For the latest installment of our Responsible AI Programme we welcomed Thorsten Jonas, founder of the Sustainable UX network, for a presentation titled “From invisible to accountable: designing digital products in the age of AI”.
The talk reflects a question many designers, UX specialists and developers face today; “What can AI do for me and my work and is it possible to use it in a responsible way?”
To answer this question, Thorsten advocates curiosity, education and experimentation. He sees it as essential to look past the hype and take a balanced view of the potential benefits and costs for users as well as lesser seen externalities.
Thorsten highlighted a number of areas to consider.
- The limitations of AI. Incorrect outputs and unexpected behaviours are part and parcel of LLM models and need to be factored in
- Efficiency gains need to be carefully assessed. AI performing the groundwork required for a task might be quicker but verifying and tweaking outputs can be time consuming and sometimes offset that initial efficiency.
- Environmental and social costs. Individual users of AI add to the cumulative resource footprint of the industry as well as participating in a system that relies on exploitative labour practices in the global south.
- The flattening of outputs. Thorsten pointed to research that showed GenAI outputs tend to rely on the same patterns, over and over again. Creativity in writing and graphic design can be lost in the name of efficiency.
- The competency cost of using AI. If AI is doing the “thinking” for us, what does it cost us as design professionals?
So given these questions, what can design professionals control when it comes to AI?
- The point at which we actually use AI. Thorsten summed this up as “Brain first, AI second”. We need to think about the problem and explore potential solutions before turning to AI, rather than using it as a starting point and working backwards.
- Choosing assistance rather than automation. He advocates for using AI to help tweak and iterate work rather than blindly produce it.
- The choices we make. We get to choose the tools we use, how we use them and how we discuss them. Our choices as an industry helps influence the direction the industry will take.
- What model you use. Using frontier models for all tasks is akin to driving a huge SUV for a journey you could walk. Smaller, open source models offer legitimate alternatives.
A full recap of the presentation is available here.
Reflections on the topics raised in the presentation continued in this thread of the #greener-ai channel, our dedicated space to discuss “How can we build & use AI sustainably.”
Huge thanks to Thorsten for his time and insights. He will be in Turin on the 26th September delivering a talk titled “The true Price of AI”.
Thank you also to ustwo for their continuing support and partnership with CAT on the Responsible AI series. Keep an eye on our #announcements channel in CAT Slack and on our newsletter for further events.