Adobe has officially launched a groundbreaking service called Adobe AI Foundry, aimed at revolutionizing how enterprises utilize generative AI through its flagship model, Firefly. This new initiative is designed to create bespoke versions of Firefly that are tailored specifically to the unique needs and branding of individual companies. Unlike traditional methods of model customization, which often involve simple fine-tuning, Adobe’s approach—termed “deep tuning”—involves a comprehensive rearchitecting and retraining process that integrates a company’s proprietary data, brand tone, visual style, and intellectual property (IP).
The launch of Adobe AI Foundry comes in response to increasing demand from enterprise clients for more sophisticated and nuanced AI solutions. As businesses seek to leverage AI technologies to enhance their creative workflows and customer engagement, they require models that not only understand their specific needs but also embody their brand identity. Adobe recognizes this need and has positioned AI Foundry as a solution that goes beyond mere adjustments to existing models.
At the core of Adobe AI Foundry is the Firefly model, which serves as the foundation for these custom solutions. However, what sets Foundry apart is its ability to understand and process multiple concepts simultaneously, making it a multimodal model capable of handling various types of content, including images and videos. This is a significant advancement over traditional custom Firefly models, which were limited to single-concept tasks and primarily focused on image generation.
Hannah Elsakr, Vice President of GenAI New Business Ventures at Adobe, emphasized the importance of this initiative during a recent interview. She explained that the idea for AI Foundry emerged from conversations with enterprise customers who expressed a desire for more advanced and tailored versions of Firefly. Recognizing the complexity of enterprise needs, Adobe decided to take on the responsibility of rearchitecting the models rather than leaving it to the clients. This hands-on approach ensures that the resulting models are not only effective but also aligned with the specific goals and requirements of each business.
One of the key features of Adobe AI Foundry is its commitment to data security and IP protection. According to Elsakr, Adobe will retrain its Firefly models using the enterprise’s proprietary data while keeping that information separate and secure. This means that the enterprise retains ownership of the output generated by the model, ensuring that sensitive data is not inadvertently integrated back into the base model. This level of security is crucial for enterprises that are cautious about sharing their proprietary information with external vendors.
The deployment of Foundry versions of Firefly will be facilitated through Adobe’s API solution, known as Firefly Services. This API provides a scalable way for enterprises to integrate their customized generative AI models into their operations, allowing for seamless interaction between the AI and existing workflows. By leveraging this API, companies can harness the power of AI to generate content that resonates with their audience while maintaining brand consistency.
Adobe’s deep tuning methodology is a significant departure from conventional fine-tuning practices. Elsakr describes it as a “surgical” process that involves reopening the Firefly-based models to incorporate the enterprise’s unique data and insights. This continuous pre-training approach allows Adobe to adjust the model’s parameters in a way that enhances its understanding of the brand’s voice and visual identity. The result is a model that is not only more capable but also deeply attuned to the nuances of the enterprise’s branding.
The training pipeline for AI Foundry involves collaboration between Adobe’s embedded teams and the enterprise clients. These teams work closely with clients to identify the specific data needed for retraining the model. Once the data is securely transferred and tagged, it is ingested into the base model, and Adobe initiates a pre-training run. This collaborative effort ensures that the final product is a true reflection of the enterprise’s brand and operational needs.
Two early adopters of Adobe AI Foundry are Home Depot and Walt Disney Imagineering, the research and development arm of Disney responsible for its theme parks. Both organizations are leveraging the capabilities of AI Foundry to enhance their creative processes and improve customer engagement. Molly Battin, Senior Vice President and Chief Marketing Officer at Home Depot, expressed enthusiasm for the potential of Adobe’s AI Foundry, stating that it represents an exciting step forward in embracing cutting-edge technologies to deepen customer engagement and deliver impactful content across digital channels.
As enterprises increasingly turn to AI for assistance in their creative endeavors, the demand for customization and specificity in AI models is growing. Fine-tuning and model customization have become essential tools for organizations looking to align large language models with their unique needs. Traditionally, enterprises have taken on the responsibility of fine-tuning themselves, connecting to the model’s API and retraining it based on their own data and preferences. However, Adobe’s AI Foundry shifts this paradigm by offering a more comprehensive and hands-on approach to model customization.
Elsakr anticipates that many companies will end up utilizing multiple versions of Firefly to meet their diverse needs. For instance, a company may use the Foundry version for broad, brand-specific projects, a custom Firefly for specific single-concept tasks, and the base Firefly for general creative needs. This flexibility allows enterprises to choose the right model for the right task, optimizing their use of generative AI across various applications.
The introduction of Adobe AI Foundry reflects a broader trend in the enterprise AI landscape, where organizations are moving away from generic models and towards deeply integrated, brand-aware systems. As businesses recognize the value of AI in enhancing their creative workflows and customer interactions, the demand for tailored solutions will only continue to grow. Adobe’s commitment to providing these solutions through AI Foundry positions the company as a leader in the evolving field of generative AI.
In conclusion, Adobe AI Foundry represents a significant advancement in the realm of enterprise AI customization. By offering a service that enables businesses to create deeply personalized versions of the Firefly model, Adobe is addressing the complex needs of modern enterprises. With its focus on data security, collaborative retraining, and multimodal capabilities, AI Foundry is poised to transform how organizations leverage generative AI to enhance their creative processes and engage with customers. As more enterprises adopt this innovative approach, the future of AI in business looks promising, with the potential for even greater integration and sophistication in the years to come.