07How about you, Johnny?
03:08Well, I may have a slightly different perspective than my colleagues on the panel, just simply because I suppose I'm
03:16at the beginning of the chain.
03:17I'm with the creatives who start.
03:19They write a book.
03:20They direct a play or write a play.
03:23And I suppose from an agency's point of view, and Curtis Brown is 125 years old, I suppose I'm the
03:31legacy guy on this panel.
03:33But I don't see legacy as a negative thing.
03:37I think legacy for me means expertise and art and culture.
03:42So for me, I look at AI, and I think there's two ways of looking at it.
03:47One is AI as an enhancement or a tool, as a speeding up for publishing and for studios who are
03:55making movies and for directors and actors.
03:57Or AI as a replacement to the creativity.
04:02And the danger there will be a dumbing down and making it very difficult to find the really good writers
04:09and really good creators.
04:11So there are two ways of looking at it.
04:13And I think the creative industries, from the creator's point of view, is in the middle.
04:18I've got to say, there isn't a leaning towards AI with enthusiasm.
04:23On the other hand, I think I'm coming from the point of view of the first category, which is AI
04:29could be a fantastic tool for writers and for actors and directors.
04:33And I can explain why later.
04:35But I think, to be honest, we're at a real revolutionary moment because this is the first technology, in my
04:41view, that could create its own consciousness.
04:45And therefore, make its own decisions.
04:47Whereas before, the revolutions in my industry, publishing, is whether you have an e-book or an audio version or
04:55a paperback.
04:56These are just distribution revolutions.
04:58This is a fundamental creative revolution.
05:01I want to make sure we come back to that.
05:03Arla?
05:04Yeah, I mean, that's a way to leave us on a big note there.
05:08So, I think, you know, it's such a big topic, and that's such a big question.
05:14But I think one tangible example of how we're looking at leveraging AI within Xtreme right now is I oversee
05:20our technical documentation team.
05:22We have over 720,000 pages of technical documentation.
05:25We have a huge portfolio.
05:27And one of the difficulties, and as we talk about, is productivity.
05:32So, we think about how many writers does it take to actually produce that level of documentation.
05:37The opportunity for AI right now is our ability to not only increase the number of pages, but also increase
05:44the amount of feedback we get about the documentation.
05:47Is it right?
05:48Is it wrong?
05:49And then also update that.
05:51So, one of the things that we're doing at Xtreme is we are actually looking at how we write our
05:56JIRA tickets.
05:56Our JIRA tickets are what our product managers create, and it tells the entire team what they should be creating,
06:04and then our design teams and engineering teams build it from there.
06:07We are partnering with another vendor, and we are going to use AI to take the content from the JIRA
06:13tickets to actually develop our technical documentation.
06:16So, we are expecting that we're going to be able to produce technical documentation now what can take us up
06:22to months done in minutes.
06:23The other interesting thing, and part of that, is it takes the role of our writer and makes them not
06:30a writer, but an editor.
06:31And I think that the obstacle that a lot of companies are going to face as they implement AI strategies
06:36like this is that you're really going to see an obstacle around change management.
06:41You have to think about an entirely new workflow.
06:43How the employee approaches what they're doing completely changes.
06:47It's a different framework, and I think that is going to be the obstacle as more and more companies look
06:53at how they best leverage AI.
06:55Fantastic.
06:56Alberto.
06:56I think the beauty of this panel is that we approach AI from different angles.
07:02So, I come from research and development, so I'm really interested in product innovation and how we can leverage artificial
07:08intelligence to make better products.
07:10I mean, fundamentally, Unilever is about creating products that make life better and easier for about 3.4 billion people
07:19that we serve on a daily basis.
07:22So, we want to make sure that our products are the best out there.
07:25And when you look at artificial intelligence, we try to apply it across the entire product innovation lifecycle.
07:32You know, how can we predict consumer needs?
07:35How can we predict superiority of our products?
07:37How can we formulate or reformulate our products at speed?
07:43How can we optimize the way that we scale up our innovations in manufacturing and distribution globally?
07:49So, that's where AI is already today having a big impact in terms of how we automate things and the
07:54type of decisions that we make.
