Ed H. Chi, a distinguished researcher associated with Google, participated in a press conference held in Taiwan. He is actively engaged in the development of the conversational AI, Bard, and underscored Google's robust commitment to the advancement of AI software and hardware. Google employs its proprietary Tensor Processing Unit (TPU) for in-depth exploration of deep learning products.
Originally hailing from Taiwan and based at Google's headquarters in the United States, Ed H. Chi leads the machine learning division at Google DeepMind. The team's focus spans large language models (LLM), neural network recommendation systems, Bard, and associated research.
Chi's academic accolades encompass securing bachelor's, master's, and doctoral degrees from the University of Minnesota in a span of just six and a half years. Since 2013, he has made significant contributions, spearheading over 700 enhancements spanning diverse Google services, including YouTube, news, advertising, and the Google Play store.
In February 2023, Google unveiled Bard, a dynamic conversational AI, progressively enhancing its features over time. Bard's functionalities encompass supporting programming, furnishing location-specific insights (with responses potentially differing by location), and accommodating more than 40 languages.
Nonetheless, certain observers find Google Bard comparatively less impressive in contrast to OpenAI's ChatGPT. These observers are inquisitive about Google's strategic initiatives to fortify its position within the domain of AI.
Although Bard is presently accessible for utilization, it is currently designated as an experimental feature rather than an official product. The team is enthusiastically seeking a deeper comprehension of user expectations. The training and inference procedures for Bard and its associated LLM, virtually all products related to deep learning, harness Google's exclusive TPUs.
Google introduced TPUs into their data centers in 2016. On the mobile front, integration of TPUs commenced with the debut of the Pixel 6 series in 2021, marking a departure from the prior utilization of Qualcomm Snapdragon chips.
Google unveiled its PaLM 2 and Gemini models during its I/O conference in May. It was specifically highlighted that PaLM 2 supports enhancements across 25 products, including Bard. Meanwhile, the Gemini model is still undergoing training and is poised to eventually accommodate multimodal applications and divergent parameter configurations. Noteworthy is Google DeepMind CEO Demis Hassabis' earlier assertion that Gemini's capabilities will outpace the models employed in ChatGPT.
Ed H. Chi stressed the pivotal role of the Gemini model as a cornerstone project within DeepMind, destined to shape Google's LLM strategy and provide support for Bard. Nevertheless, the formal emergence of Gemini as a product is yet to be realized.