In a remarkable turn of events, Google's Bard has outperformed OpenAI's GPT-4 in the recent rankings on the LMSYS Leaderboard, signaling a significant shift in the landscape of chatbots. The leap made by Bard, now the second highest-scoring chatbot, is attributed to its recent update with Google's powerful Gemini Pro large multimodal model, shaking the long-standing dominance of GPT-4 and GPT-4 Turbo.
The LMSYS Leaderboard, maintained by the Large Model Systems Organization (LMSYS Org), has been a battleground for large language models, fostering anonymous and randomized battles akin to competitive games. Developed by the University of California, Berkeley, in partnership with the University of California, San Diego, and Carnegie Mellon University, LMSYS Org recognized Bard's ascent as a "remarkable achievement."
Bard's surge to the second position is particularly noteworthy as it challenges the stronghold of GPT-4 Turbo and GPT-4, which have consistently occupied the top two spots. Bard's recent success is tied to the incorporation of Google's Gemini Pro large multimodal model, an upgrade that catapulted it past not only GPT-4 but also Claude, another formidable competitor.
Google's Gemini Pro, introduced last December, replaced the previous PaLM 2 models, showcasing Google's commitment to advancing the capabilities of its chatbots. Bard's rise marks a significant milestone as it becomes only the second model on the leaderboard to achieve a score exceeding 1200, intensifying the competition in the Chatbot Arena.
The Chatbot Arena, utilizing the Elo rating system commonly employed in chess, is a benchmark platform for evaluating large language models through crowdsourced, randomized battles. The rankings reflect the models' performance in diverse scenarios, providing insights into their adaptability and proficiency.
Bard's triumph extends beyond its victory over GPT-4; it also surpassed Claude, securing a higher rank than Anthropic's Claude 2.1 and GPT 3.5 Turbo. The announcement by LMSYS Org captures the excitement surrounding Bard's unexpected rise, anticipating the release of Gemini Ultra to intensify the competition further.
After a shaky start, Google's Bard has undergone routine updates and integrates with various Google applications such as YouTube and Docs. Responding to user feedback on Reddit, Google is actively enhancing Bard's features with plans for dedicated mobile apps, custom instructions, and image generation.
While OpenAI's GPT-4 maintains its dominance on Stanford's HELM Leaderboard, GPT-4 Turbo follows closely in second place. Notably, the previous model that powered Bard, PaLM 2, faced tough competition from non-OpenAI models on the HELM leaderboard.
The race for supremacy in chatbots has become more dynamic than ever, with Google's Bard emerging as a formidable contender. This sets the stage for further innovations and advancements in large language models.
In a groundbreaking move, OpenAI has introduced a new feature to ChatGPT, empowering users to seamlessly incorporate GPTs (Generative Pre-trained Transformers) into any conversation by tagging them with the @ command. This innovative enhancement allows GPTs to join discussions with a comprehensive understanding of the ongoing dialogue, catering to diverse needs and queries.
OpenAI's latest feature enables users to tag specific GPTs using the @ command, bringing them into the ChatGPT conversation. The tagged GPTs become an integral part of the ongoing discussion and possess a thorough grasp of the conversation's context. Users can summon different GPTs based on their specific requirements, adding a dynamic layer to the conversational experience.
It's essential to note that this feature is currently exclusive to paying customers with access to browsing, creating, and utilizing GPTs. While this groundbreaking functionality elevates ChatGPT's capabilities, it is yet to be available to all users.
OpenAI introduced this feature through a post on X (formerly Twitter), stating, "You can now bring GPTs into any conversation in ChatGPT - simply type @ and select the GPT. This allows you to add relevant GPTs with the full context of the conversation."
This announcement follows OpenAI's recent launch of the GPT store, which is designed to assist users in discovering popular and useful GPTs akin to an app store. The GPT store was introduced as a means for developers to potentially monetize their custom GPTs in the future. However, OpenAI acknowledges that widespread adoption and usage of GPTs are crucial before implementing such monetization features.
Generative Pre-trained Transformers (GPTs) represent custom versions of ChatGPT that users can create to tailor chatbots according to their specific requirements. These GPTs, accessible to paying customers, can utilize the internet, DALL-E, and code interpreters. Beyond their inherent capabilities, developers can define custom actions by providing APIs to the GPTs, enhancing their functionality and versatility.
OpenAI launched the GPT store earlier, a repository where users can find popular and beneficial GPTs. Over 3 million custom versions of ChatGPT were created within approximately three months of the launch of GPTs, fostering a vibrant ecosystem of customizable chatbots.
While the potential for developers to monetize their custom GPTs is on the horizon, OpenAI remains focused on bolstering the adoption of GPTs within the user community. TechCrunch reports that custom GPTs constitute only 2.7% of all traffic on the OpenAI website, with a declining trend each month since their initial launch.
As OpenAI continues to innovate and introduce features like summoning GPTs into conversations, the trajectory of GPT adoption and the evolution of ChatGPT's capabilities are poised to shape the landscape of AI-driven conversational agents.