DeepMind’s CEO, Demis Hassabis, believes that the company’s new chatbot, Gemini, will surpass ChatGPT by combining large-language models with problem-solving through reinforcement learning. Gemini will utilize techniques from AlphaGo, which famously beat a chess champion in 2016. The goal is for Gemini to make tough decisions and outperform other large-language models such as GPT-4 through reinforcement learning. DeepMind sees AI’s reliance on learning from text rather than feedback as a limitation and aims to address this by incorporating AlphaGo’s ability to learn from previous actions and feedback. Google recently merged DeepMind with its Brain AI department to accelerate AI development and plans to compete directly with ChatGPT and Microsoft’s similar tools by launching an enhanced version of its AI tool, Bard. DeepMind is striving to establish a new standard with Gemini, following its acquisition by Google in 2014. DeepMind has achieved notable progress in its robotics division, including testing an algorithm that can manipulate robot arms. However, Hassabis emphasizes the need to balance progress with the potential risks of AI and plans to expose DeepMind’s technology to outside research for a better understanding of these risks. In related news, OpenAI has its own chatbot, ChatGPT, and aims to target corporate commercial opportunities. OpenAI’s CEO, Sam Altman, envisions ChatGPT as a virtual assistant capable of drafting emails with updated business information. This puts ChatGPT in direct competition with Salesforce, Microsoft’s business partner, and Google’s smart response feature. Salesforce has already launched ChatGPT for Slack, while Microsoft has incorporated Copilot into its Office product suite. Additionally, Zoom’s IQ tool uses OpenAI’s model to summarize meetings and generate email responses.
Continue reading on Beincrypto.com
By
By
By
By
By