Sam Altman, CEO of OpenAI, has responded to criticism of GPT-5, claiming that the latest language model marks a significant advancement in scientific artificial intelligence. Despite its challenging launch in August 2025, Altman asserts that GPT-5 is already beneficial in real-world research, aiding scientists in accelerating breakthroughs in areas like physics, biology, and mathematics. In an interview with Wired, Altman acknowledged that initial feedback largely focused on user experience and technical difficulties, which overshadowed GPT-5’s primary goal of enhancing scientific knowledge. He stated, “The vibes were kind of bad at launch, but now they’re great.” The rollout of GPT-5 encountered various issues, including technical problems during OpenAI’s live demonstration, leading to user complaints about its engagement level compared to GPT-4.
Critics suggested that OpenAI had set unrealistic expectations and that GPT-5 failed to meet them, with some speculating that interest in generative AI was waning. However, Altman contends that these assessments overlook the model’s intended purpose. He emphasized that GPT-5 is not just an incremental upgrade; it represents a major milestone, where users are impressed by its capability to contribute to significant scientific tasks. Altman noted, “GPT-5 is the first time where people are, ‘Holy fuck. It’s doing this important piece of physics.’ Or a biologist is saying, ‘Wow, it just really helped me figure this thing out.’” Although OpenAI has not provided specific instances, Altman views these advancements as pivotal moments, signaling AI’s potential to aid scientific discovery.
Greg Brockman, OpenAI’s president, supported this view, indicating that GPT-5 excels in specialized fields rather than general conversations. Altman pointed out that most users may not immediately recognize the depth of the model’s capabilities. Mark Chen, OpenAI’s head of research, mentioned that GPT-5 performs on par with top Math Olympians, showing a significant enhancement over GPT-4. Critics have raised concerns that GPT-5 exposes the limitations of simply scaling models, but OpenAI asserts that the improvements stem from more sophisticated training methods rather than increased size. The model employs reinforcement learning with guidance from human experts and can generate and learn from its own data.
Brockman explained that when a model is less capable, the focus is on training a larger version, but as it becomes smarter, the goal shifts to leveraging its capabilities. OpenAI is investing heavily in expanding its data centers in Texas and other locations, gearing up for the development of even more advanced models. Brockman compared this process to rocket science, where each upgrade becomes increasingly complex yet essential for advancing AI. Altman also shared that OpenAI is changing its perspective on Artificial General Intelligence (AGI). Rather than seeing AGI as a definitive goal, it is now viewed as a dynamic process that transforms the global economy.
Altman stated, “The scientific progress definition is really a big deal for the world,” stressing that the real breakthroughs will stem from AI-facilitated discoveries rather than merely replacing human work. Looking to the future, Altman expressed confidence: “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.”