07:55Now, from an R&D, from a research and development perspective, we believe that AI is giving us an edge
08:01when it comes to delivering scientific breakthroughs
08:04that can meet consumer needs in ways that we believe are superior to competition.
08:12And that's where, you know, because we see also a change in consumer behavior as well.
08:17You know, consumers want to understand more and more what the science is behind the brand.
08:22You know, is this a good product for me?
08:24Is this a good product for the planet?
08:25So, and that's where we believe R&D and the application of AI within an R&D context can really
08:32influence purchase decisions and drive growth.
08:35And that's what we're focusing on.
08:36It's amazing.
08:37So, you've each talked about efficiency.
08:40Efficiency being sort of a key growth factor.
08:44What are some other competitive advantages of AI?
08:47And I think, also, if you could talk a little bit about best practices.
08:51If I'm sitting in the audience today and I'm thinking about where to begin on behalf of my company or
08:57as a marketing initiative,
08:59what are some best practices to sort of begin this sequence of ideas?
09:04Jays?
09:05You know, our first thought when we think about that question is personalization.
09:09There's this opportunity to make things personal for people, to allow people to tell their story.
09:14You know, Unilever is an excellent example of telling a story behind a brand and explaining and making it personal
09:22to the individuals.
09:23You can do that at scale with a generative AI tool.
09:27And then when you do that, you need to make sure that you've got in place systems that is making
09:32sure that it doesn't become ultra-creepy.
09:35You know, that you can now use it, you can have humans involved, and I know that big companies that
09:41do this well do,
09:42have humans involved to make sure that you test it ethically.
09:45It's one of the programs that I'm responsible for at Adobe, where we test it for ethics, we test it
09:50for bias,
09:51we test it for all these things that humans want when they see these types of images and marketing campaigns.
09:58Fantastic.
09:59Johnny?
10:01Best practice, I think, is trust.
10:03I believe that the creator needs ownership and control.
10:09And the moment they give that to either a publisher, a streamer, a tech company,
10:14there has to be a trust that that value is going to come back to the author, and there will
10:19be some control.
10:20As we know, OpenAI has sort of, in a sense, opened a new conversation about where did they get the
10:27training models.
10:28They scraped the internet and, in some views, stole the copyright of writers.
10:35So it is my job as an agent to protect the copyright of these creators,
10:41because I fundamentally believe that creativity, art, and culture is unique,
10:48and we need to celebrate the uniqueness.
10:50And if we have AI that essentially dilutes it, then what is that going to mean for our enjoyment of
10:58books, films, music, and art?
10:59So it's quite a philosophical, but having said that, there are many areas within agenting and publishing and movie
11:07that you really can move forward.
11:11You know, and there's, I'm seeing it already, you know, there are scanning devices for actors and over their voices
11:17so that they may well, beyond their death, be able, their families and their estates, be able to carry on
11:24licensing their image.
11:25And in fact, it's already been done.
11:27I mean, Harrison Ford in Indiana Jones and Robert De Niro in The Irishman.
11:33And we heard this week about Scarlett Johansson stopping her voice or an imitation of her voice.
11:39So there's, it's really very now, this issue.
11:43And so practices, best practices is crucial.
11:47And for me, it's about understanding the nature of creativity, which sounds a bit grand, but it is about the
11:55unique,
11:56unique, for example, I don't know how many people here saw Baby Reindeer on Netflix or One Day.
12:03Those were two projects we were behind and they could never have come out of AI.
12:10They came out of those authors and those creators' experiences over years in that medium at that moment.
12:18So for me, AI is about how do we work together and in some way create new things together
12:24or allow the author to control their version of AI?
12:30Well, double underline from both of you on the human element.
12:34So, Carla, how about you?
12:36So I think there are a lot of different ways to slice this question.
12:39I'd say from a business perspective, one of the things to think about when you're thinking about AI
12:44and best practice is thinking about where your data is coming from.
12:49Everybody has heard before data is the new gold.
12:51I really hate that phrase.
12:53It is not the new gold.
12:54The new gold is actually AI, and I think a lot of businesses out there are really concerned.
12:59Does that mean I have to have everything digitized?
13:01All of my data is a mess.
