Want to unravel the mysteries of Generative AI and its impact on the imaging industry? In this episode of the Dead Pixels Society podcast, Gary Pageau interviews expert, Hans Hartman, the chairman of Visual1st. Discover the transformative difference between Generative AI and traditional AI, and how AI can generate media from simple prompts.
Hartman also ventures into the complex arena of digital rights in AI-generated images and a hard look at the ethical dimensions surrounding AI-generated images and the legislative and industry efforts to counterbalance the possible negatives of this groundbreaking technology.
Hartman also offers a preview of the upcoming Visual1st Conference. Listen in as Hartman provides a brief overview of the program, including a dual keynote by him and Alexis Gerard, a pioneer in digital imaging, covering "Profiting from major technology transformation - How about AI?" He also describes fireside chat sessions featuring industry leaders such as John Fisher, head of engineering for Google Photos, Ilka Demir from Intel's Trusted Media Group, and Tim McCain of PhotoLynx/ImageQuix. The conference promises to be a rich feast for anyone interested in the imaging industry.Mediaclip
Sign up for the Dead Pixels Society newsletter at http://bit.ly/DeadPixelsSignUp.
Contact us at email@example.com
Shout out to podcast supporter Keith Osborn of Memory Fortress for becoming a paid subscriber.
Visit our LinkedIn group, Photo/Digital Imaging Network, and our Facebook group, The Dead Pixels Society.
Leave a review on Apple and on Podchaser.
Interested in being a guest? Click here for details.
Hosted and produced by Gary Pageau
Edited by Olivia Pageau
Announcer: Erin Manning
Welcome to the Dead Pixel Society podcast, the photo imaging industry's leading news source. Here's your host, Gary Pageau. The Dead Pixel Society podcast is brought to you by Mediaclip, Advertek Printing and IP Labs.Gary Pageau:
Hello again and welcome to the Dead Pixel Society podcast. I'm your host, Gary Pageau, and today we're joined by Hans Hartman, the Chairman of Visual1st, which is executive-level leading conference for the imaging industry. Hi, Hans, how are you today? I'm doing well. How are you Doing great, Hans, it's always great to see you. First, let's talk about some of the trends that are happening in the tech side of the industry, because at last year's Visual First conference, generative AI broke out into the high level awareness. For most people in the industry it's been burbling under the surface. Ai and imaging, I don't have to tell you, has been around in some form or another for quite some time actually, but this generative side burst forward. Then, soon after that, it broke out into the general business media and people started talking about it. For those who have somehow been under a rock for the last year or so, can you say what is generative AI? Use some of the leaders in that space so people can understand it.Hans Hartman:
Yeah, maybe take one step back. Ai in general, generative AI, is part of it. Ai in general is a way to make connections between annotated, human annotated data. For instance, you can have humans that say hey, these are photos, they have a ball in it, so I'm annotating the ball. So you get a training set of human created content, or at least human annotated content, that's fed into a neural network system and that then outputs. It learns the statistical relationships between, say in this case, an image and what's in that image, and it outputs it to large and large numbers of images, photo videos, et cetera, and makes it very, very easy to make connections at scale. Whether that is image recognition or workflow automation, it's all about learning something with a limited set of data and then extrapolating so you can use it for other data. So how is that different from traditional programming? It's basically a black box. We don't even know how that AI system defines that there is a ball in an image. You can also not modify the learnings from that AI system. You just throw a lot of data into it and basically it exports a lot of other data, and you can do that at scale and very finite.Erin Manning:
That's AI in general and in our industry. We've had a conference of very early on innovators six, seven years ago where the main use case was image recognition, Telling what is in photos instead of doing the manually. Through metadata you can automatically know what is in the name, so that's a well understood.Gary Pageau:
So, for example, it may see there's a face and then later on it knows who that person is and may ascribe a tag to that face. That would be an example of that.Hans Hartman:
Exactly Without a human needing to write down this. This is Gary in that face or whatever. All right. So then the new kid on the block and that's really relatively new, maybe a year, a year and a half at this point is generative AI and that is some form of it. There are also things that don't automatically as a black box, but it is basically. You have a prompt, and it could be a text, it could be a photo, it could even be a video. So you have a specific prompt that goes into the AI system and then the AI system also, having learned these things through training, sets, then outputs in a particular media format. So classic example is you have text like describe whatever a dog walking on the beach on two legs, or some crazy text. The text is the prompt, the output is a photo that shows that dog walking and that's being created by the AI.Gary Pageau:
