SlideShare ist ein Scribd-Unternehmen logo
1 von 29
Downloaden Sie, um offline zu lesen
Private-by-design
voice assistants
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Outline of talk
1. Lead-up to ML / voice assistant craze (tech)
2. What’s in a voice assistant? Wake Word, ASR, NLU
3. Cloud vs. on-device (Snips!)
Eric Bezzam | @EricBezzam | @snips @HdwPioneers
But first…
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Maker Kit Giveaway!
RT + follow @snips to enter the lottery!
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Maker Kit Giveaway!
RT + follow @snips to enter the lottery!
Eric Bezzam | @EricBezzam | @snips #OpenVoice
2018: 41% of US consumers (100M) own a smart speaker 

(21.5% in 2017 +90%)
source: RBC Capital Markets
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Towards voice assistants
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Towards voice assistants
Algorithms
• Perceptron (’58)
• Backprop (60s)
• CNN / LSTM (80s-90s)
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Towards voice assistants
Algorithms Hardware
70x speedup when
training with GPUs (’09)
• Perceptron (’58)
• Backprop (60s)
• CNN / LSTM (80s-90s)
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Towards voice assistants
Algorithms Hardware Internet
70x speedup when
training with GPUs (’09)
• Access to large
amounts of data
• Easier collaboration
• Perceptron (’58)
• Backprop (60s)
• CNN / LSTM (80s-90s)
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Towards voice assistants
Algorithms Hardware Internet
70x speedup when
training with GPUs (’09)
• Access to large
amounts of data
• Easier collaboration
• Perceptron (’58)
• Backprop (60s)
• CNN / LSTM (80s-90s)Deep Learning Revolution (2012)
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Towards voice assistants
Source: https://www.smartsheet.com/voice-assistants-artificial-intelligence
What’s in a voice assistant?
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Voice assistant pipeline
Wakeword
Detection
Automatic Speech
Recognition (ASR)
Natural Language
Understanding (NLU)
Action / Dialogue
—> writing an algorithm
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Voice assistant pipeline
Wakeword
Detection
Automatic Speech
Recognition (ASR)
Natural Language
Understanding (NLU)
Action / Dialogue
Eric Bezzam | @EricBezzam | @snips #OpenVoice
AUDREY: automatic digit
recognizer (1952)
Cloud vs. on-device
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Cloud based
• Rely on remote computing power for general
voice assistant (more memory and comp. power)
• Always listening, streaming when Wake Word
detected
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Cloud based
• Rely on remote computing power for general
voice assistant (more memory and comp. power)
• Always listening, streaming when Wake Word
detected
Source: 2017 Speak Easy Report
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Cloud based
• Rely on remote computing power for general
voice assistant (more memory and comp. power)
• Always listening, streaming when Wake Word
detected
Source: 2017 Speak Easy Report
Is it necessary to send voice data?
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Your voice is fundamental
to your identity
myactivity.google.com/myactivity
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Your data is safe with us
Eric Bezzam | @EricBezzam | @snips #OpenVoice
On-device (Snips)
• Rethinking solutions → specialized
models instead of general AI
• Specialized models allow smaller model
and need less data
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Cloud-level performance on a
Raspberry Pi 3 (Medium post)
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Free as in beer!
Snips Console
console.snips.ai
docs.snips.ai
NLU Training
Maker Kit
hackster.io/contests/building-for-voice
Eric Bezzam | @EricBezzam | @snips #OpenVoice
Eric Bezzam | @EricBezzam | @snips #OpenVoice
data hungry ≠ data greedy
Thank you!
docs.snips.ai
github.com/ebezzam/snips-workshop
Eric Bezzam | @EricBezzam | @snips #OpenVoice

Weitere ähnliche Inhalte

Mehr von openvoice

Mehr von openvoice (13)

