AI teddy bears that read personalized bedtime stories to kids may become the norm in 5 years, says toymaker CEO. But, privacy and security concerns must be taken seriously before releasing these toys to the market.
Generative language programs like ChatGPT have limitations, including telling jokes. Researchers found ChatGPT produced variations of one of 25 joke set-ups 90% of the time. While the AI can break down what makes a joke funny, its ability to do so was a result of prior training and memorization. ChatGPT learned a specific joke pattern, but its generation and identification of jokes focuses on their meaning and content. ChatGPT3.5 represents progress towards AI's understanding of humor, but humans still reign supreme in the realm of jokes.
OpenAI announces updates to its API and reduced prices for services. New models include extended context length and function calling capabilities to guide responses. Stay tuned for GPT-3.5 and GPT-4.
Tech giants are prohibiting employees from using ChatGPT due to security concerns, leaving workers uncertain of the legal implications. Even if not explicitly banned, utilizing this AI chatbot for sensitive data analysis may result in serious repercussions, including being fired or facing legal action. Legal and HR experts highlight that policies must be followed to avoid potential legal entanglements and being blacklisted.
Asus is set to launch their rental business AI server, AFS Appliance, using an AI language model on-site to address data control issues and security concerns. The service will be powered by Nvidia chips, equivalent to OpenAI's GPT-3.5. Monthly rental prices range between $6,000 and $10,000. Initially targeting enterprise customers in Taiwan, Asus plans to expand internationally in 2023.