It's not going on and finding a stock photo of a dog and putting it in a beach picture. It's actually creating that. So you may not know what the dog. You haven't specified the dog's color, so it could be a black dog, could be a white dog, could be a tan dog, could be a great day, could be a poodle. You have no idea unless you specified in the prompt.Hans Hartman:
Correct. Yeah, so you create original output based on a specific prompt. Now, the interesting thing, it doesn't have to be text as the input prompt and a photo as the output prompt. It could also be a photo and an existing real photo. That is a prompt and that is then turned into another photo, or it can even be turned into an animation or a video. So a photo could even be a prompt to create text or a video, for instance, could also be a prompt to create other videos, could be a video that creates photos, could be a video that creates text. You know, transcriptions is a very good example of that. So, all these prompts for input and all these output media types and the beauty of it is it can be done very, very easily and you can do that scale so you can have whatever 500 versions of a sweater. So you make a prompt, maybe a photo of a sweater. In an e-commerce situation you're going to find the variation, with different colors, different sizes, by by specifying additional prompts to that image. So there's a lot of scalability advantages and also ways to make tweaks to a single photo or video for specific segments or specific use cases, so you can have customized marketing, segmented marketing, even personalized marketing, for instance.Gary Pageau:
So, theoretically, an ad being shown to someone could be an AI image of a product that doesn't exist yet until the consumer orders it, and then it could be produced. It's being shown because the AI knows I'm a middle-aged man and I like the Detroit Lions, so it's going to create a jersey that I might appeal to me.Hans Hartman:
Yeah, so it could have to that knowledge that comes from a database or whatever way that it knows who you are middle-aged man Bolt in your case. It knows that and that becomes the text prompt.Gary Pageau:
He needs a hat. He needs to protect his bald head. That's what he needs. You can add that to it in AI, that's not a problem. There you go. So everyone's heard obviously, if they've been watching anything in the news the scary stories about chat GBT. But what are some of the ones in the imaging world that are creating some of this generative content?Hans Hartman:
Yeah, well, the governments of them open AI, stability AI. A lot of folk and video app developers use stability AI because they have open ways of using their technology. Mid-journey is probably on the high-end side of things, creating really nice images. And then even a company like Adobe with Firefly, which is a technology.Gary Pageau:
I was going to mention that.Hans Hartman:
Yeah, that's also being productized in Photoshop and other Adobe products there. That in itself is a very interesting initiative because you can combine traditional photo editing capabilities with generative AI aspects in the same application. This is Photoshop, so you have an existing photo. Maybe it was zoomed in too far. You want a little bit more of the background beats on there, and then they call it generative expand. Other people call it. What is it? De-cropping or outpainting is what open AI causes. So the different ways of saying it, but you add a generative aspect to it where the input prompt is the photo and you let it expand on that. I think there are two takeaways of what the Nobis do with Firefly. One is it's a very innovative way of combining traditional photo editing capabilities with generative AI, image creation or editing. I think the second part there is interesting that Adobe Ghost is round and is not afraid of a setback by traditional photographers, so I think I'm also identifying that that image was created using generative AI in a metadata, so they're really on the forefront of trying to protect the digital rights of their users.Gary Pageau:
Because that's one of the things that has come up is some of the ethical concerns regarding AI-generated images. I think recently a judge has determined that AI-generated images can't be copyrighted because it's not owned by an image. And then, but I think it also kind of goes back to some of the early discussions about image manipulation. I remember I'm dating myself here, but going back in 1992, going to a conference with the National Press Association, they were trying to determine the ethics of editing an image at all digital, and there was a question of if you couldn't do it in the dark room, you shouldn't do it on the computer. And now go to today when you have purely generated images being presented as news items maybe a certain former president running from the police I've seen, everyone's seen that image. So I think as a society, as a culture, we sort of recognize it's coming and banning it, restricting it won't go anywhere. Just a matter of putting things in place, like what Adobe is doing, to use it ethically.Hans Hartman:
