Files

T

koriweb d8a80f6272 chore(wiki): dangling 링크 canonical 정규화 (768파일/1200건)

이름만 다른(표기 변형) [[위키링크]]를 대상 문서의 canonical 제목으로 치환해
끊겼던 1,200개 링크를 연결. 제목/파일명 정규화 일치만 적용하고 별칭 매칭은
과병합 위험으로 제외(애매성 가드). 원본은 _link_reconcile_backup/ 에 백업.
도구: Datacollect/scripts/link_reconcile_apply.mjs

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

2026-06-08 12:24:15 +09:00

6.6 KiB

Raw Blame History

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

title

Atmospheric Intelligence (Ambient AI)

📌 한 줄 통찰

"공기 처럼 스며든 지능". 매 screen / button X — 매 environment 자체 의 interface. 매 Matter / edge AI / privacy 의 결합. 매 user 의 의식 X 의 benefit 의 enable.

📖 핵심

매 3 element (ISTAG, 2001)

Sensitivity: 매 sensor network → 매 context 인지.
Responsiveness: 매 implicit / explicit 의 즉각 반응.
Adaptive learning: 매 user habit 의 자연 학습.

매 evolution

단계	Era	매 interface
Mainframe	1960s	매 batch
PC	1980s	매 keyboard / mouse
Mobile	2010s	매 touch
Spatial / Ambient	2020s+	매 voice + gesture + context
Zero-UI	now	매 invisible

매 component

Sensor

Motion / presence: PIR, mmWave radar (privacy 친화).
Audio: 매 wake word (Alexa, Siri).
Camera: 매 vision (privacy 의 sensitive).
Environmental: temp / humidity / CO2 / VOC.
Wearable: heart rate, accelerometer.
Smartphone: location, accelerometer, app context.

Edge AI

매 cloud round-trip X.
매 latency < 100ms.
매 privacy 의 local.
매 hardware: Apple Neural Engine, Google Edge TPU, NVIDIA Jetson.

Standard

Matter (formerly CHIP): cross-vendor smart home.
Thread (mesh networking).
Zigbee / Z-Wave (legacy).
HAP (HomeKit).

LLM 의 ambient

매 voice assistant 의 next gen.
매 always-on (privacy 의 challenge).
매 small model (Phi, Gemma) on-device.
매 multimodal (vision + voice).

매 use case

Smart home: 매 lighting, 매 climate, 매 entry.
Health monitoring: 매 wearable + AI.
Office productivity: 매 occupancy, 매 booking.
Retail: 매 customer flow, 매 dwell time.
Elderly care: 매 fall detection, 매 routine.
Vehicle: 매 driver state, 매 passenger comfort.

매 privacy challenge

Always-on listening: 매 wake word 의 false trigger.
Camera / vision: 매 most invasive.
Profiling: 매 routine 의 reveal sensitive (medical, sleep, sex).
Data aggregation: 매 silent leak.
Surveillance creep: 매 state / corp.

매 mitigation

On-device inference: 매 raw data 의 leave 의 X.
Federated learning: 매 model update 만.
Differential privacy: 매 noise.
User control: 매 mic mute, 매 camera shutter (Apple).
Data minimization: 매 keep 최소.
Audit log: 매 user 의 visibility.

💻 패턴

Matter (cross-vendor)

# 매 Matter device 의 commission (Python SDK)
from chip import controller

devnode = controller.commission(
    setup_pin_code=20202021,
    discriminator=3840,
    network='Thread',
)
# 매 device 의 fabric 에 add.
# 매 across-vendor (Apple Home + Google Home + SmartThings).

Home Assistant automation (YAML)

automation:
  - alias: "매 morning routine"
    trigger:
      - platform: state
        entity_id: binary_sensor.bedroom_motion
        to: 'on'
    condition:
      - condition: time
        after: '06:00'
        before: '09:00'
      - condition: state
        entity_id: input_boolean.weekday
        state: 'on'
    action:
      - service: light.turn_on
        target: { entity_id: light.bedroom }
        data: { brightness_pct: 30, color_temp: 350 }
      - service: media_player.play_media
        target: { entity_id: media_player.bedroom_speaker }
        data: { media_content_id: spotify:playlist:morning }

Edge inference (TensorFlow Lite)

import tflite_runtime.interpreter as tflite

interpreter = tflite.Interpreter(model_path='gesture.tflite')
interpreter.allocate_tensors()

def detect_gesture(camera_frame):
    interpreter.set_tensor(0, preprocess(camera_frame))
    interpreter.invoke()
    return interpreter.get_tensor(output_details[0]['index'])

# 매 raw frame 의 leave 의 X — 매 label 만.

Privacy-preserving presence

# 매 mmWave radar (no camera)
def detect_presence(radar_frame):
    # 매 person 의 presence + count + 매 fall
    # 매 identity X — 매 raw data X
    return presence_count, fall_alert

# Apple Watch 의 fall detection 의 same approach.

Wake-word + on-device

import openwakeword

owwModel = openwakeword.Model(wakeword_models=['hey_jarvis'])

def listen():
    while True:
        audio_chunk = mic.read(0.5)
        prediction = owwModel.predict(audio_chunk)
        if prediction['hey_jarvis'] > 0.5:
            trigger_assistant()  # 매 cloud 시작

→ 매 wake word 까지 매 on-device. 매 cloud 의 explicit consent.

🤔 결정 기준

상황	Approach
Smart home	Matter + Home Assistant
Privacy-critical	Edge AI + on-device
Cross-vendor	Matter
Voice assistant	Wake word (local) + cloud
Health monitoring	Wearable + edge ML
Elderly care	mmWave (no camera)

기본값: 매 edge-first + 매 user control + 매 minimum data.

🔗 Graph

부모: Ubiquitous-Computing · HCI · 클라우드 인프라 및 IaC 운영 표준
변형: Spatial Computing · Zero-UI
응용: Edge-AI
Adjacent: Privacy · Federated-Learning · Differential-Privacy

🤖 LLM 활용

언제: 매 ambient device design. 매 smart home automation. 매 IoT privacy review. 매 voice assistant integration. 언제 X: 매 explicit user attention 필요 task. 매 highly visual interaction.

❌ 안티패턴

Camera-first: 매 most invasive 의 default.
Cloud-everything: 매 latency + privacy + offline 의 fail.
No mute / shutter: 매 user control X.
Vendor lock-in: 매 Matter X.
Profiling 의 broad: 매 sensitive routine 의 leak.
No data minimization: 매 silent 의 hoard.

🧪 검증 / 중복

Verified (Matter spec, Apple HomeKit, Google Nest).
신뢰도 B.
Related: Smart-Home · Edge-AI · Privacy · Matter.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — sensor + Matter + Edge AI + privacy mitigation

6.6 KiB Raw Blame History