From: redpointai

Artificial intelligence (AI) is fundamentally transforming creative tooling across various domains, including images, video, and music [00:00:00]. Scott Belsky, founder of Behance and Chief Product Officer and Chief Strategy Officer at Adobe, discussed the impact of AI on creative workflows, the opportunities for innovation, and the challenges faced in this evolving landscape [00:00:05].

Current State of AI in Creative Tools at Adobe

Adobe’s strategy for AI in creative tools focuses on three core pillars: interfaces, models, and underlying data [00:03:00].

  • Models: Adobe developed its own family of generative models called Firefly, which are trained exclusively on licensed content [00:03:10]. A compensation program exists for content contributors, and Adobe aims to provide indemnification to all customers, emphasizing the ethical training of these models [00:03:20].
  • Interfaces: These models are integrated into products like Photoshop [00:03:34]. Adobe also makes Firefly models available to third parties and large enterprises for complex, at-scale workflows [00:03:42].
  • Control Capabilities: Key features like “structure reference” and “Style Match” are being built to give users more precise control over AI-generated content [00:03:59].
  • Custom Models: A significant focus is on enabling customers to train their own custom models, such as Nickelodeon training a Firefly version on SpongeBob SquarePants to generate new content without concern [00:04:06].
  • LLM Partnerships: For large language models (LLMs), Adobe partners with external providers, surfacing these capabilities through products like Acrobat and AI assistants for marketing analytics [00:04:26].

The focus on commercial safety and data compliance is a key differentiator for Firefly, ensuring users don’t inadvertently generate copyrighted content (e.g., Spider-Man), which is a top concern for enterprise customers [00:04:41].

User Adoption and Workflows

One surprising observation is how users combine different AI features in unexpected ways [00:01:10]. For example, users discovered they could use Project Neo (a 3D illustration program) to create 3D structures and then use them as “structure references” for Adobe Firefly, allowing for extreme precision in image generation [00:01:15]. This highlights a shift from focusing on the “next best model” to emphasizing the controls applied on top of the model [00:02:09].

AI tools are empowering creative professionals to explore a far wider “surface area of possibility” much more quickly [00:06:31]. An example is the “Generative Recolor” feature in Illustrator, which allows users to instantly apply hundreds or thousands of color palettes to vector creations, a process that previously took days [00:06:42]. This frees up creative professionals for higher-order exploration of concepts rather than rudimentary tasks [00:07:04].

The Role of “Taste” and Future UI

In an era where rudimentary skills are increasingly offloaded to AI, “taste” becomes more important than ever [00:08:50]. The goal of creative tools is to enable users to flex their taste through features like prompt augmentation and onboarding experiences, translating their mind’s eye into better outcomes [00:09:12].

The future of AI-driven interfaces involves “UI on demand” [00:13:36]. Instead of a standard UI, the AI will generate custom interfaces to fine-tune outputs based on specific user requests, which then disappear when the task is complete [00:13:27].

Challenges and Opportunities in Specific AI Areas

Video Models

While many startups are touting video models, achieving “professional grade” quality for commercial use cases remains a significant challenge [00:14:17]. There’s a notable gap between “cool happy path demos” and pragmatic, commercially viable quality [00:15:02].

Instead of full video generation, immediate value lies in smaller, pragmatic improvements, such as extending a video scene by a few seconds in a tool like Premier Pro [00:15:31].

Music Models

Many music AI startups focus heavily on the AI aspect but often miss the human element [00:22:31]. The reason people repeatedly listen to a song is often due to the human story or folklore behind it [00:22:38]. AI should lower the barrier for music creation but also allow for human storytelling to foster resonance between artists and consumers [00:22:56].

Broader Market Opportunities and Challenges

The creative and marketing space is moving towards hyper-personalized digital experiences at scale, where websites and communications are tailored to individual users [00:17:22]. This requires connected customer data, marketing workflows for deployment, measurement, and optimization, and creative stacks with “guard rails” like “brand check AI” to prevent inappropriate content [00:17:43].

  • Startup Opportunities: Startups have a significant opportunity in helping small businesses operate like much larger enterprises, leveraging AI tools to automate marketing and creative tasks [00:19:04]. There are also opportunities in tackling “antiquated processes” in sectors like law or government [00:19:29].
  • The Model Landscape: While initial thoughts suggested a few dominant AI models, the reality is likely thousands of specialized models, many becoming commoditized or focusing on niche use cases [00:10:00] and operating “on the edge” [00:19:19]. Creative tools should allow users to choose different models for different needs, recognizing that not all use cases require commercially safe, licensed data (e.g., ideation or mood boarding) [00:10:17].
  • Incumbents vs. Startups: Companies with existing workflow products and proprietary data sets (like Adobe) have an advantage in integrating and fine-tuning AI models [00:12:51]. However, startups can innovate by focusing on interfaces and niche data sets, especially where established players don’t have the same advantages [00:16:51].

Evolution of Content Consumption and Human Creativity

The coming “flood” of hyper-personalized content will lead to an “inundated” audience [00:20:18]. This commoditization of content will likely drive a craving for “scarcity, meaning, and craft” in digital experiences [00:20:57]. This could “reboot the role of humans in content creation,” emphasizing handcrafted and authentic experiences [00:21:19]. Shared social experiences, like discussing a common piece of entertainment, are also unlikely to disappear [00:22:04].

Future Milestones and Considerations

Customers are moving beyond simply wanting AI to “do things for them” (like automating Photoshop tasks) [00:24:10]. The next tier involves AI offering “proactive suggestions” for creative directions or potential issues, acting as a “superpower” for discovery [00:24:30]. Universal desires include increased speed and quality [00:24:56].

Cost and Pricing

Inference costs are a consideration for AI features, driving efforts to reduce costs and increase efficiencies [00:25:23]. However, product development is not constrained by cost if the value to the customer is high enough [00:25:47]. Adobe uses a “generative credits” model, allowing users an initial amount with existing plans and offering upgrades for more intensive use [00:26:00]. The goal is to keep these credits accessible to make AI a ubiquitous part of everyone’s workflow [00:27:25].

Biggest Surprises

The biggest surprise in building AI features has been the impact of “the devil’s in the defaults” [00:28:13]. Making AI features like “Generative Fill” a prominent default in Photoshop significantly boosted utilization beyond forecasts [00:28:22].

Changed Minds

Initial assumptions about a few dominant AI models have changed; it’s now expected that there will be thousands of models, many of which will be specialized or open source [00:29:07]. As models rapidly increase in capability, most common use cases fall below the frontier of what these models can do, shifting focus towards cost and accessibility [00:29:31].

Adobe views frontier models as platforms, where advancements in their capabilities directly improve Adobe’s products [00:30:33]. By also building proprietary models where Adobe has world expertise, they can offer a full-stack advantage [00:30:43].

Broader AI Applications and Personalization

Beyond creative tools, AI offers significant promise in fields like identifying mineral deposits, where antiquated processes can be dramatically improved in speed and reliability [00:31:08].

Another exciting opportunity for AI lies in scaling personalization in physical spaces [00:32:36]. Just as Uber made “everyone’s personal driver” accessible, AI could enable the magical experience of feeling “known” at a restaurant or store, remembering long-tail preferences, making previously exclusive services available to all [00:33:09]. This involves scaling experiences that were once only available to big companies or through high-end services [00:33:42].

Scott Belsky shares his insights monthly through his newsletter, “Implications” [00:34:28].