Siri AI

Original Article ↗ Hacker News Discussion ↗

Overall reaction to Siri AI

  • Many find the keynote and website demos underwhelming: they mostly show timers, directions, photo search, light Shortcuts, and message lookups that feel like things Siri should have done years ago.
  • Some see real potential in deeper OS integration and cross‑app context (Spotlight semantic indexing, knowledge‑graph‑like features) and think this is the first serious “Star Trek computer” step if it actually works.

Performance, reliability, and UX

  • Latency in Apple’s marketing video looks unrealistically low; real‑world beta reports say it’s substantially slower than old Siri for simple tasks, likely due to more server‑side calls.
  • Longstanding Siri frustrations remain a major theme: basic failures with alarms, home control, media playback, and settings search have eroded trust. Several people say Apple needs to fix core dictation, keyboard, and settings search before layering AI on top.
  • Battery‑life impact of on‑device models is a concern, but currently unclear.

Branding and naming

  • “Siri AI” and “Apple Intelligence” are widely mocked as uninspired (“AI AI”) and confusing; many argue Siri’s brand is so tarnished Apple should have rebranded entirely.
  • Others say consumers care more about what it does than what it’s called.

Hardware and feature gating

  • Confusion over device matrix: base Apple Intelligence works on iPhone 15 Pro/16 and M1+ Macs, but the “most powerful” on‑device model (e.g., expressive voices, advanced dictation) requires iPhone 17 Pro/Air and M3/M4‑class hardware with ≥12 GB RAM.
  • Users who bought recent devices marketed as “built for Apple Intelligence” feel misled; lawsuits are mentioned.

EU, DMA, and third‑party AI

  • Siri AI on iOS/iPadOS is delayed in the EU; Apple blames DMA requirements to give competing assistants equivalent deep access to personal data and app actions.
  • Some see this as a genuine privacy issue; others call it Apple’s bargaining chip to protect lock‑in and App Store economics.
  • There’s debate over whether a permission‑based API could safely expose the same capabilities to OpenAI/Anthropic/etc., balancing user choice vs. privacy.

Dependence on Google Gemini

  • Multiple commenters think most cloud‑side capabilities are effectively Gemini (or Gemma) under an Apple skin, citing:
    • Image style resemblance to Gemini’s “nano banana.”
    • Text verbosity and table‑heavy answers similar to Gemini.
  • This raises questions about how much is truly Apple‑developed versus Google‑licensed, and how “Private Cloud Compute” interacts with Google‑hosted models.

Shortcuts, agents, and use cases

  • Strong interest in Siri orchestrating Shortcuts as an “agentic harness” across apps; some call it the first mainstream OpenClaw/Hermes competitor.
  • Others find Shortcuts’ block‑based editor painful and want a real scripting language or even user‑defined “Siri APIs” so they can wire the assistant into arbitrary workflows or local models.
  • Many note that current consumer demos (rewriting emails, removing objects from photos, summarizing a 1‑line message) feel trivial versus truly hard, high‑value tasks (complex travel booking, multi‑step personal automation).

Trust, safety, and opt‑out

  • Concerns about hallucinations, mis‑acting agents (e.g., password auto‑change, mass deletions), and prompt‑injection from emails/webpages.
  • Some praise Apple’s Private Cloud Compute design and “no data retained” claims; others remain skeptical and want a global off switch or ability to plug in their own local/third‑party models instead.
  • Overall sentiment: cautious curiosity, but Apple has a long way to go to overcome a decade of Siri disappointment and to prove this is more than a slow, restricted Gemini wrapper.