AI Models Pursue “Superintelligence,” But Still Fail on Simple Questions
OpenAI plans to increase the price of its ChatGPT Plus subscription to $44 per month
Jensen Huang, CEO of Nvidiastated in a recent interview with ‘Wired’ that “everyone should have an AI tutor.” This tutor, according to him, can be a chatbot that helps structure and synthesize complex information.
Huang showed a particular interest in biomedical research, especially AI-assisted drug discovery, an area where language models play a critical role in identifying therapeutic opportunities faster than traditional methods. To keep up with advances in this field, Huang uses chatbots, which They allow you to organize information effectively and drill down into specific details more quickly.
Although the best-known chatbots are usually ChatGPT or Bard, Huang surprised by revealing his preference for Perplexity AIa less popular model. Developed by Aravind Srinivas, former OpenAI researcher, Perplexity is billed as “the world’s first conversational response engine.” Huang highlighted its simplicity and ability to facilitate research, calling Perplexity his personal “tutor.”
Huang’s interest in these tools is understood in the context of Nvidia’s central role in the development of artificial intelligence. Nvidia, with its GPUs and data centers, has created the essential infrastructure to run large AI models that power chatbots like ChatGPT and Perplexity. The company, according to Huang, continues to develop projects such as “AI factories”, infrastructures that will be crucial for the future of technology.
Huang also addressed the evolution of data processing in the industry. He mentioned the importance of overcoming the limitations of Moore’s Law, which has traditionally set the pace of growth in chip power. Nvidia has taken steps, such as the acquisition of Mellanox, to circumvent these limits and improve data center performance.