The timeline surrounding GPT-5 continues to be a moving target, but a recent interview with Microsoft AI CEO Mustafa Suleyman offers insights into what GPT-5 and its successor might entail. In a discussion with AI and tech investor Seth Rosenberg, Suleyman delved into the future of generative AI and the concept of defining intelligence. Rosenberg inquired about the idea of autonomous agents and the current gap between existing chatbots and achieving that level of autonomy.
Suleyman responded by emphasizing the challenges in getting these models to consistently follow instructions with nuance over extended periods. He believes achieving this will require significantly more computational power for training, suggesting that GPT-6, not GPT-5, will be the model capable of taking consistent action. He estimates this to be about two years away.
These comments diverge somewhat from a recent interview given by OpenAI CTO Mira Murati, who alluded to a next-gen model without specifically naming it GPT-5. She described the progression as moving from toddler-level intelligence (GPT-3) to smart high-schooler intelligence (GPT-4), with the next phase aiming for Ph.D.-level intelligence for specific tasks.
While both Murati and Suleyman seem to be referencing the same development milestone, the timeline remains unclear. Suleyman’s mention of both GPT-5 and GPT-6, with GPT-6 only two years away, raises questions about the immediate future of GPT-5. Is it coming later this year followed by GPT-6 next year, or are we waiting another two years for a significant leap?
Beyond the timing, Suleyman also shared his views on the future of AI. He expressed skepticism about fully autonomous agents, considering them potentially dangerous. Instead, he advocates for “narrow lanes of autonomy,” where AI agents are deployed for specific tasks with defined limitations, and regulation plays a crucial role in maintaining control.
Suleyman also touched upon his current work at Microsoft with Copilot, focusing on refining OpenAI models and enhancing memory and personalization capabilities. Having joined Microsoft in March, Suleyman brings his extensive experience in AI, previously co-founding and leading AI at DeepMind, a company acquired by Google.
The release date of GPT-5 remains shrouded in uncertainty. GPT-4 launched in March 2023, and OpenAI currently seems more focused on expanding its ecosystem and exploring multimodal capabilities. The integration into Apple Intelligence, the upcoming low-latency voice chat, and a recent acquisition hinting at PC and desktop plans are indicative of this shift.
It appears that while the next generation of AI models is on the horizon, the exact timing and the specifics of GPT-5 remain open questions.
Editors’ Recommendations: DuckDuckGo’s new AI service keeps your chatbot conversations private, OpenAI strikes major deal with News Corp to boost ChatGPT.
