In late February 2024, Google paused Gemini’s image generation of people. Prabhakar Raghavan, Google’s Senior Vice President, Knowledge & Information, shared why.
First, let me stress that this post is not about whether generative AI should exclude some people groups based on race, age, or other characteristics. That’s a whole other (thousand) posts. I am annoyed that when creating with generative AI, white young people are the default. Even when I prompt for “diverse” people, tools still skew young and Eurocentric. Users have to go out of their way to include images of people from Pakistan or Belize and show folx who are older than 23 years.
Creating Conditions for Reason
This post is about how crucial logic is for creating generative AI tools. It’s right there in Prabhakar’s blog post.
So what went wrong? In short, two things. First, our tuning to ensure that Gemini showed a range of people failed to account for cases that should clearly not show a range. And second, over time, the model became way more cautious than we intended and refused to answer certain prompts entirely — wrongly interpreting some very anodyne prompts as sensitive.
Prabhakar Raghavan
As I followed the developing story, the importance of logic loomed in the back of my head. Yeah, like, syllogisms and all that. Classical rhetoric and the like. Reasoning and critical thinking.
Designers must establish conditions when developing generative AI tools—just like when they select stock photos or rules for a board game. What content is “in” and what is not “in” when generative AI creates an image? What people are “in” and which are out? What ideas should be enhanced, and which ones should be reduced? What is admissible, and what should be excluded?
It’s fascinating how creating these systems brings human reasoning to the fore. It’s shedding light on how designers set boundaries and constraints at a systems level.
Decisions have intended and unintended consequences. When those decisions guide how an agent (AI technologies) beaves, learns, and creates, designers’ decisions take on a lot more weight.
Be the first to comment.