OpenAI CEO Sam Altman has recently shared insights into what users can expect from the highly anticipated GPT-5 artificial intelligence model. While expressing cautious optimism, Altman is confident that GPT-5 will represent a significant advancement over its predecessor, GPT-4, and will not encounter insurmountable challenges. He emphasizes that GPT-5 is poised to address many of the issues currently plaguing GPT-4, particularly its struggles with reasoning and error prevention. “A lot of the things that GPT-4 gets wrong,” he notes, “can’t do much in the way of reasoning, sometimes just sort of totally goes off the rails and makes a dumb mistake, like even a six-year-old would never make.”
Drawing a parallel to the early days of the iPhone, Altman suggests that current AI models are still in their infancy, with vast untapped potential. He points out that while today’s models, including GPT-5, are relatively small compared to what’s possible in the future, significant growth and improvement are on the horizon. Interestingly, Altman’s recent remarks represent a shift from his previous stance on model size. Last year, he hinted that the era of gigantic models might be coming to an end. However, he now suggests that models will likely continue to grow, fueled by substantial investments in computing power and energy. This change in perspective reflects the dynamic nature of AI development and the evolving strategies within the field.
Altman expresses confidence that GPT-5 will address many of GPT-4’s shortcomings, particularly in reasoning and error prevention. However, he acknowledges that much work remains to be done. “We are optimistic, but we still have a lot of work to do on it,” he states, underscoring the ongoing efforts to refine and enhance the model. During his discussion, Altman also tackles some of the biggest controversies surrounding AI, particularly content licensing. He highlights OpenAI’s approach, which involves agreements with publishers to license news content for ChatGPT in exchange for training data for the models. This approach stands in contrast to companies like Google, which claim that AI-driven traffic benefits publishers—a claim met with skepticism by many, including Altman.
Altman tempers expectations regarding AI’s impact on the internet and the broader economy. He suggests that while AI may not drastically change internet use, it could still lead to significant shifts. “I think maybe AI is going to not super significantly but somewhat significantly change the way people use the internet,” Altman says. He points out that this could necessitate the evolution of economic models, highlighting the broader implications of AI beyond just training data. Altman also suggests that GPT-5 is just the beginning of a series of advancements aimed at building more sophisticated and capable AI systems. The coming months will be crucial in determining whether GPT-5 can deliver on its promise of a significant leap forward, addressing the limitations of its predecessors and paving the way for more advanced AI applications.
