Artificial Intelligence | Swipe Insight
Perplexity AI has launched a self-hosted version of the Chinese DeepSeek R1 reasoning model for Pro users. The announcement faced backlash due to concerns about data accessibility to the Chinese government and censorship. CEO Aravind Srinivas clarified that all DeepSeek usage occurs in data centers in the USA and Europe, ensuring no data goes to China. While the model can be downloaded for local use, it may be censored based on its training data.
Meta Enhances AI Assistant with Personalization Features
Meta has enhanced its AI assistant with personalization features for tailored user experiences. A memory feature allows the AI to retain user-shared information during one-on-one conversations on WhatsApp and Messenger, learning preferences for future interactions. These features are available on Facebook, Messenger, Instagram, and WhatsApp for iOS and Android users in the U.S. and Canada.
DeepSeek Unveils R1 Model Technical Details
DeepSeek, a small Chinese AI lab, shocked the tech world by unveiling its R1 model's technical details, making its founder, Liang Wenfeng, a national figure. This model learns autonomously on a limited budget, posing a challenge to US firms like OpenAI. DeepSeek prioritizes research over profit and effectively utilizes local GPU resources. However, questions remain about its ability to stay competitive as US companies enhance their AI capabilities.
Perplexity AI Submits Revised Proposal to Merge with TikTok
Perplexity AI has submitted a revised proposal to merge with TikTok, allowing the U.S. government to own up to 50 percent of the new entity. The plan includes an initial public offering of at least $300 million, with ByteDance possibly retaining ownership. This revision follows feedback from the Trump administration. TikTok briefly went down due to a law requiring ByteDance to sell or face a ban, but the deadline was extended after Trump's intervention.
OpenAI Launches Operator for Autonomous Tasks
OpenAI has launched Operator, an AI agent that performs tasks autonomously using its own browser. Currently in research preview for Pro users in the U.S., it can handle tasks like filling forms and ordering groceries. Powered by the Computer-Using Agent model, Operator interacts with web interfaces and can self-correct. Safety measures ensure user control during sensitive actions. Operator aims to improve efficiency and engagement.
