Apple shuffles AI executive ranks in bid to turn around Siri

Original Article ↗ Hacker News Discussion ↗

State of Siri and “Apple Intelligence”

  • Siri is widely viewed as stagnant or regressing: struggles with basic queries (“what month is it?”, timers, simple lists), unreliable voice recognition, and stripped‑out “fun” behavior.
  • Many see the “Siri” brand as irreparably damaged; some argue Apple should retire the name as Microsoft did with Cortana.
  • Apple Intelligence launch is seen as over‑promised and under‑delivered: key features (agentic Siri, personal context, 3rd‑party integration) slipped, are region‑limited, or function weakly, prompting ridicule and even a lawsuit over advertising.
  • Some users report disabling Apple Intelligence due to instability or no visible benefit.

Why Apple Seems Behind in LLMs

  • Explanations offered:
    • Privacy positioning makes large‑scale cloud training and rich server‑side features harder, though some think “privacy” is mostly marketing.
    • Brand expectations: Apple can’t ship something as error‑prone as current LLMs without reputational risk.
    • Talent and compensation: others outbid Apple for top ML researchers.
    • Internal focus on on‑device models before hardware and RAM are really ready.
  • Counterpoint: Apple is actually strong in non‑LLM ML (ANC, crash/fall detection, imaging), but those successes are mostly outside the central AI org and invisible to users as “AI”.

Org Structure, Leadership, and Culture

  • Several commenters frame the reorg as a symptom of deep bureaucracy: too many “deciders/discussers,” not enough empowered “doers.”
  • Longstanding reports of Siri org turf wars, rule‑based legacy, and resistance to new approaches; some ex‑employees describe it as a “mini‑dinosaur” that should have been rebuilt from scratch.
  • Debate over leadership: nostalgia for a Jobs‑style product tyrant versus criticism of current execs as performative, risk‑averse, and driven by services revenue and Wall Street, not product quality.

Hardware Strength vs Software/OS Weakness

  • Broad agreement that Apple’s silicon and hardware integration are excellent and well‑positioned for on‑device AI in a few years.
  • In contrast, macOS/iOS software quality is perceived as deteriorating: bug backlogs, yearly OS churn, and features that ship half‑baked and stay that way.
  • Some argue Apple needs a “Snow Leopard”‑style multi‑year bug‑fix push far more than flashy AI features.

Voice Assistants and Product-Market Fit

  • Many doubt that full‑time voice agents are what mainstream users want; most real‑world usage is timers, reminders, basic navigation and music.
  • Others point to Android/Gemini and Alexa as proof that assistants can be genuinely useful when they integrate deeply with email, calendar, and documents.
  • Core unsolved problems: reliability (hallucinations), security (prompt injection, tool access), and making agentic behavior safe enough to, e.g., modify calendars or send emails autonomously.

Alternative Opportunities and Missed Leverage

  • Several see huge untapped potential in Apple’s existing automation hooks:
    • macOS AppleScript / Apple Events and iOS App Intents could be natural backends for an LLM‑driven agent; instead Apple has pushed Shortcuts while neglecting richer scripting.
  • Some argue Apple should:
    • Open the Siri interface to third‑party LLMs, or even lean on external models (ChatGPT, Claude) rather than insisting on first‑party everything.
    • Focus first on rock‑solid speech‑to‑text and simple, composable commands, then layer LLMs on top as creative assistants, not omniscient agents.
  • A minority remains optimistic: if any OS vendor can eventually integrate AI deeply and privately at the system level, they argue, it’s Apple—once it fixes the org and software fundamentals.