In February this year, Google revealed Bard, an experimental conversational AI service, powered by LaMDA, to compete with OpenAI’s ChatGPT. Bard draws on information from the web to provide fresh, responses. Instead of using a full language model like OpenAI GPT-4, Google used lightweight model version of LaMDA. This smaller model requires significantly less computing power, allowing Google to scale to more users.
In March, Google opened up its AI-powered chat experience to the general public. Google Bard was made available at bard.google.com in the U.S. and U.K.
However, Bard was less capable when compared to ChatGPT. Recently, Alphabet CEO Sundar Pichai revealed that Google would soon upgrade Bard to the more advanced PaLM models, which would enhance its capabilities in areas such as reasoning, coding, and answering math questions, k, bringing Bard closer to the level of ChatGPT and Bing chatbot.
Today, CNBC reported that Google will reveal PaLM 2, its direct competitor to OpenAI GPT-4 at Google I/O conference this week. Also, Bard will be using PaLM 2 to offer new generative AI experiences including creative writing, text analysis, coding, solving math problems and more. PaLM 2 will support more than 100 languages including Japanese and Korean, and was developed under the internal codename “Unified Language Model” inside Google.
Google is also working on several different versions of Bard including a multi-modal version called “Multi-Bard”, “Big Bard” and “Giant Bard.”
Finally, Google will be adding the capability to create image based on text description, rivalling Bing Image Creator and other Dall-E based services.