A mysterious AI chatbot called “gpt2-chatbot” was briefly available online before disappearing again. The chatbot quietly made its debut on the LMSYS Chatbot Arena website – a website used to benchmark, compare and rank various AI systems.
Based on its name, some speculate that the tool could be an earlier version of OpenAI’s GPT-2 chatbot language model. However, users have noted that the language model appears to be as powerful or even more powerful than GPT-4, OpenAI’s newer and more advanced language model.
In fact, some Internet users found that the language model performed better than GPT-4 in certain tests. This has led to speculation that the “gpt2 chatbot” could be an early prototype of GPT-5 or perhaps an updated, advanced version of GPT-4, which can essentially be considered GPT-4.5.
Thank you for the incredible enthusiasm of our community! We really didn’t see that coming.
Just a few things to clarify:
– In line with our policy, we have worked with several model developers in the past to provide the community with access to unpublished models/checkpoints (e.g.…)
— lmsys.org (@lmsysorg) April 30, 2024
But users who managed to test the model before it was taken offline found that there was surprisingly little information about what the language model was and where it came from. Still, it didn’t take long for the language model to be taken offline again, with LMSYS saying in a tweet: “In keeping with our policy, we have worked with several model developers in the past to provide the community with access to unreleased models/checkpoints.” (e.g. mistral-next, gpt2-chatbot) for preview testing.”
The site then added that it had to “temporarily” disable the gpt2 chatbot due to “high traffic and high capacity limits.”
Speculation about “gpt2 chatbot” is increasing
Thanks to a subsequent tweet According to OpenAI CEO Sam Altman, the language model appears to be something new rather than an earlier model of GPT-2.
Altman wrote, “I have a soft spot for GPT-2,” before later editing the tweet to appear as “gpt-2.” And to add more fuel to the fire, OpenAI employee Steven Heidel wrote: tweet say: “if gpt-2.”
Based on these answers, it seems more likely that this is an unpublished model, as suggested by LMSYS.
Featured Image: Franz Bachinger out of Pixabay