13:03Everybody, for as long as that phrase has been around for at least 20 years now,
13:08all of us have not been at home cleaning our data.
13:11I know that's not what I like to do on a Friday night.
13:14But the point is you don't have to standardize all of your data.
13:18When you're thinking about building an AI that is specific for your customer, for that personalization,
13:24choose the parts of the data that you think are going to be the most relevant and focus on those
13:29areas.
13:30You don't have to completely digitize everything in your back office.
13:33that will take you 15 years and tens of millions of dollars, so just don't do it.
13:38So I think from a business perspective, that's one thing.
13:40From a creative perspective, I would say get comfortable with AI and lean into it.
13:45Educate yourself.
13:46Because I think one of the things that's out in the media right now is that people,
13:51and we tend to position AI as this technology that we have no control over.
13:56But actually, generative AI is a technology that we have to teach, that you actually have to train.
14:02And you can train based on the tone of voice for your company, for the tone of voice for your
14:07brand.
14:07That's where the personalization comes in.
14:09And also the control where the data comes from.
14:12I would also recommend that, I know there's a lot of, people are very scared about transparency,
14:17and how do we know these foundational models, where are they pulling this data from?
14:22Stanford has an institute for human-centered AI, and they actually publish a document and a study every single year
14:32about the transparency of the major foundational models in the world.
14:36And they do it on a multiple of vectors.
14:38One of them is transparency, design, what's the bias within the foundational model.
14:43Educate yourself on this, and it becomes more comfortable to understand how to leverage it and lean into it.
14:50Can I just ask you a question on that?
14:51About labeling, would it be useful at some point to have, like, you know, like GM foods are labeled in
14:57supermarkets,
14:58would we be able to be at a point where we'll say to people,
15:01this is partially AI-generated or wholly generated?
15:05Absolutely. I think that is something that we're going to have to look at as an industry.
15:09And one of the things that, you know, we partner with Google and Intel and also AWS,
15:14we look at how can we make sure that content is more transparent,
15:18and I think that's where we're going to end up going, is being able to label that.
15:22There's another company that I've worked with called Credo.ai.
15:25One of the things that they are doing is building a model that will be able to pull
15:30and recognize all the sources from your answer, in your answer.
15:34So you'll be able to understand exactly how that answer was developed
15:38and where the information was pulled from.
15:40So it's a really exciting space to see where we go next.
15:43I'm going to come back to you in a moment, Alberto,
15:45but Jace, I know Adobe has a framework for this,
15:48and I wanted to see if you might answer Johnny's question.
15:52Yeah, and I'm glad you raised it.
15:54About four years ago, Adobe worked with Microsoft and the New York Times
15:59and several other companies to set up a global standard for nutrition labels
16:03for images, video, audio, and documents.
16:06All right.
16:07Contentauthenticity.org, if people want to check it out,
16:10it's a, to create a global standard that you can tell whether things,
16:14when you turn on the content credential,
16:16you can tell whether it's wholly AI generated, partially AI generated,
16:20what are some of the AI touch-ups that have gone into it.
16:23You can have things like identity, source, you can compare and contrast,
16:28you can take an original and use a slider and go back and forth
16:31from the original source to what did it turn into.
16:33That's a whole functionality built for the internet
16:35so that consumers have a way to understand what they're seeing.
16:39And it feels like it might be really powerful as applied to clients.
16:43Yeah.
16:43Okay, so Alberto, I actually want to ask you a bit of a more nuanced question.
16:47So you're an R&D, so if there are marketers in the audience,
16:52what are some of the real-life experiences you're having in the R&D area
16:57that potentially could apply to, say, I'm a marketer
17:01and I'm trying to understand benefits and use cases here?
17:04Yeah, absolutely.
17:05So we work with marketing very closely together, obviously,
17:08when it comes to product innovation.
17:12And examples of how we apply generative AI
17:15that actually resonate with marketeers,
17:18I'll give you three cases.
17:21The first one is about new claims generation, for example.
17:25When we think about what are the innovations,
17:28what are the products that we need to work on
17:29that if we were to develop them and manufacture them at scale
17:33would give us product superiority.
17:35This is becoming more and more a data-driven approach
17:37and a generative approach.
