Google employees test Big Bard, a smarter chatbot that takes on ChatGPT
Big Bard: Googlers’ Secret AI Upgrade
Google has asked employees to test its public‑ready chatbot, Bard, but insiders are already using a more powerful internal version called “Big Bard.” The two chatbots share the same LaMDA language model, yet Big Bard’s larger parameters give it a noticeably smarter and more human‑like feel.
What Makes Big Bard Stand Out?
- Richer Responses – Big Bard produces longer, more detailed answers than the lightweight Bard model.
- More Conversational Tone – It chats in a casual, informal style, sometimes even swearing.
- Higher Intelligence – Despite using the same LaMDA core, the increased parameter count makes it appear more intelligent.
Why Google Wants a “Lightweight” Public Version
CEO Sundar Pichai announced that the first public release of Bard will be based on a “lightweight” model requiring less computing power.
Key reasons for limiting the launch include:
- Cost Control – AI searches could be ten times more expensive than traditional keyword searches, according to Alphabet chair John Hennessy.
- Risk Management – A smaller initial rollout helps guard against possible inaccuracies or rogue behavior that could spark negative headlines.
Internal Testing Guidelines
When employees test Big Bard, they receive a directive to use the standard Bard model for the official assessment. Some leaders have explicitly requested that staff only work with the regular version.
Earlier internal tests even named the prototype “Apprentice Bard.”
Related Stories
Business Insider delivers the breakthroughs you crave
Artificial‑intelligence competition shows no sign of easing
- Andreas Braun, the chief technology officer for Microsoft Germany, revealed that OpenAI plans to launch the successor to its language model, GPT‑4, next week.
- GPT‑4 will be multimodal, meaning it will process video, images, and audio in addition to text.
- A spokesperson from Google did not answer a request for comment.
Got insider info? Share it securely
Contact him via encrypted email at hlangley@protonmail.com or through Signal/Telegram on +1 628‑228‑1836.