The anticipated evolution of ChatGPT, rumored as GPT-5, has been a hot topic for over a year. Recently, OpenAI’s Chief Technology Officer, Mira Murati, shed light on its capabilities.
In an interview with Dartmouth Engineering shared on X (formerly Twitter), Murati compared the leap from GPT-4 to GPT-5 to progressing from high school to university. She explained that while GPT-3 systems were at a toddler-level intelligence, GPT-4 is more like a smart high-schooler. In the next couple of years, they expect AI systems to reach Ph.D. level intelligence for specific tasks. This analogy highlights significant advancements in AI capabilities.
When asked about the timeline, Murati suggested that GPT-5 might not be available until late 2025 or early 2026, which might disappoint those expecting an earlier release. Initial rumors suggested a late 2023 launch, which then shifted to a summer 2024 release. However, this turned out to be for GPT-4o, a notable update but not the major leap that GPT-5 promises.
According to Microsoft CTO Kevin Scott, future AI systems like GPT-5 will be “capable of passing Ph.D. exams” due to enhanced memory and reasoning skills. Murati clarified that the “Ph.D.-level” intelligence will only apply to certain tasks. She mentioned that these systems are already human-level in specific tasks, but there are still many tasks where they are not.
In summary, GPT-5 is set to represent a significant step forward, potentially achieving advanced task-specific intelligence by late 2025 or early 2026. While this may seem far off, the advancements in AI capabilities are promising and indicate a bright future for AI development.