17:39So we start thinking about, okay, you know, what are the claims?
17:42What are the benefits that people are looking at?
17:44What are the packed designs that people would be interested in?
17:49And a lot of it is becoming more and more machine-generated.
17:52Obviously, the human is in the loop
17:54and ultimately we need to make our calls.
17:56But the amount of support that we get from generative approaches
17:59that feed from both proprietary data
18:02as well as external data is quite impressive.
18:06Now, if you go a little bit more downstream,
18:09generation of chemistry.
18:10So not just generation of claims that I was talking about,
18:12it's generation of chemistry.
18:13So we use generative AI LLMs to feed them chemistry
18:18and get chemistry out.
18:19This is LLMs that actually talk the language of chemistry.
18:22And more and more, the way that we formulate new products
18:25is going to be supported by generative.
18:28And then finally, you know, knowledge management as well.
18:31So we've developed a custom co-pilot together with Microsoft.
18:36It's almost like a small language model
18:38that actually goes really deep into our scientific knowledge.
18:41we've fed it with 150,000 documents,
18:45100 years of history of formulation that Unilever has.
18:48And now you can query it, you can interrogate the data,
18:51you can synthesize it, you can visualize it
18:53at no time, at super fast speed.
18:57So imagine the level of support that our scientists,
19:00together with working with our marketers,
19:02can actually get in terms of thinking of new innovation ideas
19:05and making them a reality.
19:07But maybe conclude,
19:09because we were talking about competitive advantage,
19:10we believe that also this technology
19:13is somewhat leveling the playing field,
19:15because everybody has access to this technology.
19:17So what is it that will differentiate
19:19the companies that win
19:21versus the companies that are a bit left behind?
19:24We do believe in R&D that it's a combination of two.
19:27It's your domain expertise,
19:28the ability for you to use these tools
19:30like no one else can,
19:32and interpret the results of these tools
19:34and interact with them
19:37in ways that maybe other people
19:39or other companies may not be able to,
19:41who may have less of an expertise.
19:43And also the richness, accuracy, and reliability
19:46of your own proprietary data,
19:49that you can fit these models
19:49and actually get to proprietary results
19:52that are not accessible by all companies.
19:54Fantastic.
19:56So we're all talking about
19:58turning up the volume on efficiency
20:00and the ability to look at our data,
20:02but how do we make sure
20:04that AI is creating something unique,
20:07not just more stuff, faster?
20:11Jace, I'll start with you.
20:12Yeah, and I wanted to just quickly address Johnny's
20:16and then I'll come to yours.
20:17He had talked about web scraping,
20:19and I just wanted to state that
20:20there are AI systems.
20:22Adobe has built one.
20:24It's a generative AI system
20:25that isn't web scraped.
20:26It's composed all of licensed content.
20:29It's made for artists by artists,
20:32and it's a wonderful tool.
20:34It's called Firefly.
20:36But to your point,
20:38I think that this idea
20:40of making something commercially safe
20:42has real value in the marketplace.
20:44This commercially safe concept
20:47of enterprises know
20:49that they can use it
20:49and not have to face lawsuits
20:51even at these early stages
20:53of generative AI,
20:54but also consumers at home
20:56can get into these products,
20:58play with them,
20:59and not feel like they're at risk themselves
21:01for the things that they'll be creating
21:03using these tools,
21:04and that provides a real advantage
21:05for everybody in the market.
21:07Absolutely.
21:09I'm not sure how to answer this
21:10because in my experience,
21:13writers come to the point
21:16where they're ready to show a publisher
21:18and then show the readers
21:19through a whole network of experiences
21:21and I can see the efficiencies.
21:26Absolutely.
21:26In fact, there was a piece
21:27in The Spectator last week
21:29or a few weeks ago
21:30by a writer
21:31who put their novel
21:33through Gemini Pro,
21:35I think it was,
21:36to get an edit,
21:37and they got an editorial report
21:38which was actually better
21:40than their human editor's report.
21:42So I don't know whether that's good or bad,
21:44but in some ways it's efficiency
21:45because we might as well get the best
21:47out of this,
21:48you know,
21:49we only use a tiny part of our brain
21:51and if AI can expedite the bits we can't,
21:54but it's got to come from our brain.