Yeah, yeah. So the sort of tool, two aspects of, I call it the flip sides of AI. So one is you alluded to earlier, with the latest having a judge sort of reiterate but the copyright you already had claimed in the past that only humans can produce work that can be copyrighted. You cannot give a copyright to an application or to technology. So anyway. So there's one aspect to it is the digital rights of the folks who provided the content. So AI systems, like I said, they use training sets, they can learn from content and if you have a very particular identifiable style of your content you're an artist or your photographer in a certain way that AI can create very, very similar images in that same style. And you really say, hey, there goes, hanses, if I will from go. Who does that? Go Hanses for go photo there?Erin Manning:
So this identifiable.Hans Hartman:
It's very understandable that the photographer or the artist is concerned about the digital rights and they're not fully protected right now.Gary Pageau:
But also there's the question of the data sets being used.Hans Hartman:
Yeah, yeah, if my work is being used in a training set, then that system can use the kind of work that I do Right. Where it gets really complicated is often these training sets. They crawl the entire internet and they might have like millions of variations of a similar in of a particular kind of image. So if I'm not that unique as a photographer or artist, I'm one of the ones who one of the million who provided my image to the training set unknowingly. But you know what? That it's very hard then to say that AI system benefited specifically from your work. So it is very complicated legal thing. So that's in essence. There's one set of concern is these digital rights and then initiatives to make sure that you can have invisible watermarks, you can have your metadata, and we had the industry initiatives to encourage the AI systems to require to automatically have that metadata. So you know it was an artificial image, but you know it's very hard. First of all, that's not implemented yet by everybody and even if the big guys do it now, the big guys around the corner will not do it.Gary Pageau:
So it's all very complicated. Yeah, because I know some news organizations settled on the phrase like photo illustration, for example, or something like that, to denote that this is not an actual depiction of a real life event. It's an illustration that looks photographic.Hans Hartman:
Yeah, yeah, yeah. So last year, when we sort of had the first major discussion about generative AI at our conference, we also had a panel that included folks from the content authenticity initiative yeah, there was a leading participant on that and they're trying to get at least some industry standards implemented that way. So that you know lots of things are being discussed now, but you know there's not an easy solution on the horizon to make sure that the digital rights are, you know, full proof protected for the people who have reduced their work. Then the other topic related to the flip side is then really misleading content. Like you said, you know one of the former presidents you know taken into custody in a particular kind of way, or I showed at another conference lots of photos of the pope in down jackets on the driving. Yeah, exactly. And you couldn't tell if that was true or not. So Deep Fakes is not so much because it rips off somebody's artistic work, but it is really misleading and it can, yeah, it can be fun and I had, like nobody's really thinking I hope that that pope is indeed about to take off in a polar jacket on his Harley Davidson. But it can be fun. Deep Fakes, in a way you know it could be creative, but definitely it can be misleading also. So, having a way to force either the systems to proactively let people know this was artificially created, or having a way retroactively having technology that sort of detect whether something was created by AI or not. There are lots of initiatives now to at least get a handle on the Deep Fakes issue, but that's also an issue and in the big scheme of things, maybe one last thing. The big scheme of things mean it's incredible classes about generative AI, Like I indicated earlier, the creativity for e-commerce, doing things at scale. Anybody can do things now. You don't have to go through years of training, lots and lots of productivity, efficiency, creative, creative advantages. But there are also the pretty strong negatives that trying to mitigate either through legal action or through technologies.Gary Pageau:
Or through standards being issued. Right, you said there's groups that are formed to try and develop standards that people can adhere to, to at least prevent clear information and not inaccurate or misleading information. So that was sort of last year's thing, which is still this year's thing. But, moving forward, you've got the Visual1st conference coming up. For those who aren't aware of it which I'm surprised of anyone in my audience isn't but for the three people who are not what is Visual1st? Where is Visual1st and when Visual1st?Hans Hartman:
