In the simplest terms, the Adobe Sensei AI we’ve been using for years focuses on detecting and deriving insights from patterns. It then uses the insights to automate optimal actions or help marketers make smart, data-driven decisions.

Generative AI, in comparison, learns from the data given to synthesize new content, such as new images and new copy, based on text-based instructions it’s given. It pulls in pieces of images or words from existing images and content to create something new. It might modify the colors or styles of images or text based on what you tell it to change or write copy in a form you describe — for example, writing this post in the style of Ernest Hemingway.

Generative AI can produce amazing content, but we’ve all heard news about it creating inappropriate, biased, and harmful content. We want to share how at Adobe we realize the benefits of the former, while putting guardrails in place to avoid the latter. For us, it’s important that you and our customers can confidently use content that our generative AI produces.

We’ll share more later about the guardrails for our generative AI offerings. But first, let’s look at some exciting new generative AI products that we announced at Summit and even in the last few weeks. 

Generative AI can enhance what people do — serving as a creative co-pilot, but not replacing human ingenuity. Our new products use generative AI in ways that will redefine creativity and the customer experience. 

Adobe Firefly is a family of Adobe-owned generative AI models that generate creative content. We’ve trained the first model on hundreds of millions of pieces of stock content from Adobe Stock and open-licensed content. We’ve integrated Firefly into our application and content workflows. Currently it can generate images and text effects and will soon be able to generate vectors, in-painting, out-painting, video, and 3D content.

At this time, Adobe Firefly is integrated with the following Adobe Creative Cloud solutions — Adobe Express, Adobe Photoshop, and Adobe Illustrator. With Firefly, you’ll be able to use AI to generate countless assets for your websites, ecommerce sites, marketing campaigns, and many other places. You’ll also be able to see which asset attributes and variations resonate most with your customers.

For marketers and customer experience practitioners, we announced our new Adobe Sensei GenAI Services. These services can be used across Adobe Experience Cloud solutions to increase productivity, content velocity, and personalization. With generative AI, teams can quickly produce marketing copy, chat experiences, captions, and other content as well as audiences and journeys to reach customers anywhere in any stage of their journeys.

The services use several large language learning models (LLMs) to generate text-based experiences. But as part of Adobe Experience Platform, Sensei GenAi Services also bring together a brand’s data and content under a common language model. This enables brands to train generative AI models on their brand insights to create content tailored to their use cases. 

Here’s an early look at some of the capabilities you’ll be seeing in Experience Cloud solutions:

• Experience Manager Assets. Use natural language text prompts to generate images and assets as well as stylize and create text effects. Produce content variations by updating colors, objects, and other imagery and adapt content for use across channels.
• Experience Manager Sites. Generate content for copywriting, re-writes, and summaries, and change content tone at a click. Scale creation of content variations while optimizing for personas across channels and create content that applies brand voice, the latest product information, and uses customer profiles and behaviors.
• Adobe Marketo Engage. Equip your chatbot with the ability to answer questions in a natural, context-based, and on-brand way. Package up information to prepare sales reps for customer meetings and live chats with summaries of customer intent, topics previously discussed, content and assets shared, and campaign goals. Provide conversational topics, questions, and responses for reps to use in live chat sessions, too.
• Adobe Real-Time CDP. Discover, suggest, and if wanted, automatically create valuable missed segments; simulate customer journeys based on the performance of past campaigns and customer profile preferences; and use conversational insights to continuously refine segments. 
• Adobe Journey Optimizer. Automatically generate customer journeys based on objective, industry, and persona with out-of-the-box use case templates. With Adobe Firefly, create commercially usable, secure, and ethically sound content for use in journeys. Automatically produce on-brand copy for use in messages created in the message designer, with recommendations about the copy that will perform best.
• Adobe Customer Journey Analytics. Benefit from easy-to-understand natural language insights, surface valuable trends with Line visualization, and gain text-based insights within context to make them more easily understood.

At Adobe, we take great care and thought to responsibly and ethically add generative AI capabilities to the solutions you and our customers rely on daily. We’ve always focused on developing any of our innovative AI offerings on a foundation of accountability, responsibility, and transparency. We apply these same principles to our generative AI offerings.

Here are some actions we’ve taken to add those guardrails around use of generative AI, along with critical related topics we’re addressing:

• AI ethics framework. As we develop our AI capabilities, including those that use generative AI, we adhere to a set of clearly defined AI ethics principles and have a rigorous review process.
• Safe and inclusive training data. We train our generative AI models on safe and inclusive data sets, and we test the output with the oversight of our AI Ethics Team.
• Transparency. It’s important that people know where content comes from to avoid the issue of misinformation that is today so prevalent. Content credentials let each creator add data such as their name, creation date, and tools used. Any content generated or modified by AI automatically shows as AI-generated in content credentials.
• Content Authenticity Initiative. To give creators credit for their work, several years ago we started the Content Authenticity Initiative, which now has 900 participating organizations. As a result of this initiative, creative professionals can choose if they want Adobe generative AI to train on their content. In addition, we’re examining the best way to compensate creators when Firefly uses their work in the content it generates.

Read this blog post for a more in-depth look at how Adobe is taking a responsible and ethical approach to creating capabilities using generative AI. 

To ensure that the content we deliver meets our customers’ high expectations, we’re working with several top industry partners — from our own Adobe Stock for Firefly images to the large language learning models from Microsoft Azure and OpenAI for text in conversational experiences, and NVIDIA for co-developing advanced generative AI models.

To learn more about all that we’re doing with Adobe generative AI, please check out the following articles and posts and sign up for the free Adobe Firefly beta in Adobe Express!

• Watch the Summit opening keynote
• Read about Generative AI on Adobe.com
• Test drive Adobe Firefly and provide feedback — join the free Firefly beta in Adobe Express