10 Hypotheses for Future-Proof Gen AI Product Design

There’s a big risk of getting distracted or losing one’s way in the new generative AI space, especially during the curiosity phase we’re in now. You can over-solve for all the new tech or fly-by-night applications that are fun but don’t add sustained value. This space will have many first movers but far fewer last movers.

So, what does it take to have longevity in building generative AI products — ones that persist past the hype curve? Apple has rigorously applied the principle “form follows function” to make a dent in how hardware enables human creativity and interaction. The door is now open for generative AI companies to reinterpret that principle anew. This isn’t solely for the sake of profits or headlines: doing this right can positively impact how AI products are received and increase benefits for users.

‍What’s the Function of AI Products?

So what is the “function” and “form” of generative AI? Having read all the customer feedback I can at the generative AI company I work at, Jasper, I suggest the function of generative AI is to create an experience which includes: 

• Content that creators and those consuming it love most (the superlative is key because if it’s not the best content curated for readers, it might be spam).
• Easy, quick, and enjoyable experiences generating content.
• A feeling of connection to the work through a required level of “authorship,” depending on how intimate the work is.
• Confidence in the ethical soundness in both the process and output.

‍What’s the Form of AI Products?

It’s difficult to know what great design for generative AI products looks like given the diversity of its users — so we should experiment and learn through trial and error. In this spirit of experimentation and variety, I offer 10 hypotheses about enduring design of products in this space. These principles, inspired by Dieter Rams’ 10 principles of good design, are simply hypotheses. I invite others to experiment, comment, and improve on what’s presented below.

Hypothetical Principles for Generative AI Product Design:

Makes better content: It’s clear generative AI will produce a lot of content, but no one wants it to usher in a new age of spam. Content must be read by someone to be useful and that will always have somewhat of a ceiling since people can’t infinitely digest content. Therefore, generative AI must be about developing a content creation process that enhances a creator’s work and abilities, which results in better content for viewers.

Is “anti-fragile”: This space is already so innovative that generative AI product design should anticipate upcoming technological and market changes as well as shifting ethical considerations. This is not only future-proofing but deliberately planning to improve with each development instead of being disrupted by it. 

Gives voice to the colloquial: When radio and TV first emerged, well-traveled writers like Orwell and Steinbeck observed that listening to a few main channels destroyed local languages and dialects. Now, everyone can all generate content using products which might literally have the same voice. And we have the power to design these products around understanding and magnifying local voices, languages, and art styles, instead of blending them into monotony. Both paths are possible at this point and we must push for the former, even at the individual level.

Has integrity: An AI product should actively tell users when it’s being asked to generate things it should not be generating. This idea is especially important when considering work that could have real-world stakes, like content with legal or medical implications. AI should also advise users on proper attribution or royalties to the sources used in its generations. Falsehood or harm should be mitigated, which of course includes any toxic or violence-supporting outputs.

Is environmentally friendly: AI product design should conserve resources and minimize physical and visual pollution throughout the lifecycle of the product. Therefore, generative AI should strive to use the least energy possible to contribute to the preservation of the environment.

Inspires with interaction: People can interact with AI to create in the same varied ways a writer or graphic artist interacts with a pen or a trusted collaborator, fitting the user’s needs in a space between being a blank-slate tool and a full-fledged thought partner. These interactions are intended to inspire, enabling creators to create and learners to learn without the traditional “hooked” product side effects of addiction or misuse. Interaction is so important and rich in this space that the future of the UX field could be changed by it.

Is predictable: An issue with generative AI products today is they feel like they have a mind of their own instead of anticipating what the user wants. Features that allow users to have more intuitive, meaningful control (or not) over their outputs will become increasingly important. 

Has a memory: To endure in the future, products must be built on the past. While the products are “anti-fragile” and constantly updating, the way they use data can be long-lasting. Perhaps the best way to reach this goal is for these products is to learn from past user information and choices and continue to evolve. This would improve product stickiness and also allow each user to produce differentiated content based on their unique memories, even though the models behind the outputs might be shared by millions. 

Meets users where they are: Detail orientation for generative AI products probably focuses less on how and what users produce and more on where they are as the base models and systems grow increasingly versatile. We must think about the systems people already use — be they digital or physical — and how they inherently reduce friction. Think of a ubiquitous tool like headphones infused with technology that instantly translates any language spoken into them. And small details like voice preservation techniques could drastically change the value and experience of using those headphones.

Less is more: “Less, but better. Simple as possible but not simpler. Good design elevates the essential functions of a product.” No change from Rams' principles.

‍

Designing well-rounded AI will not only improve experiences and value for readers, observers, learners, and (hopefully) society but help build better products and companies. The above thoughts on design principles hopefully add to the conversation on building these tools to offer enduring value for the people who use them. 

Perhaps the best way is to focus on the purpose of creativity and content — helping people spread ideas, learn, interact, feel, and grow. Despite AI’s powers to scale infinitely, it’s about enriching humans and continuously bolstering their creativity, thus furthering the truth of Maya Angelou’s statement “You can't use up creativity. The more you use, the more you have.”