On Thursday, OpenAI launched ChatGPT Pro, a premium $200-a-month subscription that stirred mixed reactions in the AI community. The plan includes access to the new "o1 pro mode," an enhanced version of OpenAI's o1 reasoning model, which claims to provide more reliable and comprehensive answers to complex questions in science, math, and coding.
While OpenAI touts its improved capabilities, skepticism arose quickly. Some questioned whether o1 pro mode's answers justify the steep price, with computer scientist Simon Willison asking for clear examples where the standard o1 model fails but o1 pro succeeds. Critics noted that early tests revealed flaws, such as struggles with Sudoku and basic optical illusion puzzles. Benchmark tests showed only slight improvements in consistency, even under stricter evaluation criteria.
OpenAI CEO Sam Altman clarified that ChatGPT Pro isn't intended for the average user, who would likely find the free or Plus tiers sufficient. Instead, the Pro plan seems targeted at niche audiences, such as experts requiring high reliability. OpenAI has offered free access to select medical researchers, emphasizing potential applications where errors could have significant consequences.
Despite its potential, some experts question the high cost. Ameet Talwalkar of Carnegie Mellon sees the tenfold price increase as risky, while UCLA’s Guy Van den Broeck doubts whether expensive reasoning models will become standard. The vague marketing pitch—claiming it excels at solving “the hardest problems” and demonstrating “intelligence”—also left many unconvinced.
OpenAI's former research chief, Bob McGrew, suggested the model's potential might be limited by its simple, text-based interface, making its "extra intelligence" harder to leverage. However, the $200 price tag sets high expectations, and so far, social media feedback suggests ChatGPT Pro has yet to deliver a compelling value proposition.
Post a Comment