21:56It's about who makes the choices,
21:58which is why I keep going back
21:59to perhaps the philosophical things
22:01and I'm really pleased to hear
22:02about your ethos
22:04because I think artists will gravitate
22:06towards people like you
22:08than they will against those ones
22:10who are basically seeing them
22:12as a resource
22:12to just increase distribution.
22:18You know,
22:18it isn't a distribution game.
22:20It's actually a quality game.
22:21That's the,
22:22you know,
22:23it's a 125-year-old agency.
22:24We have people from John Steinbeck
22:27to Winston Churchill
22:28to John le Carré
22:30and the only reason
22:30that those people are still relevant
22:32is because they wrote good books
22:33and what I worry about with AI,
22:36generative AI,
22:38unless it's really done cleverly
22:39and with collaboration,
22:43it could reduce that possibility
22:45because there are too many
22:47bad books out there.
22:48There's too much bad art
22:51and I think we're all in the market
22:53for making AI better art
22:56than worse.
22:58It also increases, though,
23:00the importance of curation.
23:01So, you know,
23:02your role, Johnny,
23:04at Curtis Brown Group,
23:05you're having to decide
23:07the good from the bad,
23:08the real from the fake,
23:09the true from the false,
23:10and so you become more important,
23:13in fact.
23:13But the only,
23:14I totally agree with you,
23:15I would say that,
23:16but I also,
23:17the flip side
23:18is the discoverability,
23:20and I'm sure you have this
23:20in all your industries,
23:22is how on earth do you find things
23:24when everybody is releasing material,
23:27but it tends to rise to the top.
23:29I remember when Amazon
23:30released their self-publishing program,
23:33it still worked
23:35and people still read books.
23:36Right.
23:37I think,
23:38so I think I might take this
23:40in a bit more
23:41of a philosophical direction,
23:43but I love what Roberto said
23:44because just because
23:47you have AI
23:48doesn't mean that
23:50it's going to do
23:51your homework for you.
23:52You still have to know
23:53whether or not
23:53that homework is correct.
23:55And I think that
23:56there is a concern
23:57about whether or not
23:58what's the quality
23:59that's going to be produced
24:00out there,
24:00but what if we looked at AI
24:03not as an opportunity of,
24:05oh, I don't have to do
24:06that work anymore,
24:06but rather,
24:07I don't have to do that work,
24:08and now I have the opportunity
24:10to think bigger.
24:11That is really what
24:12I think is the opportunity
24:14with artificial intelligence.
24:16As a small example,
24:18our technical writers
24:18have the opportunity
24:19to grow into editors.
24:21You have to become the teacher.
24:23You're no longer the student,
24:24and that challenges us
24:26and gives us an opportunity
24:29to think bigger,
24:31think more creatively,
24:33spend more time
24:34thinking creatively.
24:35I think, you know,
24:36I'm going to go out on a limb
24:37and say maybe creatives
24:39are actually going to
24:40to get better.
24:41Reality TV is here.
24:42Everybody loves a little bit
24:43of, you know,
24:45reality TV on a Friday night
24:47or whatever it is,
24:48but I think that really,
24:50rather than us looking at,
24:51well, how is this going,
24:53this takes something
24:54away from me,
24:55but what if it is
24:56a gift of an opportunity
24:57to think a bigger thought
24:59or think more creatively
25:00or collaborate more
25:01so that we can create
25:03bigger and better
25:05ideas in chemistry,
25:07writing,
25:08television,
25:09and all sorts of things
25:10because we're able
25:11to spend more time thinking
25:13and less time doing.
25:15That's fantastic.
25:16Alberto.
25:16So I guess from our perspective,
25:18we're really passionate
25:19to unlock scientific creativity,
25:22which is human-led,
25:24as you can imagine.
25:25So we have a two-prone strategy
25:28to do this.
25:30As you just heard,
25:31that we're interested
25:33in automating repetitive tasks,
25:37you know,
25:37transactional workflows
25:38that someone else should do.
25:40You know,
25:42they're relatively easy
25:43to automate.
25:43They consume usually
25:44a lot of human resources,
25:46so our researchers hate it.