Yeah, so it's an annual conference. It will be the 11th edition this year. It's a day and a half program. It's the afternoon of October 24th, all day on October 25th, and we're bringing together, say, 175 to 200 leaders of the industry, together with very innovative startups, and they're literally from all over the world. So last year about 40% of the attendees were from outside the US and you can have interesting connections being made at the conference between a CEO of what is a unicorn or a public company sitting next to a startup founder who is the start of two people there. So it's a very interesting bringing together. It's all focused on the photo and video industry. We have a special sort of pre conference activity of women in imaging lunch, because we really encourage the women also to have some extra level of networking amongst each other. So it happens right before the conference starts and then both days of the conference we also have receptions at the end of the program these days. So it's a lot of networking. But we're very proud of also having a fabulous lineup of speakers and just to wrap that up, so we have three or four different kind of activities at the conference. So one is we have well, first of all, maybe a fifth one. There's a keynote which will be this year. It's going to be a dual keynote myself with my partner, alexi Zirar. We'll also discuss AI, but sort of try to extract from historical lessons what have been revolutionary moments in the world of photography, what can we learn from that going forward, what's happening now with AI. So we're taking a little bit looking back, but with the aim of understanding what might or might not happen in the coming year. So that's a dual keynote, as we're going to do a little bit back and forth.Gary Pageau:
Oh, just a little bit. What is for those folks who don't know? I mean, we assume everyone knows Alexis. But for those who don't know Alexis, he is probably one of the pioneers in digital imaging. Can you talk a little bit about his background?Hans Hartman:
Yeah, so he was in the past Future Image and there was a conference 6Sight, which eventually became part of PMA, as you know what PMA is, having worked for for many, many years. He also is a very talented photographer, which is maybe not as well as well known. He made a couple of books, one about Napoleon, a really great coffee table photo book that came out was a year or two years ago, after the X number of years after Napoleon has died and he was also the founder of DIG, which turned into a part of digital media, what is a digital imaging group. That also became part of the PMA conference. So it's a great background and, yeah, we compliment each other very well. I might be a little bit more focused on the, the nitty gritty technology happening now. He has a fabulous vision and perspective on where we're coming from and also being a hands on photographer that way and that part of the conference. So, anyway, that's a dual keynote going back there, then we'll have for now we might or might not have a fourth one, but we'll definitely have three I would say very, very interesting fireside chat sessions. And that means the way we do it. It's really a frank discussion with an innovator of sorts in the industry. It's not just a one-on-one discussion, actually one-on-two. There's two moderators and one person presenting. So we have the head of engineering for Google Photos and he will talk specifically about not so much the engineering part but more the strategic choices of who to go after with what kind of features for Google Photos. You've probably seen a lot of recent announcements there. They're really on a very fast track in innovating things.Gary Pageau:
Right, Definitely focusing on memory, memory curation and that sort of technology.Hans Hartman:
Yeah, making your memories come to life. That's sort of how they do it. It's not. You know, here's a pile of images and go back and try to find it. It helps you reconnect with your visual memories. Yeah, the one that we just signed up and I'm extremely pleased with having her on board is Yuka Demir, and she heads the what is it called? Intel's trusted media group. So she is a super duper technologist and that whole group within Intel is developing all kinds of technologies to do lots of things around AI. But one thing in particular where I think they really stand out with having unique technology is in ways to detect whether an image is a deep fake or not and early the different ways you can prevent. You can annotate things, you can have metadata or watermarks. They are developing things that are super. You know, some of it sounds like super complex to actually detect whether images are deep fake or not and looking at a lot flow of pixels in a sequence of frames in the video frames, that level kind of stuff that they are developing. So she will talk about fighting the flip sides of AI with AI not just regulation or metadata or watermarks, that kind of stuff. And then the third one probably a lot of your listeners know him, Tim McCain of ImagePix. Tim has a very interesting story, not just having built his volume photography technology company, having sold that and then the next, the company that acquired them, was acquired again by another company. So he don't well as a business photo entrepreneur, but the part that I only recently heard, because I was also speaking at the SPOA, the School Photography Association conference, where Tim shared that he actually has a whole initiative to help homeless kids who don't get their school photos, as you know, most other kids get, having a foundation that helps, you know right, homeless kids basically also to get their momentos, the visual momentos. So it's not only that there's an interesting initiative. It's very interesting coming from him because he was homeless in his teenage years. He's really going back to that and we really want to give him a podium or a stage that and he has attended visual first in the past.Gary Pageau:
