Chatbot rivalry is as intense as ever, with DeepMind claiming it could beat ChatGPT
DeepMind’s Gemini will do the planning, solve problems and analyse text
While ChatGPT has garnered significant global interest, DeepMind, the research laboratory owned by Google, asserts that its upcoming large language model will potentially match or surpass OpenAI’s achievements.
Recent updates indicate that DeepMind is leveraging techniques utilised in AlphaGo, an AI system developed by DeepMind that famously defeated a professional human player in the board game Go. This technology will be used to create a chatbot named Gemini. The chatbot is being designed with the intention of rivalling the capabilities of ChatGPT.
According to DeepMind CEO Demis Hassabis, if all goes according to the plan, Gemini will be able to plan or solve problems as well as analyse text. Hassabis explained that, in essence, Gemini merges the advantageous aspects of AlphaGo-like systems with the remarkable language capabilities exhibited by large models. He further expressed enthusiasm for the project, mentioning that they have incorporated innovations that promise to be highly intriguing.
Knight speculates that Gemini, which was briefly previewed during Google’s I/O developer conference in May, will utilise advancements in reinforcement learning to overcome challenges faced by current language models. Reinforcement learning entails providing rewards or punishments to an AI system based on its performance, aiming to instruct the system on the desired actions. This approach is anticipated to enhance Gemini’s capabilities in tackling the most challenging tasks.