Open Voice #06 - Emma Heitbrink - Bartimeus
Open Voice #06 - Emma Heitbrink - BartimeusOpen Voice #06 - Emma Heitbrink - Bartimeus
Open Voice #06 - Emma Heitbrink - Bartimeus
 
Open Voice #06 - Maarten Lens-FitzGerald - a State of Voice
Open Voice #06 - Maarten Lens-FitzGerald - a State of VoiceOpen Voice #06 - Maarten Lens-FitzGerald - a State of Voice
Open Voice #06 - Maarten Lens-FitzGerald - a State of Voice
 
Open Voice #06 - Maarten den Braber
Open Voice #06 - Maarten den BraberOpen Voice #06 - Maarten den Braber
Open Voice #06 - Maarten den Braber
 
Open Voice #05 - Maarten Lens-FitzGerald - a state of voice
Open Voice #05 - Maarten Lens-FitzGerald - a state of voiceOpen Voice #05 - Maarten Lens-FitzGerald - a state of voice
Open Voice #05 - Maarten Lens-FitzGerald - a state of voice
 
Open Voice #05 - Guido Jongen (Nuance)
Open Voice #05 - Guido Jongen (Nuance)Open Voice #05 - Guido Jongen (Nuance)
Open Voice #05 - Guido Jongen (Nuance)
 
Open Voice #04 - Sophie Kleber - Huge
Open Voice #04 - Sophie Kleber - HugeOpen Voice #04 - Sophie Kleber - Huge
Open Voice #04 - Sophie Kleber - Huge
 
Open Voice #03 - Tim van de Rijdt (Google)
Open Voice #03 - Tim van de Rijdt (Google)Open Voice #03 - Tim van de Rijdt (Google)
Open Voice #03 - Tim van de Rijdt (Google)
 
Open Voice #03 - Timo Snoek (PostNL)
Open Voice #03 - Timo Snoek (PostNL)Open Voice #03 - Timo Snoek (PostNL)
Open Voice #03 - Timo Snoek (PostNL)
 
Open Voice #02 - Maarten Lens-FitzGerald - State of Voice
Open Voice #02 - Maarten Lens-FitzGerald - State of VoiceOpen Voice #02 - Maarten Lens-FitzGerald - State of Voice
Open Voice #02 - Maarten Lens-FitzGerald - State of Voice
 
Open Voice #02 - Vanja Mlaco - Transavia
Open Voice #02 - Vanja Mlaco - TransaviaOpen Voice #02 - Vanja Mlaco - Transavia
Open Voice #02 - Vanja Mlaco - Transavia
 
Open Voice #02 - Michiel Cremers - House of Music
Open Voice #02 - Michiel Cremers - House of MusicOpen Voice #02 - Michiel Cremers - House of Music
Open Voice #02 - Michiel Cremers - House of Music
 
OpenVoice #01 - Ben Sauer - Voice UI: The past, the future and ... what now?
OpenVoice #01 - Ben Sauer - Voice UI: The past, the future and ... what now?OpenVoice #01 - Ben Sauer - Voice UI: The past, the future and ... what now?
OpenVoice #01 - Ben Sauer - Voice UI: The past, the future and ... what now?
 
OpenVoice #01 - Maarten Lens-FitzGerald - Are you talking to me
OpenVoice #01 - Maarten Lens-FitzGerald - Are you talking to me OpenVoice #01 - Maarten Lens-FitzGerald - Are you talking to me
OpenVoice #01 - Maarten Lens-FitzGerald - Are you talking to me
 

Kürzlich hochgeladen

Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Safe Software
 
Architecting Cloud Native Applications
Architecting Cloud Native ApplicationsArchitecting Cloud Native Applications
Architecting Cloud Native Applications
WSO2
 

Kürzlich hochgeladen (20)

Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...
 
