Sam Altman asked ChatGPT users what they want to see in 2025, and dropped some clues about what to expect
OpenAI's CEO took to X to hear what people want from ChatGPT in 2025, and his replies are full of clues about what to expect
Since its launch in 2022, ChatGPT has been at the tip of the spear when it comes to AI's presence in the mainstream. ChatGPT is to AI as Google is to search engines. Yes, there are others, but OpenAI's chatbot has seemingly captured the zeitgeist of the LLM sphere.
In the two years since its public launch, ChatGPT hasn't just been amassing a considerable userbase (which as of August 2024 is reported to be over 200 million active users per week, double the figure that was reported in November 2023), it has released several new models including GPT-4, 4o, and o1—each inching the OpenAI's chatbot ever closer to the company's long-standing goal of developing AGI (Artificial General Intelligence).
Not only that, ChatGPT gained several new features including, advanced voice mode, custom instructions, user-made GPTs accessible through the new GPT Store, advanced editing and writing tools through Canvas, mobile apps, and the ability to browse the web for up-to-date information about any given topic, effectively eliminating the chatbot's knowledge cutoff date.
The latter has now evolved into ChatGPT search, a feature similar to Google's AI overview for Search, only OpenAI's version is yet to recommend users eat rocks and glue or make chlorine gas in their washing machines.
Taking all of this into account, along with the launch of the text-to-video model Sora and the new 1-800 ChatGPT phoneline, it's been a busy and feature-packed few years for the OpenAI team.
So what's next? Well, while OpenAI still has many irons in the fire with upcoming projects, headlined by the upcoming release of GPT-o3, Sam Altman took to X on Christmas Eve to ask the internet at large, "What would you like OpenAI to build/fix in 2025?"
Several suggestions caught the attention of the OpenAI CEO enough for him to answer, some of which may tell us more about what to expect from ChatGPT next year and beyond.
Stay in the know with Laptop Mag
Get our in-depth reviews, helpful tips, great deals, and the biggest news stories delivered to your inbox.
what would you like openai to build/fix in 2025?December 25, 2024
ChatGPT: CEO Sam Altman fields suggestions on X
Open AI CEO Sam Altman gained several thousand replies to his late Christmas Eve post, with a number being responded to personally. It's in those responses we can gain more insight into what we might expect to see from OpenAI over the next 12 months.
- Strong 4o replacement- GPT/o/sora all seamless integrated in chatgpt- 50/70 bucks plan, a middleground?- Long context- Knowledge cutoff! Aggressive updates! PLEASEYou guys are doing AMAZING work, the ultimate wishlist only has AGI, but we'll get there nail after nail!December 25, 2024
Replying to the above post, Altman stated "Very good list! Hopefully, you'll be quite happy with us next year overall."
But what exactly from that list can we expect to see? Well, as far as seeing a strong GPT-4o replacement, the final reveal for OpenAI's "12 Days of OpenAI" did reveal that GPT-o3 and GPT-o3 Mini would be the next model we'll see from the company, expected to be released in January 2025.
Whether or not this model satisfies the commenter's "Long context" request remains to be seen. However, in another reply, Altman suggests, "Rate limits up, context up, and costs down!" GPT-o3 is claimed to be three times as accurate as the o1 model launched in September 2024.
As for AGI, we know this is OpenAI's ultimate goal. Still, a 2025 achievement of this feels like wishful thinking.
The vector store you guys have for assistants api is really good - make that its own retrieval api. Would instantly become the top retrieval product.Video input modality would be nice.You’re already doing agents, but excited for those.Unexpected? Let’s see a hardware play.December 25, 2024
Altman responded to this post using only the "Eyes" emoji (👀), suggesting that there's something here for us to look out for.
Aside from the API talk, I'd say that there's a high chance that Altman is hinting at video input for ChatGPT, especially with GPT-4o's multimodal capabilities.
We heard in November that OpenAI is also planning to launch an AI agent tool to help automate certain tasks for users, so we expect to hear more about that in 2025 also.
However, Altman may be giving a nod to the talk of hardware. It's something of an open secret that OpenAI is currently working with Apple hardware alum Jony Ive on developing a product they hope could become the "iPhone of AI." Will we see this product revealed in 2025?
Just make sora really goodDecember 25, 2024
Given that Sora has only just had its public launch and is already facing stiff competition from Veo 2, Kling, and Hailuo AI, it's no wonder that Altman is ready to confirm that the fledgling text-to-video generator will be seeing improvements, simply replying "coming."
Release gpt-4o image generation supportDecember 25, 2024
While Altman is a little short on giving an actual guarantee to this suggestion, the OpenAI CEO's "Hope to!" response suggests that GPT-4o image generation is at least on the cards.
Family accounts. Let me create accounts for my kids with guard rails. Let their curiosity take off, but within reasonable limits, as determined by the parent. Maybe we could even get insights about our kids from their usage!December 25, 2024
On the topic of family accounts for ChatGPT, Altman appears to be sold. "This is a good idea!" doesn't exactly scream "We're on the case," but it would allow parents to introduce younger members of the family to OpenAI's chatbot in a safe and controlled environment.
As for gaining insights about your kids from the chatbot, personally, that seems slightly intrusive to entrust to OpenAI, but the general idea is sound.
What's next
While Altman wasn't overly forthcoming with his replies, his X thread certainly drew in a lot of suggestions for how OpenAI can refine the ChatGPT experience further.
There seems to be a lot of interest in pulling Sora's capabilities into the main ChatGPT model, similar to how the text-to-image generator Dall.E has been.
When it comes to better-performing models, we'll be seeing GPT-o3 go public in January next year, and 2025 should also see the release of the much anticipated GPT-5.
If you'd like to see more of the many suggestions made by ChatGPT users and a handful of other responses from Altman, you can catch the full thread on X.
More from Laptop Mag
- Scarlett Johansson: ChatGPT stirred controversy with "her" voice in 2024
- "Is that Instagram photo AI?" Meta tried to police AI fakery in 2024 with an ever-changing policy
- The New York Times made headlines by suing OpenAI this year, but the case against AI isn't so black and white
Rael Hornby, potentially influenced by far too many LucasArts titles at an early age, once thought he’d grow up to be a mighty pirate. However, after several interventions with close friends and family members, you’re now much more likely to see his name attached to the bylines of tech articles. While not maintaining a double life as an aspiring writer by day and indie game dev by night, you’ll find him sat in a corner somewhere muttering to himself about microtransactions or hunting down promising indie games on Twitter.