• The Tensor
  • Posts
  • Apple’s new iPhones ship with Siri that actually works

Apple’s new iPhones ship with Siri that actually works

Anthropic’s CPO wants Claude to be your next co-worker

Welcome to today’s edition of The Tensor, get smarter with executive insights on the latest in AI & Tech Industry, 5 min reads, 3x a week.

In today's Tensor:

  • New iPhone 16s ship with Siri that actually works

  • Anthropic’s CPO wants Claude to be your co-worker

  • Can you generate better ideas than an AI 🤔? Stanford research says no.

  • Quick Bytes: Sony announces PS5 Pro with AI upscaling, AI startup Glean raises $260M to build “Google for Work” and 3 more.

Read Time: 4.5 mins

The scoop: Fall is here, and right on queue, Apple has once again come out with it’s “Most Advanced iPhones ever”. They announced new iPhone 16s, Watch 10 and more which are slight improvements over the previous versions, but the event was centre stage was Apple Intelligence - AI, and with that the last of the Mag7 jumps into the AI ring.

What’s new:

  • The iPhones ship with A18/A18 pro, which have a 30% CPU performance increase YoY, and double the AI performance.

  • Writing Tools enable quick proofreading, rewriting, changing tones and even creating custom emojis anywhere you type.

  • New Visual AI: iPhones now have an additional camera button and visual intelligence, which lets you capture photos, get quick insights, and even search visually, coming on par with releases earlier this year from Samsung and Google.

  • Siri is finally good! Powered by Apple’s own LLMs, in my testing it’s much better at understanding my queries and pulls in relevant context, finally I can use it to do more than set timers.

Why it matters: Apple's 15-year silicon efforts are paying off huge, with some of the most advanced edge processors for AI, enabling private and real-time on-device computing. This local processing aligns with their brand and strategy focused on privacy. The main differentiator where other big tech companies will struggle to catch up.

Bottom line: While some users might question the need for AI-powered emojis, Apple's intelligent invasion is a bold step towards a future where our devices are more like more useful than ever.

The scoop: Mike Krieger, Instagram co-founder and Anthropic's new Chief Product Officer, outlined his strategy for developing AI products that live up to the hype on an interview with the Verge. After Instagram and Artifact, he joined Anthropic because of the safety-minded approach and research prowess. Krieger aims to transform Claude from a capable chatbot into an essential workplace tool.

Highlights:

  • Enterprise focus: Claude for Enterprise is launching soon, offering features like SSO, SCIM, and data management. Krieger sees potential for AI to become a trusted coworker, handling complex tasks autonomously.

  • Novel pricing: Exploring value-based billing models beyond subscriptions. Krieger envisions pricing tied to actual value delivered, potentially charging based on task completion or time saved.

  • Long-term capabilities: Emphasis on evolving AI from "intelligent amnesiacs" to assistants that can plan and execute over longer time horizons without constant human input.

  • Responsible development: Commitment to AI safety is baked into Anthropic's processes, with extensive testing and ethical considerations integrated into product development.

Why it matters: Anthropic's product strategy is strong, focusing on practical and work-focused applications and positioning itself as a serious contender in the rapidly expanding enterprise AI market. It’s focus on safe AI and fast innovation is quickly making it the leading AI company alongside OpenAI

Bottom line: With Krieger at the helm, Anthropic is betting that the path to AI success isn't through your phone's home screen, but through your office's front door. It's less "Hey Siri" and more "Hey, can you finish this report by EOD?”

The scoop: A recent Stanford study is getting traction on the discovery that Large Language Models (LLMs) can generate more novel research ideas than human experts in NLP. However, these AI-generated ideas were judged slightly less feasible, highlighting both the potential and limitations of AI in research ideation.

Details:

  • Over 100 NLP researchers gathered to participate in the study and were asked to generate novel ideas and review AI generated ones by Claude Sonnet 3.5.

  • Researchers grounded the LLM in existing literature and allowed it access to research papers.

  • The ideas underwent blind review, with evaluators unaware of whether humans or AI generated them.

  • The result? LLM ideas were significantly more novel (p < 0.05) but slightly less feasible than human ideas, showcasing AI’s creative potential but also it’s limitations.

Bottom line: This study opens new questions. Can you use LLMs reliabily in research? Can they think? What it does point to is that ideating with LLMs is like having a brainstorming session with Einstein but you still need to do the math yourself.