But now, there are two: im-a-good-gpt2-chatbot and im-also-a-good-gpt2-chatbot.
These models exhibited the same GPT-4 level capabilities, with some saying theyre even better than the original.
These models are even more fleeting than the first gpt2-chatbot.
Photo: Steve Jennings (Getty Images)
Most AI models on LMSYS you’re free to select from a dropdown menu.
However, the only way to access im-a-good-gpt2-chatbot and im-also-a-good-gpt2-chatbot is by navigating toLMSYS Chatbot Arena (battle).
There you could enter a prompt and hope one of the chatbots comes up randomly.
A question that tests pattern recognition. Llama got it wrong, but gpt2-chatbot was correct. (im-a-good-gpt2-chatbot: left ; Llama-right)Screenshot: LMSYS
Despite the strange nature of these AI chatbots, theyve once again drawn massive attention from the AI community.
LMSYS Org typically only works with major AI model providers, offering many of them anonymous testing services.
OpenAI and LMSYS Org did not immediately respond to Gizmodos request for comment.
Asking to identify itself. This is not always reliable, because many AI models are built on GPT-4 training data. (left: im-a-good-gpt2-chatbot ; right: Llama)Screenshot: LMSYS
Online communities have also been impressed with im-a-good-gpt2-chatbot and im-also-a-good-gpt2-chatbot.
Just like the original gpt2-chatbot, some say these AI models outperform current versions of ChatGPT.
Someone enters the room and kills one of them.
A question examining reasoning capabilities. (left: im-a-good-gpt2-chatbot ; right: Llama)Screenshot: LMSYS
Nobody leaves the room.
How many killers are left in the room?
However, the report states the company is now considering postponing the event.
The report didnt identify what the event was about.
However, Altman did not clarify whether this was or wasnt an OpenAI product.
Gizmodo was able to use im-a-good-gpt2-chatbot in the LMSYS battle arena.
It answered some questions more accurately than Metas Llama-3b-70b-instruct.
Still, theres very little information about where these gpt2-chatbots came from.
Regardless, new AI models are on the horizon, even if we dont know what their origins are.
News from the future, delivered to your present.
Why the Hell Is OpenAI Building an X Clone?
OpenAI is reportedly planning on making a social media platform because content to train on ain’t cheap.
Can This Be Right?