25:48So if you can automate it,
25:49you can free up bandwidth
25:50for people to do creative work.
25:53Now,
25:53when it comes to driving
25:54scientific creativity,
25:55we believe we can boost it
25:57even more
25:57through the use of AI,
25:59AI as a tool,
26:00where the human
26:01is not just only in the loop,
26:03they're the orchestrators of it.
26:05So humans,
26:06with all technology waves,
26:08it's been more or less
26:09the same thing.
26:09Humans need to position it
26:10themselves on top
26:11of the technology
26:12and then use it as a tool,
26:14you know,
26:14to amplify impact.
26:16So the humans,
26:19you know,
26:19will continue to set
26:20the priorities,
26:22the direction,
26:22vet the outcome,
26:24iterate further,
26:25and that's why we believe,
26:27you know,
26:28these tools can be
26:30augmentative
26:30of scientific creativity,
26:32and that's something
26:33that we've been already
26:33experimenting today.
26:36And frankly,
26:36I'll tell you something,
26:38people love it,
26:39the fact that they don't have
26:40to do any more
26:41lab experiments,
26:42because it can be done
26:44through computational models,
26:46and they get time
26:46to think of
26:47what is the next big thing
26:49that I should be,
26:49you know,
26:50innovating on.
26:51So I was going to ask
26:52a quick question
26:53about trust
26:53and sort of discernment,
26:55because, you know,
26:56it's easy to paint
26:57a broad brushstroke
26:58across all companies
26:59that are using
27:00generative AI today
27:01and say they're
27:02scraping the internet,
27:03I heard that today.
27:05So I guess
27:06for whomever
27:08wants to take this question,
27:09what is the best way
27:11to diligence
27:13a tool
27:14when you approach
27:16a tool
27:16to understand
27:18what is the source data
27:19and is this
27:20an ethical tool
27:21and how do I
27:22trust this tool?
27:24So maybe,
27:24Jace,
27:24you could start.
27:25I'd be happy to.
27:27I mentioned
27:27that I manage
27:28a team at Adobe
27:28called the
27:29Ethical Innovation Team.
27:30It's really filled
27:31with people
27:32who are meant
27:33to break our AI
27:34before it goes
27:36out into the wild,
27:37and they're really good
27:38at breaking it.
27:38You know,
27:39they are really good
27:39at trying to get it
27:40to do things
27:41that it shouldn't do.
27:41They follow the acronym
27:43ART.
27:43It's accountability,
27:45responsibility,
27:46and transparency.
27:47And that accountability
27:48piece is us being liable
27:50for what we put
27:50into the marketplace,
27:51and that transparency
27:53piece is we'll show
27:54consumers or enterprises
27:56or whoever is
27:57buying those tools
27:58as much as they need
27:59to know to feel
28:00comfortable about
28:00using that tool
28:01and then take their
28:03feedback for improvement.
28:04We want every organization,
28:06even whether they're
28:07small or large,
28:08to have people in it
28:09who are practicing
28:10this art.
28:11They are trying
28:12to break AI systems
28:13and then providing
28:14feedback to those
28:15who make it
28:16so that they can get
28:17a comfort level
28:18with what's going on
28:19with a tool
28:20that they're using
28:21that runs their business.
28:22Right.
28:23So Adobe Trustworthy,
28:24where else should I
28:25be looking for,
28:26you know,
28:27language that helps
28:28me understand
28:29whether I should
28:30be using a tool
28:30and it's safe to use?
28:32Well, I think I know
28:33Microsoft has also
28:35with their tool
28:36they already disclose
28:37where their data
28:38is coming from
28:39and how that answer
28:41has been cultivated
28:43in the response
28:44for the AI.
28:44I think you're going
28:45to see that more
28:46and more from the
28:46big four
28:47as it becomes
28:47more and more
28:48popular leveraging AI.
28:51Transparency is
28:52at the heart of this
28:53and I think that
28:54we're going to see
28:55standards start to develop
28:58but for right now
28:58I would recommend,
28:59you know,
29:00like I said,
29:00the Stanford Institute
29:01for Human-Centered AI,
29:03they produce a report
29:05to keep on looking
29:06for the industry
29:07to produce more
29:08and more transparency
29:09reports because I think
29:10this is top of mind
29:11for everybody
29:12out there right now,
29:13all companies,
29:14individuals,
29:15consumers.
