SearchGPT enhances search capabilities, Meta SAM2 can segment videos, Apple ushers in the era of local LLMs — the top 3 AI news stories of the week
Our latest AI Digest covers the biggest breaking AI news of the week. Ihar Nestsiarenia, Lead Machine Learning Engineer at EPAM, comments on key stories.
#1 — SearchGPT is announced
OpenAI has introduced SearchGPT, a prototype that integrates AI with real-time web data for enhanced search capabilities. It is similar to Perplexity and recent Google AI search updates announced at the Google I/O 2024 conference. SearchGPT plans to provide up-to-date information and answers to questions that incorporate resource links. A key feature is OpenAI’s plan to partner with publishers and content creators to potentially resolve content usage issues. This approach may shape the industry. Perplexity also plans to start sharing advertising revenue with media companies whose content is used in the platform's responses. For now, you can only join the waitlist for SearchGPT, since it is initially being made available to just 10,000 test users.
#2 — Meta SAM 2 can segment videos now
Meta Segment Anything Model 2 (SAM 2) extends the impressive capabilities of the original SAM to video. Using a transformer architecture with streaming memory, SAM 2 is the first tool able to process videos in real time, creating detailed maps of objects. This model differs from typical visual image processing — it offers more precise and predictable results. It is particularly useful for robotic systems and video analytics. Using SAM saves millions of hours that would otherwise have been required for manual annotation.
#3 — Apple ushers in the era of local LLMs
The latest iOS 18.1 developer beta rolls out Apple Intelligence, notable for its integration across products and its new features. One key feature is the use of local LLMs enhanced by specialized adapters (LoRA) for tasks like summarization and error correction. Even with a relatively small model hosted on your phone, Apple’s approach enhances the model’s utility and performance. This update marks the beginning of the local LLM era. In addition, Apple has enabled function calling, external APIs for developers, and has released a comprehensive technical report.
.png?auto=webp)
