A new mysterious chatbot seems to be the talk of the town it seems. And, no, it’s not OpenAI’s highly speculated GPT-5. Although we are aware that the Sam Altman-led company is quietly working on its most advanced AI model, there is nothing official about it, not yet at least. The chatbot that has been making waves in the AI community is known as “gpt2-chatbot”.
The chatbot first appeared on Chatbot Arena, an LLM benchmark platform and it reportedly showcased advanced capabilities. Gpt2-chatbot’s advanced capabilities have also fanned rumours that it could be an early OpenAI stealth test of GPT-4.5.
Based on the details found on social media handles and popular AI newsletters such as ‘The Rundown, the model lacks official documentation or attribution. However, it reportedly identified itself as ChatGPT and was created by OpenAI when asked.
Those who have used it have heaped praise on the model for its advanced abilities of reasoning, maths, coding, and even ASCII art skills which reportedly matched or exceeded GPT-4 and above models during testing.
Reportedly, Stanford University student and AI researcher Andrew Gao claimed that the gpt2-chatbot solved an International Math Olympiad problem. On the other hand, reportedly, Wharton professor Ethan Moillick also shared similar thoughts on gpt2’s performance when compared to GPT-4.
Speculations also indicate that OpenAI could be quietly benchmarking for a new model before it is made publicly available. To add fuel to the fire, Sam Altman also took to his X (formerly Twitter) account to post – “i do have a soft spot for gpt2.”
i do have a soft spot for gpt2
— Sam Altman (@sama) April 30, 2024
The latest development is important. It needs to be noted that there is always a certain degree of mystery around the workings of OpenAI. It could be possible that this model is just the rumour mills going abuzz over nothing. However, real accounts praising the impressive capabilities of the new model have turned it into a matter of contention in the AI community.