After OpenAI’s eventful 12-day festival, company CEO Sam Altman held an open discussion about X and asked users, “What do you want OpenAI to build/fix in 2025?” I asked. As expected, the post went viral, with many users openly expressing their wants and needs for AI.
One user quickly expressed the need for a drag-and-drop user interface with multiple separate models that could be chained together to generate and adjust output. This discussion was previously brought up on OpenAI’s developer forums, where a user said, “You will get better results if you split the task into multiple models, each optimized to perform one task well. I think so,” he said.
The executive called the users’ suggestions “interesting ideas” that he hadn’t considered. Another user prefers a more powerful GPT-4o alternative, seamless integration of GPT/o/Sora into ChatGPT, longer contexts, proactive updates, and a “mid-range” paid subscription plan around $50-70. We have shared a comprehensive list of requests, including: OpenAI CEO Sam Altman suggested that users would be very happy with the company’s progress this year, albeit inexplicably.
Interestingly, another user called on executives to loosen the guardrails on next-generation AI models, claiming it would be “cleaner.” Sam Altman suggested the need for an “adult mode.”
family account. Let me create an account for my kids with guardrails. Let your child’s curiosity be free within the reasonable limits determined by the parents. Maybe you can even get some insight into your kids from how they use them! December 25, 2024
At the same time, another user requested a family account that would allow them to create an account for their child with guardrails. Users stated that this approach would elicit curiosity. However, users said it is most important for parents to set limits to prevent the system from getting out of control.
Finally, @davidlach of X called for a proper agent, prompting Sam Altman to say “Happy 2025.” As you may know, OpenAI is reportedly preparing to ship an AI agent called Operator in January 2025. Late last year, chief product officer Kevin Weil said:
“I think 2025 will be the year that agent systems finally go mainstream.”
What to expect from OpenAI in 2025 and beyond
Sam Altman posted a seemingly endless thread on X asking users what they’d like to see OpenAI build or fix in 2025, but the executive said there are a few updates we can expect in 2025. or benchmarks.
OpenAI may release a successor to GPT-4o with better features. ChatGPT makers’ advanced AI models may have better memory for longer contexts when generating prompts. An “adult mode” with fewer guardrails may also be possible, but its development raises safety and privacy concerns. Finally, Sora is getting an important update that could make it even more capable and have longer videos.
Common themes: AGIagents are much improved 4o upgrades are much improved Longer memory Longer contexts “adult mode” Deep investigation improvements Soramore personalization (interestingly, many great updates coming soon) was never or barely mentioned!) https://t.co/lMZmlZif66December 30, 2024
Of course, AGI was at the top of Sam Altman’s list after the ChatGPT maker widely published its o1 inference model, despite rumors that it may have already reached the coveted benchmark. Interestingly, around the same time, another report stated that OpenAI plans to eliminate strict AGI clauses that could sever its partnership with Microsoft in order to secure future funding and access to cloud computing. It was suggested that.
Based on the information being highlighted, OpenAI may already have reached the coveted benchmark, or at least is one step away from it. Interestingly, Sam Altman says that AGI will be possible within the next five years with today’s hardware and will be successful with surprisingly little impact on society.
That being said, OpenAI may have more surprises for users in 2025. As Sam Altman pointed out, “Many great upcoming updates are either never mentioned or barely mentioned.”
