Fireside Chat, Roundtable, Interview
Roundtable #7: Spotify, Adobe and Linkedin on How AI Changes The Future of Product & Design | E1097
- The market will likely transition from a few mega-cloud models to a diverse ecosystem of long-tail models, including local and open-source variants, with routing logic directing queries to the most specialized and cost-efficient option.
- Model size will continue to scale significantly, though parallel trends show smaller models improving their performance to exceed previous larger versions, with efficiency gains driven by neural hardware and verticalized software potentially offsetting the end of Moore's Law.
- Companies are expected to reduce costs and improve margins over time as efficiency increases, potentially leading to the development of proprietary models for specific use cases while partnering for general large language models.
- Future architecture may involve embedding entire user histories into single models for prediction while retaining specialized models for generic tasks like rendering and voice, though this approach may coexist with separate optimized systems.
- Long-term success will depend on high-fidelity user data, as the volume and quality of user history become critical for powering proprietary models and enabling powerful queries.
- Enterprise AI adoption is predicted to follow a sharper S-curve than cloud computing, driven by step-function breakthroughs rather than linear progress, with a transition to business models based on task completion rather than seat-based licensing.
- By 2030, job tasks are expected to change by at least 65%, marking a significant acceleration compared to the 25% shift observed in the previous five to six years.
- Product development will face a critical shift where AI becomes the product rather than a feature, requiring organizations to unlearn deterministic design principles and embrace probabilistic, persona-focused interfaces.
- New roles for product leaders will prioritize data quality as the "oxygen" for AI, necessitating a holistic approach to experience design that treats tone and interaction as brand elements.
- Innovation in the application and "middle tier" layers is anticipated to surge, driven by startups building routers that manage multiple models and cost centers to mask complexity and improve efficiency.
- Business models may face disruption from the potential obsolescence of current advertising paradigms in a conversational world, with new frameworks emerging that could challenge incumbents.
- Implementation of AI-first strategies must originate from the CEO level to effectively retool companies and foster the necessary cultural shift away from traditional control mechanisms.
- Risks include the failure of incumbents to move quickly enough compared to startups, the inability to develop fast enough in the new landscape, and the challenges associated with the conservation of complexity in simplifying user experiences.
- Designers and product teams must rapidly acquire expertise in model capabilities, such as GPT-4, to build fault-tolerant experiences and redefine UIs that capture better signals through tone and interaction.