Feature
TechRadar’s latest assistant feature is deliberately playful: ask ChatGPT, Claude, Gemini, and Grok which science-fiction AIs they most resemble, then compare the self-portraits. The piece is useful less as evidence of “what models think” and more as a window into how each product wants to be framed — helper, infrastructure, or irreverent sidekick.
The prompt, as TechRadar describes it, asked the models to ignore physical form and interface and focus on behavior, personality, empathy, values, goals, and relationship with humans — then rank a top three. That framing matters: it steers answers toward narrative archetypes rather than product specs.
“help, don’t dominate”
We’re not reprinting each model’s full ranked list here — those answers live in TechRadar’s original piece and can shift with model version or prompt wording. What the reporting does establish at a high level: ChatGPT and Claude converged on nearly the same top three in different order; Gemini leaned toward neutral, ego-free infrastructure; Grok leaned toward a witty sidekick archetype (including JARVIS).
TechRadar’s own caveat is the right one for tool buyers: treat the exercise as cultural feedback, not concrete evidence of inner life. Models trained on fiction will echo fiction — including flattering ones. An assistant that could enable dystopian outcomes could also claim to be the helpful robot from Moon.
Read the original
- TechRadar — sci-fi AI self-portraits — Full model answers and methodology