A mysterious new AI chatbot referred to as “gpt2-chatbot” is popping heads this week after it grew to become out there on a serious massive language mannequin benchmarking website, LMSYS Org. Nobody is aware of the place it got here from, however many think about it to have roughly the identical capabilities as OpenAI’s GPT-4. This places gpt2-chatbot in a uncommon class of AI fashions that solely a handful of builders worldwide have been in a position to obtain.
“Nobody is aware of who made it or what it’s, however I’ve been enjoying with it a bit and it seems to be in the identical tough skill degree as GPT-4,” Ethan Mollick, a Professor researching synthetic intelligence on the Wharton Faculty of the College of Pennsylvania, mentioned in a tweet on Monday.
On-line AI communities have gone wild concerning the nameless gpt2-chatbot. One X person claims that gpt2-chatbot almost coded an ideal clone of the cell sport Flappy Chicken. One other X person says it solved an Worldwide Math Olympiad downside in a single shot. On lengthy Reddit threads, customers are speculating wildly concerning the origins of the gpt2-chatbot and arguing over whether or not it’s from OpenAI, Google, or Anthropic. There’s no proof for these claims, however tweets from OpenAI CEO Sam Altman and different executives have simply added gasoline to the hearth.
You may check out the gpt2-chatbot your self at LMSYS Org’s web site. Navigate to “Direct Chat” or “Area (side-by-side)” and choose it from the dropdown menu. LMSYS Org says in its coverage weblog that sure AI mannequin builders can check nameless unreleased fashions earlier than a broader launch. This has led many to consider that gpt2-chatbot is an nameless mannequin from a serious AI developer.
“Simply to make clear, following our coverage, we’ve partnered with a number of mannequin builders to convey their new fashions to our platform for neighborhood preview testing,” mentioned LMSYS Org in a tweet on Monday, responding to a thread about gpt2-chatbot. “These fashions are strictly for testing and received’t be listed on the leaderboard till they go public.”
LMYSYS Org and OpenAI didn’t instantly reply to Gizmodo’s request for remark.
In Gizmodo’s restricted testing, we discovered the gpt2-chatbot has capabilities which might be just like main AI fashions from Anthropic and OpenAI. It exhibited habits unique to superior massive language fashions, reasoning effectively and outlining detailed plans for classy duties. Listed below are a few of our examples evaluating gpt2-chatbot (left) and Anthropic’s Claude Opus mannequin (proper).
A pc engineering professor on the College of Wisconsin discovered that gpt2-chatbot may carry out a activity that different main AI fashions couldn’t. Dimitris Papailiopoulos requested gpt2-chatbot to resolve a math riddle that entails studying some inexplicit guidelines. AI largely struggles to reply questions like this.
In the end, there’s little or no data out there concerning the gpt2-chatbot simply but. Nonetheless, it appears clear that an influence participant is behind this AI mannequin. Within the coming weeks, the creator and origins of the gpt2-chatbot will doubtless turn out to be public. This might imply a brand new AI mannequin is on the horizon or perhaps there’s a new AI developer on the scene.