29:16Fantastic.
29:17Okay, well,
29:18I promised a speed round
29:19at the end.
29:20So, you know,
29:21I was thinking about
29:22how much FOMO
29:23there always is
29:24with respect
29:25to a new technology
29:26and Gen AI
29:27is certainly
29:27no exception
29:28to that rule.
29:29So, I want to ask
29:31each of you
29:32to fill in the blank.
29:33If you're not blank,
29:36you're already late
29:37with respect
29:38to generative AI.
29:39If you're not doing
29:41fill in the blank
29:42in your daily
29:43kind of practice.
29:44So, Jace,
29:45I'll start with you.
29:47I'll fill in the blank
29:48with Photoshop
29:49because we were
29:51talking earlier
29:52about the wow moment.
29:53What was the wow moment
29:54for generative AI?
29:55As someone who could
29:57never use Photoshop,
29:59until it last year
30:00when it got
30:01generative AI in it,
30:03I now became
30:04a user of a tool
30:05I had been aspiring
30:06to use for years.
30:08So, that's my
30:10fill in the blank.
30:11Fantastic.
30:11Johnny?
30:12I suppose
30:13my one is obvious.
30:15If you're not
30:15reading good books,
30:17you're not going
30:18to do anything.
30:18I think we're all
30:19in the same business
30:20where it comes from
30:22the original creator
30:24and we have to cherish
30:25and protect
30:26and defend them.
30:28I would say
30:29the last thing
30:30to think about
30:30is if you are not
30:31understanding AI,
30:32you're already behind.
30:33You need to go out there
30:34and educate yourself
30:35because every business
30:37that knows how
30:39to leverage AI
30:40is going to outproduce
30:41and be more productive
30:42and more effective
30:43than any business
30:44that's not leveraging AI.
30:45So, if you aren't
30:46already learning about it,
30:47you're behind.
30:49I'm just going to ask you
30:50a follow-up on that.
30:51What's one thing
30:52we can each do today
30:54to start educating ourselves?
30:55What's one experience
30:56that we could have today?
30:58One experience today,
31:00I don't work for this company,
31:02but I'm going to give them
31:03a shout-out anyway,
31:04is IBM's Watson.
31:05They produce a great video series
31:07that explains RAG,
31:09it explains how AI works,
31:11it's very easy,
31:12it's digestible content,
31:13it's just eight minutes,
31:14you can go on YouTube,
31:15it's free,
31:16go and educate yourself
31:17on how models work.
31:19Thank you.
31:20Alberto.
31:20Yes, so one thing
31:21that I personally do,
31:22which actually I love,
31:23is I tend to go
31:24for long walks
31:25and I listen to my books
31:27via audio
31:29and a long time ago
31:30I downloaded,
31:31I mean,
31:31most of you have it already,
31:32sort of chat GPT
31:33on your iPhone
31:34and I just talk to it.
31:35I just go with my headphones
31:37for long walks
31:37and I just ask it things
31:39and then I have conversations,
31:40I mean,
31:41with it
31:42to learn about new topics.
31:44I'm speaking at a conference
31:45in a couple of weeks
31:46and I was just like
31:47educating myself
31:49on the topic
31:49and it's great,
31:51it's absolutely amazing.
31:53Every time,
31:54it's kind of tailored,
31:55bespoke to your needs,
31:56obviously,
31:56because you're asking questions
31:57but you're having
31:58a dialogue with it
31:59and you're having it
32:00in natural language
32:01and you're speaking
32:01to someone really
32:02and I found it always
32:03mind-blowing.
32:04It's relatively basic
32:06as a use case
32:06but it really
32:08always surprises me.
32:10That's fantastic.
32:11I had some conversations
32:12with AI this morning
32:13and I will tell you
32:13it wasn't as fascinating
32:14as talking to each one of you.
32:16so thank you
32:17for all of your
32:18generous contributions
32:20this afternoon
32:21and thank you to Viva Tech
32:22and enjoy the rest of your day.
32:24Thank you.
Commentaires