AI summary by TwinMind
Karen Lellouche Tordjman: Siri, Alexa, Google ... what comes next? | TED
TED
10m 26s
Opening Scenario: Future Smart Assistant Capabilities
- Voice assistant provides personalized morning notification about pants on sale
- Based on smart glasses data from previous week showing interest in similar item
- Suggests skipping movies to afford purchase
- Offers to find party host gift based on budget and Instagram profile
- Speaker finds scenario both thrilling and scary as working mom
- Exciting: eliminate Google searches, Amazon scrolling, budget calculations, calendar optimization
- Potential capabilities: heart rate-based Starbucks order adjustments, lunch/steps-based workout tailoring, coordinated evening plans with friends' assistants
- Emphasizes need for regulation and user control over data sharing
Current State of AI Recommendation Engines
- Already in widespread daily use
- 70% of YouTube viewing time on algorithm-recommended videos
- Existing tools: wristbands for sleep/workout tracking, Google Home/Amazon Alexa, car voice systems
- Common limitation: all tools operate in single specific area (home, car, health)
- Current generation "stays in the lane"
Two Critical Missing Technological Components
Voice Recognition Challenges
- Current assistants don't understand everything users say
- Siri still provides confused answers to simple questions from children
- Human language understanding difficulties:
- Context-dependent meaning beyond just words
- Accents and background noise complications
- Example: difficulty for Americans understanding French-accented English
Breadth of Recommendations Challenge
- Current systems limited to specific tasks vs. true companion functionality
- Today: recommendations within single category (dress vs. dress, book vs. book)
- Future need: cross-category recommendations (book vs. dress)
- Required capabilities: meeting notes, milk reordering, mental health coaching
Industry Investment and Progress
Voice Technology Development
- Google 2018: launched investment program for Google Home suite startups
- Invested in 15+ companies since launch
- Amazon: 10,000 employees working on Alexa voice technologies
- Expected future capabilities:
- Context understanding (music "change" vs. temperature "change")
- Long, complex request processing with mid-sentence changes
- Far field speech recognition across rooms with background noise
- Mood detection (joy, sadness, annoyance) and emotional mimicry
- Amazon Echo speaker technology driving tremendous recent progress
- Natural language processing advances enabling refined voice interactions
Algorithm Requirements for Breadth
- Need powerful algorithms processing myriad data points
- Required functions: pattern identification, action modeling, user feedback learning
- Cross-category recommendation capability essential
Business Implications
Three Key Imperatives for All Companies
- Data: accelerated data collection and processing
- Tech: algorithm development and voice-enabled interactions
- Trust: consumer confidence for recommendation acceptance
- Not limited to smart assistant companies - affects every business
Future Predictions and Tipping Point
- Voice + breadth challenge resolution = tipping point for mass adoption
- Smart assistants will become as essential as smartphones today
- Key difference: embedded in every device (smartphone, car, mirrors, fridge, glasses)
- Evolution from convenient tool to essential daily life helper
- Users won't be able to live without smart assistants
