Apple Hits Pause on Siri’s AI Leap

Apple's anticipated "personalized Siri" features, a key part of Apple Intelligence, have been postponed until next year's operating systems, iOS 19 and MacOS 16. Initially previewed at WWDC 2024, the upgrades promised exciting capabilities, like Siri intuitively accessing personal data—flight times, podcast recommendations, and more.

According to Apple spokeswoman Jacqueline Roy, these advanced enhancements need extra time. Recent incremental updates, such as conversational improvements, typing to Siri, and ChatGPT integration, have shown some progress. Yet, they don't meet the groundbreaking promise Apple initially showcased.

Apple rarely admits delays, underscoring the complexity here. The new Siri must accurately and privately handle sensitive personal data. For instance, asking Siri, "When is Mom’s flight landing?" involves extracting precise flight details from emails or messages, then cross-checking them against real-time flight information. Current Siri can't reliably perform these tasks.

Historically, Apple avoids announcing products prematurely. Past setbacks—like delays with AirPods and the white iPhone 4—reinforced this cautious approach. However, heightened industry excitement over generative AI likely influenced Apple’s early reveal of Apple Intelligence.

Though disappointing for users, Apple's decision seems prudent. Delivering reliable, privacy-conscious technology later is wiser than hastily shipping an incomplete feature. Apple's long-term reputation relies heavily on user trust and precision, not just meeting deadlines.

Google Unveils Gemini Embedding, Elevating AI Capabilities

Google unveiled an experimental embedding model called Gemini Embedding on Friday. It's now available through Google's Gemini developer API.

Embedding models translate words and phrases into numerical data, capturing their meaning.

These embeddings are key in applications like document retrieval and classification. They can reduce costs and speed up processes.

Gemini Embedding is Google's first embedding model trained specifically on its advanced Gemini AI family. According to Google, this training gives Gemini Embedding a deep understanding of language nuances.

It's versatile enough to perform well in finance, science, legal contexts, and general search tasks.

Google says Gemini Embedding outperforms their previous top model, text-embedding-004. It also matches top competitors on benchmark tests. Gemini Embedding can handle longer text and code segments. Additionally, it supports over 100 languages—twice as many as before.

Currently, Gemini Embedding remains experimental. Its capabilities and availability might change. Google plans a stable, general release in the coming months.

Larry Page Launches Dynatomics to Revolutionize Manufacturing with AI

Larry Page, Google's billionaire co-founder, is launching a new AI startup named Dynatomics, aimed at revolutionizing product manufacturing through artificial intelligence.

Dynatomics plans to use AI to create highly optimized designs that can streamline factory production, according to The Information.

The company is led by Chris Anderson, former CTO of Kittyhawk—Page’s previous attempt at electric flying vehicles, which ultimately shut down.

The venture joins a growing trend of startups applying AI to manufacturing. Companies like Orbital Materials are already using AI to discover advanced materials for battery technology and carbon capture systems.

Meanwhile, PhysicsX provides AI-powered simulations to automotive and aerospace engineers.

Page stepped away from Alphabet’s day-to-day leadership in 2019, handing control to current CEO Sundar Pichai. Yet, he and Sergey Brin, his co-founder, remain influential behind the scenes.

Brin, notably active within Google DeepMind, recently encouraged staff at the AI division to adopt 12-hour workdays.

Clearly, Google's co-founders aren't done pushing boundaries—even if it's now from behind the curtain.

ChatGPT Usage Surges Again Amid New Features and Mobile Growth

OpenAI’s ChatGPT has experienced a strong resurgence in user growth through late 2024 and early 2025, according to a recent report by venture capital firm Andreessen Horowitz (a16z). Initially launched in November 2022, ChatGPT quickly gained popularity, becoming the fastest app ever to reach 100 million monthly active users in just two months.

The momentum continued steadily, hitting 100 million weekly active users by November 2023. While growth slowed temporarily, taking nine months to double to 200 million users, ChatGPT has rapidly accelerated again. By December 2024, weekly active users climbed to 300 million, and by February 2025, the number reached 400 million—doubling again in under six months.

This recent surge is attributed to substantial updates and new features enhancing ChatGPT’s capabilities. Key developments include the introduction of GPT-4o, which added multimodal functions and triggered significant user spikes between April and May 2024. Further boosts in adoption followed with Advanced Voice Mode's release from July to August, and the o1 model series between September and October.

Mobile use of ChatGPT has demonstrated notably steady growth. Over the past year, the mobile app consistently added 5% to 15% new users monthly. Currently, mobile devices account for 175 million out of ChatGPT’s 400 million weekly active users.

Despite ChatGPT’s dominance, competition is increasingly visible. DeepSeek, a rising rival, quickly ascended to the No. 2 spot globally within just ten days after launching, significantly impacting ChatGPT’s mobile user base. Sensor Tower's mobile engagement data indicates that DeepSeek users are slightly more active than users of competing platforms like Perplexity and Claude, although ChatGPT still leads in overall engagement.

The report further provides insights and rankings for developers, highlighting recommended AI tools and identifying the top-performing generative AI apps across web and mobile. ChatGPT remains solidly in the top position for unique monthly visits and monthly active mobile users, as confirmed by Similarweb's market data.