Key Takeaways
- Sam Altman admits GPT-5 had a rocky launch but claims it’s now performing well
- Experts argue GPT-5 delivered only incremental improvements, falling short of AGI promises
- OpenAI president Greg Brockman says GPT-5’s gains came from RLHF, not just scaling
- Altman confidently predicts GPT-6 will be a “significant” upgrade over GPT-5
OpenAI CEO Sam Altman has acknowledged GPT-5’s problematic launch while expressing confidence that the next iteration, GPT-6, will represent a major advancement. Despite initial setbacks, Altman insists the model is now performing well and contributing to scientific discovery.
From Rocky Start to Recovery
Sam Altman openly admitted that OpenAI “totally screwed up” aspects of GPT-5’s rollout, requiring the company to make behavioral adjustments to create warmer responses. However, he now claims “the vibes were kind of bad at launch. But now they’re great.”
Altman emphasized GPT-5’s unique contribution: “There’s something important happening that did not happen with any pre-GPT-5 model, which is the beginning of AI helping accelerate the rate of discovering new science.”
Expert Criticism and Modest Gains
Many AI experts and users found GPT-5 underwhelming, describing its improvements as incremental rather than revolutionary. New York University professor Gary Marcus, a leading critic, stated: “GPT-5 was the most hyped AI system of all time. It was supposed to deliver two things, AGI and PhD-level cognition, and it didn’t deliver either of those.”
Marcus argued that GPT-5’s modest performance gains demonstrate OpenAI can no longer rely solely on scaling data and computing power to achieve artificial general intelligence.
OpenAI’s Defense and Technical Approach
In response to criticism, OpenAI president Greg Brockman contended that GPT-5’s capabilities resulted from reinforcement learning based on human feedback (RLHF) rather than simply larger datasets or more computing power.
Looking Ahead to GPT-6
Despite the controversy, Altman remains optimistic about OpenAI’s future models. “What I can tell you with confidence is GPT-6 will be significantly better than GPT-5, and GPT-7 will be significantly better than GPT-6,” he stated. “And we have a pretty good track record on these.”