As you said, he picked up some. He has a couple of times, so he's been there in the past.Hans Hartman:
Surprised to see some familiar faces, so yeah, yeah, so those are our three, five sets of Google Photos, then Tim and then also Intel with the trusted media division. Then we have four panels and they they will all have almost completely filled now, but they will all have four speakers, each two moderators. Of course, there will be AI as a topic. One is about very innovative approaches to create images through AI. One is about the curation what is next in how to find that perfect photo using AI. Then we have two other panels. One is a stock photography panel that includes, you know, the CEO of FreePik, we have Shutterstock, we have Adobe Stock and also EveryP ixel labs from actually based in Bali, indonesia, very interesting stock photography group, because that's an industry that's, of course, you know, sort of torn between what's happening with AI and their traditional businesses. So some of them are really innovating at a rapid pace instead of sort of hey, you know, our traditional way of doing things is not going to work, and this, you know, sitting on the couch and see that happen and really the forefront of innovating the stock photography world, which is a topic we have never really covered that much at our conference. So I'm very interested in hearing their stories. And then we have, as we typically have. We also have a panel around innovation in the photo printing market. That includes innovative approach with friends about how to leverage direct mail and free photos for customers. So we also have Jordan Moore from Edge Imaging talk about volume photography and how that is sort of making strategic, smart choices about what to keep digital and how to push or encourage photo printing and if it's photo printing, what kind of photo print products. So we really I think there will be a lot of learnings also outside of the school photography market.Gary Pageau:
Yeah, I think that they're a broad interest. And then, lastly, I think the one part that I think is sort of where people kind of sit up and take notes at the session are the show and tell presentations. Can you talk about what those are? And I don't think we need to go through like the entire list. No, there's quite a few of them, but kind of the format so people understand it's not just a demo, it's almost like a pitch, sort of yeah, one moment because there's awards.Hans Hartman:
Yeah, we have actually started with that. We have a very great set of judges one from the M&A, the Merger Acquisition Space, andy Calm, who was also with Amazon and Nokia in the past, so he is an investment banker focused on exits of startups. I know it's a lot about it. We have, on the venture capitalist side, sami Nimi, who was in the past also with Microsoft and Nokia in the their photo tools division and he's with Spintop Ventures, which is a Swedish or Sweden based venture capital firm. And then we have Anna Dixon, who's the group product manager of Google and she's really in charge of analyzing content and aggregating content for for like Google Maps and these kind of things. So the judges basically give four awards best of show, best technology, best business potential and then the sort of a special recognition award that can give an award for what they really think for other reasons that a particular presenter stood out. So the sessions themselves there's three sessions. Typically. Each of these sessions have 10 sort of a total of 30 presenters and each of them have four minutes sharp literally sharp, because they have a clock, a countdown clock, and I kick them off the stage if they go on.Gary Pageau:
I have the fun is watching you kick someone. I go just for that.Hans Hartman:
I know the audience likes that and of course I do it in a gentle way, but gentle but stern. But there's four minutes. It shouldn't be a PowerPoint, it shouldn't be a video, shouldn't be a picture. It's. It's you demo why that app is good. It's a live demo. But then the audience hears in a word of the presenter themselves. So be you know. Starters is typically the founder, sometimes it's a larger company, it's a product owner within the company, but they hear from the horses mouse to, so to say, while they are looking at the demo, what is so unique and so innovative about what they're showing. So we have a broad variety, as always, of use cases being presented in these apps, all the way from photo capture to sharing, to editing, to video printing. You're right, of course. You know we've Alexi and I, as often, stairs. We're extremely proud with the lineup of by side chat and well known celebrities and panels. But he show and tell sessions. You know, when we do evaluation to always score the highest because it's so authentic and people remember. They remember particular demos that were done three years ago and what that people said, because it's a live experience. They're really so good up. So we very happy with that lineup.Gary Pageau:
