From two new releases of GPT-4, o1 to a new browser and internal drama, ChatGPT and parent company OpenAI had a notable second year. What’s next?
ADVERTISEMENTOn November 30, 2022, OpenAI launched its first ChatGPT model to the world.
What should have started as a test for OpenAI models quickly became the chatbot that has become synonymous with development generative artificial intelligence (genAI).
In 2024, OpenAI saw the launch of several new versions, including GPT-4 which came with faster intelligence in text, voice and vision, and o1, a new series of models that can understand through complex tasks in science, coding and mathematics. .
Just a few weeks ago, OpenAI released SearchGPT, a search engine extension that provides “fast, timely responses” to user queries with relevant web sources, bypassing search engines entirely.
The company has been dealing with some pretty internal problems behind the scenes, with the resignation of co-founder Ilya Sutskever, the disbanding of the company’s super-intelligence research team, and a series of lawsuits from US news companies for alleged copyright infringement.
So what’s next for OpenAI ChatGPT as it tries to up the ante in its third year? Euronews Next breaks it down.
OpenAI’s next ‘giant breakthrough’
In the latest Reddit Ask Me Anything (AMA). monthfounder of OpenAI Sam Altman and his colleagues gave a little insight into their priorities for the third year.
Kevin Weil, director of product at OpenAI, told the AMA that the “big topic” of 2025 will be whether ChatGPT can perform tasks on its own.
Altman has suggested that it could look like an autonomous agent, which he would consider the company’s next “giant breakthrough”.
“AI agents,” called agent AI, will allow companies to design large-scale language models (LLMs) that power their systems to automate tasks in the workplace.
It’s a feat that some of OpenAI’s rivals, such as Google Cloud Vertex AI, are already working on agents, LinkedIn and Microsoft. Media reports say that the next update to Google Gemini, Project Jarvis, could also include autonomous agents.
Kate Devlin, professor of artificial intelligence and society at King’s College London, said there has been a mixed reaction to the AI agents that OpenAI’s competitors have already launched.
“Some people are really positive about this and see it as a complete game changer because it allows for much more of a personal assistant type of situation where you can outsource a lot of your tasks to AI,” she said.
“There are some people who are wary of that and don’t like the idea of giving that much input or control to the model.”
The next model of ChatGPT: What we know so far
There have been some predictions that OpenAI will release a brand new model sometime before the end of the year.
ADVERTISEMENTAltman told Reddit OR October that there are “very good releases” coming later in the year, but “none that we would call ChatGPT-5.”
The priority, according to Altman, is to “prioritize the delivery” of other models, such as GPT4.0 and 0.1 that launched this year.
Reporting from Verge a few weeks later the release of Orion, which is believed to be the successor to both GPT-4o and o1, puts it in December for select companies working closely with OpenAI to be able to build their own products and features.
Tadao Nagasaki, CEO of OpenAI Japan, teased in September a future model of ChatGPT that will be “100 times more powerful” than GPT-4, according to local media say.
ADVERTISEMENTDevlin said it’s possible that AI agent activity could be included in any future Orion release.
Other updates to look for could be Sora, the AI text-to-video model, which has been delayed until now.
The software, which is still being “refined” according to Weil, was leaked this week.
Altman said the following version DALL-Eimage generation software from OpenAI, is “worth the wait” but still has no release plan.
ADVERTISEMENTShould OpenAI consider specialization or reduction?
One of the things that OpenAI and other companies will have to watch out for in 2025 are some of the overall flaws that are taking over the industry, Devlin continued.
“[AI companies] they need more compute, they need more power, they need more data,” Devlin said. “So, how much can I do with limitations on those things?”
One option that OpenAI might consider is downsizing to small or medium LLMs that are less intensive.
These smaller models could “curate data” to make it more specific or useful in certain domains, such as law or health, Devlin said.
ADVERTISEMENT“Instead of just increasing and increasing and increasing, it’s time to think about what we’ve got so far because there are definite benefits … but we know there are risks and maybe we should take some time to assess those risks,” Devlin said .
Abdul Sadka, director of the Sir Peter Rigby Digital Futures Institute at Aston University, said OpenAI should keep ChatGPT “generic” so that individual companies or sectors can refine it to the specific data sets they want it to be used for.
However, Sadka said he could see OpenAI expanding to give ChatGPT more “modalities,” such as the ability to recognize medical images to “potentially give you diagnostic reports on any underlying condition” a patient may have.
To help ChatGPT become more specialized, Sadka said companies using the software could build an external knowledge base that the AI hasn’t seen before to “reduce the likelihood of … hallucinations,” a term used to explain the AI’s best responses to does not recognize questions.
ADVERTISEMENT