What to Anticipate from OpenAI in 2025: Insights from CEO Sam Altman

After wrapping up its lively 12-day Shipmas celebration, OpenAI’s CEO Sam Altman took to X and posed a question to the community: “What would you like OpenAI to build or fix by 2025?” This sparked a whirlwind of responses as users eagerly shared their hopes and needs regarding AI developments.

One of the standout suggestions came from a user who envisioned a drag-and-drop interface featuring multiple models that could be linked to create and refine outputs. This concept had already sparked discussions in OpenAI’s Developer Forum, with users suggesting that better results could be achieved by allocating specific tasks to individual models optimized for those functions.

Altman found this proposal intriguing, admitting it was something he hadn’t considered. Another participant provided a detailed wish list, requesting a fortified GPT-4o successor, an integration of GPT/o/Sora with ChatGPT, longer context handling, frequent updates, and a more affordable subscription plan in the $50-$70 range. Although the specifics weren’t entirely clear, Altman hinted that this year would bring developments that should please the user.

A different user boldly requested a reduction in restrictions on forthcoming AI models, believing this would lead to a “cleaner” outcome. Altman acknowledged that there could be a need for a “grown-up mode” in their AI offerings.

In a related conversation, another participant proposed family accounts, designed to allow parents to create child-safe accounts with appropriate limitations. This system could nurture young curiosity while ensuring parental control over what their children access.

Finally, as @davidlach on X called for proper AI agents, Altman responded with a cheerful “Happy 2025.” Notably, OpenAI is reportedly set to launch an AI agent known as Operator in January 2025. Kevin Weil, OpenAI’s Chief Product Officer, had previously remarked on this, suggesting that 2025 might be the year when agentic systems truly achieve mainstream success.

As we look ahead to what OpenAI might deliver in 2025 and beyond, it’s realistic to anticipate several exciting updates and innovations. There’s talk of a successor to GPT-4o, poised to offer heightened capabilities. OpenAI’s sophisticated models could come with enhanced memories and extended prompt contexts. A potential “grown-up mode” might introduce fewer restrictions, though this does come with safety and privacy challenges. Additionally, Sora might experience significant enhancements, possibly extending to longer video processing.

The anticipation surrounding AGI also deserves attention. Despite swirling speculation that OpenAI might have already reached a crucial benchmark with its o1 reasoning model, the company seems poised to push forward. Interestingly, another report suggested that OpenAI is considering removing a strict AGI clause that currently poses challenges to its Microsoft partnership, in the hopes of securing more funding and cloud computing resources.

Altman suggested that hitting the AGI milestone is feasible with today’s hardware within the next five years, and he projected that its arrival might go largely unnoticed by society. Yet, the biggest surprises from OpenAI in 2025 might still be under wraps, as Altman teased, “Many great updates we have coming were mentioned not at all or very little!”