Dec 6, 2023 · 28m · mad
Adobe’s AI Revolution: Scott Belsky on Firefly, Photoshop & the Future of Creativity
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the MAD Podcast, Adobe Chief Strategy Officer Scott Belsky joins host Matt Turck to discuss Adobe's ethical generative AI strategy across Firefly and Photoshop, the shift toward Content Credentials, and broader implications for creative professionals and venture-backed AI startups.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 13.7% of the talking time here. How this is scored →
Matt as informed peer 3.3 Guest teaching 2.5 Guest disagreement 0.4 Matt pushing back 0.6
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 8:05 Reframing database verification assumption
Scott explicitly rejects Matt's prompt asking if verification checks media against an existing database, explaining that catching bad actors via database detection is an unworkable cat-and-mouse game.
Hardest push from Matt ▶ 7:58 Double-clicking on Content Credentials mechanismMatt interrupts the broader narrative to explicitly press Scott on the exact technical mechanism behind Content Credentials.
Biggest teaching moment ▶ 8:05 Educating on asset provenance vs detectionScott educates Matt on why media trust must be built around cryptographically signed provenance metadata rather than centralized database checks or detection algorithms.
Matt holds his own ▶ 12:07 Demonstrating deep contextual background on AdobeMatt highlights his preparation by citing Adobe's 20-year history of AI development dating back to OCR alongside its $275B market cap to frame a question on organizational agility.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| AI's Role in Onboarding and Photoshop's Context Bar | 3 | 2 | 0 | 0 | Matt sets up the interview by contextualizing Adobe's place among tech giants making big bets on AI and asks Scott to differentiate products like Firefly and Sensei. Scott explains how AI solves user onboarding friction through Photoshop's new context bar. | |
| Adobe Firefly, Training Data Ethics, and Commercial Viability | 3 | 3 | 1 | 1 | Scott details Adobe's decision to train Firefly exclusively on licensed data for commercial safety, contrast with competitors, and custom tuning. Matt asks pertinent follow-up questions about expanding training sets and whether models are homegrown. | |
| Adobe Sensei, Fake Media, and Content Credentials | 3 | 4 | 1 | 2 | Matt double-clicks on Content Credentials by asking if the system checks media against a central database. Scott clarifies that bad actors cannot be caught via central databases, explaining that provenance metadata and verification are key. | |
| Firefly Model 2, Vector Generation, and Generative Match | 2 | 2 | 0 | 0 | Matt asks about recent announcements and roadmap items. Scott outlines Firefly Model 2, vector model generation, generative match features, and real-time object moving in experimental sneaks. | |
| Organizational Speed and Strategic Partnerships | 4 | 2 | 0 | 1 | Matt demonstrates host knowledge by citing Adobe's 20-year history in machine learning starting with OCR and its $275B market cap to ask how such a large enterprise moved fast. Scott explains their small cross-functional pod design strategy. | |
| The Impact of AI on Creativity, Quality, and Employment | 3 | 2 | 0 | 0 | Matt prompts Scott to share his broader vision on AI's impact on creative labor. Scott explains how AI reduces exploration time and raises quality bars, while Matt synthesizes the takeaway as 'personalization at scale.' | |
| Flow State in Creativity and Shifting Business Models | 5 | 2 | 1 | 1 | Matt cites specific concepts from Scott's writing, including removing work from workflow, shifts in business models, and caution around Cambrian explosions. Scott offers an insightful breakdown of moats, vertical AI, and local LLMs. | |
| Founder Psychology, Customer Empathy, and the Messy Middle | 3 | 3 | 0 | 0 | Matt asks about seed-stage founder evaluation and references Scott's book 'The Messy Middle' to address 2023 startup struggles. Scott highlights founder paranoid differentiation and customer empathy over solution passion. |
Statements from this episode (12)
Scott Belsky admits a significant portion of daily Photoshop downloaders get lost
“For all the people that download any of these desktop products like Photoshop every day, the number of people that get lost or can't feel successful in the product is really Significant.”
Scott Belsky: Contextual UI prompts matter as much as underlying AI models
“One of the most important insights was, and maybe as important as building a great family of models. And I can talk more about Firefly and the underlying models and how we trained it and how we didn't train it and everything else, but was simply the insight of…”
Adobe trained its Firefly AI models exclusively on licensed material
“We wanted to train off of licensed material that we had a license to train off of.”
Scott Belsky: All of Adobe's Firefly generative AI models are homegrown
“So all of our generative AI models for Firefly are homegrown.”
Belsky: Adobe automatically adds Content Credentials to AI-generated assets
“Whenever you use AI in any of our products, we automatically add credentials to the asset. That says what tool was used and what model you used.”
Belsky: AI detection algorithms will never catch media manipulators
“We're never going to catch the bad actors because anyone can, you know, make anything and it's a cat and mouse game to have algorithms that tell whether something was edited or not.”
Belsky: CAI has 2,000 partners, including Nikon, Canon, Sony, and Leica
“We now have 2000 partners, Nikon, Canon, Sony, and Leica have all come out with cameras. They now add counter-credentials to assets by default taking in their camera so that when they come into Photoshop or Lightroom or third-party products of other companies,…”
Belsky: Brands will flood the market with content, raising the engagement bar
“We're entering a world where, first of all, every brand's gonna absolutely flood the zone with content. That's gonna make the bar go up for the digital experiences that really engage us.”
Belsky: Adobe data shows AI is increasing creative hiring, not reducing it
“So in our data, we're still seeing this revolution drive hiring more as opposed to reducing the hiring.”
Belsky: Repetitive design work like banner ad formatting is already replaced by AI
“The job of someone turning squares into rectangles all day for banner ads, that job's been replaced by AI.”
Scott Belsky: Hourly business models are broken in the age of AI
“And so I'm starting to wonder if value-based pricing models are really going to replace time-based pricing models and a lot of the functions of like, very sophisticated work. And also, what are the tools that allow us to know how to price our work, right? But …”
Scott Belsky predicts local LLMs on personal devices will suffice for most
“I think that verticals will be important. I think the LMs will be abundant. And I actually think at some point, maybe relatively soon, we'll realize that some LMs, even locally running LMs on our machines are enough for a lot of the actions that we want to do …”