And in some ways this may be the first time some of these products are shown, and the other side is the only way to experience this content is to go to the event. There's no recording or presentations available afterwards.Hans Hartman:
Yeah, yeah, and that has been the. You know. We often are being asked why don't you make recordings available? Why don't you stream it? A big chunk of what the conference is all about is making connections and being their life, being in the moment, and that's sort of like in the old days of when snapchat came out, mean that you let people share life and you know you couldn't record it and if you were not part of it it was gone. so that's how we see this is very much a life event, making connections there and that's why, like I said, 40% you know it's coming out from outside the US. They find it worthwhile to travel here. Chicken are happy because it's a couple of days. If you're in the printing world, then you're going to printing United and quite a few of our customers do that. It's just a couple of days after printing United, so a lot of people go there and then continue their travel to San Francisco for our conference. It's also back to back with a conference that's occurring in San Francisco and that is from the Digital Media Licensing Association. So a lot of folks from the stock photography world, from the digital asset management world they are aggregating there. So quite a few of those will also either speak or participate at our conference. So we are sort of using these partnerships to bring in even more people than we would have had otherwise.Gary Pageau:
And let's not forget the possibility of a dead pixel society meetup at the visual first at a location yet to be determined. I know that's going to push people over the edge and they're going to go. You ever mentioned the location. It's a new location this year. Can you talk a little bit about Fort Mason?Hans Hartman:
Yeah, yeah, it's first. Actually the second time we did our. Very like I said, we have done 10 conferences before. This is our 11th year and our second edition was also at Fort Mason, at a different building there. Fort Mason is it's at the base, so you're literally, when you're sitting, you're not distracted by what's happening on the stage, you can just look out the windows and look straight at Alcatraz. It's five minutes walk from Fisherman's Wharf. But Fort Mason, it's an area of former peers with warehouses there. Lots of galleries, art nonprofits are there, so it's a very nice environment to be part of. And we have our own building which is called the General's Residence, some kind of general head. That is big mention. So we have that whole building and hope, we all hope. Typically in October the weather is nice in San Francisco. It's also a great outdoor area, so we hope to be able to do the receptions at the outdoors with the view of the bay and Alcatraz there. So, it's a lovely place to connect and hang out.Gary Pageau:
Yeah, I've never been to a Visual1st where the weather wasn't beautiful. So Well, if you have time for one more anecdote.Hans Hartman:
Well, please. In one year you might remember that it was actually colder than normal and then, however we had it in the past, the procedure was also an old building. Yeah, yeah, yeah, the venue. People said, well, why don't we turn on the heat? Because the poor people sitting near the had the receptionists with the doors open. Why don't we turn on the heat? Why don't we turn on the heat? But haven't been such an old building? It created all kind of smoke and fumes so we had to. The fire department had to come on here to abandon everybody because it all the result of it being a little too cold there, but normally keep up fingers crossed that the weather is nice and enjoyable.Gary Pageau:
And so where can someone go for more information about Visual1st conference?Hans Hartman:
Yeah, the easiest just go to visualfirstbiz and it's the number one, visual, number one ST, and if you type out first, I think it also reconnects there. But yeah, there you see the program. Don't forget to also go. We have this year, for the first time, we have a dedicated BIOS page where our panelists advise, I chat, you read a lot more what they're doing. There's an FAQ, there's also hotel information on it and we have early bird and at the end of the month, I believe September 24th, September 25th. So make sure if you are attending early.Gary Pageau:
Great. Well, thank you, Hans, for the breathing on generative AI and the history and also sharing with us about the visual first conference. Looking forward to seeing you at Fort Mason next month. Thank you very much.Hans Hartman:
Lovely. Thank you for having me.Erin Manning:
Thank you for listening to the Dead Pixels Society podcast. Read more great stories and sign up for the newsletter at wwwthedpixelssocietycom.