OpenAI’s GPTs: A Game-changing Week in AI

AI is an industry that moves quickly, so keeping up can be challenging. OpenAI held its first developer conference, unveiling new products including an updated GPT-4, text-to-speech models, and an API for image-generating DALL-E 3. The most significant announcement was OpenAI’s GPTs, allowing developers to build conversational AI systems powered by the company’s models and publish them on the GPT Store. This democratizes generative AI app creation. Samsung unveiled its own generative AI family, Samsung Gauss. Microsoft announced free AI compute for startups. YouTube is experimenting with new generative AI features. Google DeepMind’s head of robotics discussed the company’s robotic ambitions. Kai-Fu Lee’s AI startup released its first model, the open source Yi-34B. GitHub teased a customizable Copilot plan and a new partner program. Hugging Face’s two-person model team has been impressive in developing capable tools. Mozilla released its first large language model, Fakespot Chat. In terms of machine learning, Google has shown success predicting earthquakes and weather patterns. MetNet-3 uses physics-based weather models to produce accurate predictions. Researchers at the University of Kansas developed a detector for AI-generated text in chemistry journal articles. While narrow, it offers promise for future use cases.