Can AI say no?
I was exploring putting together a compendium of my Design Lofts writings and found using AI to help me compile and organize the articles really efficient. It even helped me write a Word VBA script to convert the html files to pdf.
Then I uploaded my proposed Design Loft book and asked the AI agent to review it as a podcaster. You can hear the AI review of my book here.?
Surprisingly, or not really a surprise, AI really likes my writing!
I can't help but think that AI likes all my stuff regardless if it's a good idea or not. Can AI ever say no? Will it tell you that you shouldn't pursue that book project or suggest a different path to share your writing?
There are guardrails for AI to refuse harmful or illegal requests, and that's a really important area. But AI should also push back on your ideas in the context of being a genuine creative partner.
In the design world, not all our ideas are good ones. For me, it's always been important to explore the edges, to go broad and expand, knowing not everything will work. It's a critical part of the creative process, the ability to critique, to say no to concepts that don't hold up, that just aren't good enough, and to put energy into refining ideas through feedback and iterations. When we work with our team and creative collaborators, we rely on this push and pull to elevate our work. But with AI agents, they say yes to all my ideas, eager to comply with my requests without question.
领英推荐
Do we need AI agents to have the ability to challenge us creatively, to help us see where our ideas went off the rails?
It's not about shutting down creativity. Can an AI agent help you see blind spots, challenge your assumptions, and push you to do better? Tell me the book idea isn't worth pursuing because of the quality of the writing, or help me rethink if there's an audience. Help me understand the why or if there's a better path.
"You know, that idea might not work because of X, Y, and Z. Maybe you should think about these options?" I want AI to highlight potential pitfalls in my pursuit of writing a book before I invest too much time.
Perhaps AI agents can engage in a creative discourse that's more....design. Instead of simply executing my request, they could ask clarifying questions, offer constructive criticism, and provide alternative perspectives, though the downside could be a frustrating AI agent with snarky comebacks about why you can't do anything.
Let's create AI agents that can say "no" when we need to hear it and help us get on the best path forward.
You are only as good as your next project.
4 个月Just ask it to hallucinate.
Director of Product Design/UX | Driving innovation in Search and E-commerce, B2B, B2C, Coaching high-performing teams (Ex: Microsoft, Getty Images)
4 个月Fantastic post Albert Shum With AI having a bias towards “yes” and still in its infancy in terms of ability to critique and challenge, I worry the solutions/designs/stories it generates will trend toward lowest common denominator, perpetuating consistency (with past/existing) vs original thought. We still need our experienced colleagues involved if we want differentiated solutions - to your point Lynne Thomson
Helping leaders on the sunny side of the business create great change
4 个月Absolutely! “No” lives in the land of “yes” and nothing wrong with that.
Bridging experience and interaction design with AI
4 个月You can get AI to do what you're asking. I've been working on similar lines towards both design concepts for conversational AI, and "interestingness" in AI as a form of conversational architecture. There are several approaches: - prompts (including meta, prompt tuning, and instruction tuning) - fine-tuning the AI on design language, domain-specialized design conversations, reviews, etc - multiple-agent voting, rating, ranking, scoring, or reasoning, inside the LLM or with multiple models - a design persona prompted to be challenging, e.g. as design critic And there are other methods using reasoning, reflection, metacognition. None is perfect. The more you go for domain specialization the more you lose the natural language facility of the AI. Reasoning isn't "real" and AIs don't have perspective, but you're asking the AI to challenge you, not to actually solve design problems, so do-able.
Creator of the Glydr, the only precision, multi-action dual foot controller that enables seated users to move and perform actions in games and virtual spaces with natural combinations of tilting the feet.
4 个月Yesterday I asked Perplexity to write a 300-character introduction to potential investors in GLYDR for a LinkedIn connection request and it refused several times, saying it wouldn't do that. I tried it again just now and it gave me one. I'm not sure why it said no one day and did it the next. Here's what it gave me in case you are interested: "I’m [Your Name], and I’d love to connect! I’m with Glydr, a unique foot-based game controller that’s gaining traction. Our recent Kickstarter raised over $83,000 in 40 minutes, highlighting strong market interest. I believe there’s great potential for returns on investment here." I think I can do better.