What GPT-4o means for designers

🌈 Abstract

The article discusses the new GPT-4o model, which is an upgrade to GPT-4 with enhanced multimodal capabilities. It can process text, audio, and visual inputs, and generate outputs in those formats as well. The article explores potential use cases for GPT-4o in design workflows, such as serving as a meeting assistant in user interviews and providing live design feedback during design reviews.

🙋 Q&A

[01] GPT-4o Capabilities

1. What are the key capabilities of GPT-4o compared to GPT-4?

• GPT-4o can process any combination of text, audio, and image inputs, and generate outputs in those formats
• It is a single model that connects text, vision, and audio, unlike GPT-4 which had separate models for audio-to-text, text-to-text, and text-to-audio
• GPT-4o can detect tones, multiple speakers, and background noises, which was not possible with GPT-4
• The processing speed and quality of GPT-4o is significantly faster and better than GPT-4

2. How did GPT-4o perform compared to GPT-4 in the author's tests?

• The author found that GPT-4o provided more extensive and relevant design suggestions compared to GPT-4 when asked to provide feedback on a UI design
• GPT-4o was also much faster, at least 2x faster, than GPT-4 in responding to the same prompts
• When asked to provide real-world examples of layouts similar to the UI provided, GPT-4o was able to accurately detect that the UI was for a course-related mobile app and provided relevant examples, while GPT-4 provided less relevant examples

[02] Potential Use Cases for GPT-4o

1. How could GPT-4o be used as a meeting assistant in user interviews?

• GPT-4o could not only take notes during user interviews, but also detect subtle non-verbal cues like facial expressions and tones
• This would eliminate the need to re-watch interview recordings multiple times to capture who said what and how they said it
• The detected emotions could also be included as part of the interview report

2. How could GPT-4o be used in design reviews?

• If GPT-4o can process visuals and voices, it could potentially serve as a partner in design reviews
• It could observe the designs and listen to the conversations between designers, and provide live design feedback and summarize the key points from the conversation