Mysterious ‘gpt2-chatbot’ launched. Could this be a stealth test of OpenAI GPT 4.5?
Daniel Levi
Posted On April 30, 2024
In the AI world, OpenAI is known for its secretive projects. But when a mysterious chatbot dubbed ‘gpt2-chatbot‘ suddenly emerged online, it sparked a flurry of questions: Could this be a covert trial run for OpenAI’s highly anticipated GPT 4.5?
Yesterday, the Chatbot Arena witnessed the debut of a remarkable new AI model under the enigmatic moniker “gpt2-chatbot.” With its advanced capabilities on full display, speculation quickly arose about its potential connection to OpenAI’s next big release.
Despite the absence of official documentation or attribution, the gpt2-chatbot model allegedly identifies itself as ChatGPT and its affiliation with OpenAI when prompted. Users have been quick to laud its knack for reasoning, mathematics, coding, and even ASCII art, suggesting performance on par with or surpassing that of GPT-4 and beyond in initial trials.
With gpt2-chatbot, users can ask questions to two anonymous models like ChatGPT, Claude, or Llama, and then vote on which one gave the better response. The team running gpt2-chatbot also mentioned that they’ve gathered over 700,000 votes from people to create an Elo leaderboard for LLM (large language model). It sounds like they’re using a lot of data to compare and rank these models.
“Ask any question to two anonymous models (e.g., ChatGPT, Claude, Llama) and vote for the better one! You can continue chatting until you identify a winner. Vote won’t be counted if model identity is revealed during conversation.”
Some conjecture points to the possibility of OpenAI quietly conducting benchmark tests for an upcoming model prior to its public unveiling. Adding fuel to the speculative fire, Sam Altman later hinted at his fondness for ‘gpt2.’
i do have a soft spot for gpt2
— Sam Altman (@sama) April 30, 2024
Why it’s significant: OpenAI has always shrouded itself in mystery, and while this new model may simply be the result of rampant gossip, its impressive capabilities are catching the attention of leading figures in the AI community, regardless of its origins or purpose.
