Surface-level improvements delivered with design tweaks, often described as trying to put 'lipsticks on pigs' is simply unacceptable in today’s realm of conversational interfaces. In this new era of generative AI, it's time to break free from our existing mental models of a chat window and challenge what we perceive as reality, and what’s considered “innovation”. It's a call to reimagine and redefine the very essence of conversational interfaces, creating experiences that surpass our wildest imaginations. Are you ready to step into this limitless realm of possibilities?
Design patterns empower us to challenge the norm and reshape our current reality, in a structured and innovative manner. By breaking free from traditional thinking, we explore new approaches in conversational interfaces that deliver exceptional user experiences. In this blog, we will delve into specific design patterns that we have observed and explore how they are being utilized in various contexts.
Today, even after the generative AI, conversations are frequently confined to the boundaries of a chatbot window, imposing limitations on design ideas. However, it is important to identify and overcome certain pre-concieved, leftover blockers that restrict our exploration of new possibilities:
Reimagining conversational interfaces goes beyond a standalone feature, integrating them seamlessly into the user experience for a holistic approach. It unlocks expanded possibilities, leveraging advanced capabilities such as contextual understanding and dynamic suggestions to deliver more fluid and integrated interactions. But in today’s Generative AI led world, organizations can gain a competitive advantage by offering innovative and seamless experiences that set them apart in the market.
Linus Lee's insightful talk at MLOps served as a catalyst for our own exploration and improvement in the field. Drawing inspiration from his years of experience in generative models and his current role at Notion, we have leveraged our own insights to delve deeper into the latent space of generative models and enhance our writing tools.
In the era of generative AI, bots have transcended the limitations of their training and can engage in free-form conversations with users. There is no longer a need to restrict them to a fixed set of three actions with carefully crafted verbs. However, it is crucial to be mindful of the discovery tradeoff that comes with this newfound flexibility.
In contrast to traditional chat interfaces (conversational 1.0), there is no longer a need to confine the bot to a specific corner of the interface. It has evolved to become a pervasive presence, offering contextual intelligence and support throughout the entire interface.
Contextual assistance in generative AI interfaces offer valuable benefits by providing relevant and timely actions based on the context that the user is operating in. By maintaining context across different interface elements, generative AI interfaces create seamless interactions and adapt the interface elements to optimize the user experience. Leveraging contextual actions enhances personalization, efficiency, and user satisfaction, resulting in more engaging and effective conversational experiences.
Co-creation mode in generative AI interfaces offers exciting possibilities by enabling collaborative creation between users and AI. It encourages innovation and exploration, as users and AI work together to push the boundaries of what can be created. This collaborative approach allows users to shape and refine the output, tailoring it to their specific needs and preferences. Ultimately, co-creation mode in generative AI interfaces empowers users to unleash their creativity while leveraging the capabilities of AI technology.
Literal mention, the act of explicitly referencing artifacts by name to provide context, can be harnessed to enhance generative AI interfaces. By incorporating literal mentions, users can specify particular items or entities within the conversation, allowing the AI to generate more accurate and contextually relevant responses. This level of specificity enables a deeper understanding of user intent and enables the AI to provide more personalized and tailored information. Whether it's mentioning specific products, locations, or individuals, literal mentions enable generative AI interfaces to deliver responses that align closely with user expectations, creating more meaningful and effective conversational experiences.
Feedback loop is a crucial design pattern in conversational interfaces, especially in light of concerns surrounding LLM hallucinations. By presenting users with the AI's interpretation alongside the generated results and actively seeking their feedback, we empower users to gain confidence and a deeper understanding of the AI's capabilities. This approach fosters trust, improves user experience, and enables a collaborative relationship between users and AI systems.
Despite utilizing conversational intelligence behind the scenes, Jasper.ai has a traditional point and click interface. We believed that applying these design patterns could greatly enhance the user experience. The following case study showcases our efforts in bringing this reimagination to life.
What if Jasper AI understood your brand's tone, preferred writing style, and your preferences based on the history of blogs you have created? What if it could suggest ideas for you based on expert or most popular blogs? What if it could help you create a successful blog by providing intelligence like typical reading time, structure, SEO score etc?
Imagine an assistant being available throughout that understands the context and is able to accelerate creativity by providing contextual support. In the context of the user’s focus, it is able to provide recommendations and suggestions. Users can add custom prompts to describe their ask.
Literal mention allows users to effortlessly tag expert names, blog links, documents, and other relevant references directly within the prompt itself. This feature enhances the overall clarity and accessibility of the conversation, allowing for a more dynamic and interactive exchange of information.
Feedback loop is a crucial design pattern in conversational interfaces, especially in light of concerns surrounding LLM hallucinations. By presenting users with the AI's interpretation alongside the generated results and actively seeking their feedback, we empower users to gain confidence and a deeper understanding of the AI's capabilities. This approach fosters trust, improves user experience, and enables a collaborative relationship between users and AI systems.
Together, we have an amazing opportunity to push our limits, challenge the norm, and create interfaces that shape the future of product design.
We'd love to hear your thoughts, ideas, and experiences on reimagining conversational interfaces. How do you envision the future of these interfaces? Are there any design patterns or concepts you believe hold great potential?
Generative Interfaces Beyond Chat // Linus Lee // LLMs in Production Conference