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
 
AXA XL - Insurer Innovation Award Americas 2024
AXA XL - Insurer Innovation Award Americas 2024AXA XL - Insurer Innovation Award Americas 2024
AXA XL - Insurer Innovation Award Americas 2024
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
Architecting Cloud Native Applications
Architecting Cloud Native ApplicationsArchitecting Cloud Native Applications
Architecting Cloud Native Applications
 
EMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWER
EMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWEREMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWER
EMPOWERMENT TECHNOLOGY GRADE 11 QUARTER 2 REVIEWER
 
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodPolkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
 
Ransomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdfRansomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdf
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a Fresher
 
MS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectorsMS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectors
 
ICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesICT role in 21st century education and its challenges
ICT role in 21st century education and its challenges
 
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : Uncertainty
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
Corporate and higher education May webinar.pptx
Corporate and higher education May webinar.pptxCorporate and higher education May webinar.pptx
Corporate and higher education May webinar.pptx
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CV
 
presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century education
 

Open Voice #05 - Eric Bezzam (Snips)

  • 1. Private-by-design voice assistants Eric Bezzam | @EricBezzam | @snips #OpenVoice
  • 2. Eric Bezzam | @EricBezzam | @snips #OpenVoice Outline of talk 1. Lead-up to ML / voice assistant craze (tech) 2. What’s in a voice assistant? Wake Word, ASR, NLU 3. Cloud vs. on-device (Snips!)
  • 3. Eric Bezzam | @EricBezzam | @snips @HdwPioneers But first…
  • 4. Eric Bezzam | @EricBezzam | @snips #OpenVoice Maker Kit Giveaway! RT + follow @snips to enter the lottery!
  • 5. Eric Bezzam | @EricBezzam | @snips #OpenVoice Maker Kit Giveaway! RT + follow @snips to enter the lottery!
  • 6. Eric Bezzam | @EricBezzam | @snips #OpenVoice 2018: 41% of US consumers (100M) own a smart speaker 
 (21.5% in 2017 +90%) source: RBC Capital Markets
  • 7. Eric Bezzam | @EricBezzam | @snips #OpenVoice
  • 8. Eric Bezzam | @EricBezzam | @snips #OpenVoice Towards voice assistants
  • 9. Eric Bezzam | @EricBezzam | @snips #OpenVoice Towards voice assistants Algorithms • Perceptron (’58) • Backprop (60s) • CNN / LSTM (80s-90s)
  • 10. Eric Bezzam | @EricBezzam | @snips #OpenVoice Towards voice assistants Algorithms Hardware 70x speedup when training with GPUs (’09) • Perceptron (’58) • Backprop (60s) • CNN / LSTM (80s-90s)
  • 11. Eric Bezzam | @EricBezzam | @snips #OpenVoice Towards voice assistants Algorithms Hardware Internet 70x speedup when training with GPUs (’09) • Access to large amounts of data • Easier collaboration • Perceptron (’58) • Backprop (60s) • CNN / LSTM (80s-90s)
  • 12. Eric Bezzam | @EricBezzam | @snips #OpenVoice Towards voice assistants Algorithms Hardware Internet 70x speedup when training with GPUs (’09) • Access to large amounts of data • Easier collaboration • Perceptron (’58) • Backprop (60s) • CNN / LSTM (80s-90s)Deep Learning Revolution (2012)
  • 13. Eric Bezzam | @EricBezzam | @snips #OpenVoice Towards voice assistants Source: https://www.smartsheet.com/voice-assistants-artificial-intelligence
  • 14. What’s in a voice assistant? Eric Bezzam | @EricBezzam | @snips #OpenVoice
  • 15. Eric Bezzam | @EricBezzam | @snips #OpenVoice Voice assistant pipeline Wakeword Detection Automatic Speech Recognition (ASR) Natural Language Understanding (NLU) Action / Dialogue —> writing an algorithm
  • 16. Eric Bezzam | @EricBezzam | @snips #OpenVoice Voice assistant pipeline Wakeword Detection Automatic Speech Recognition (ASR) Natural Language Understanding (NLU) Action / Dialogue
  • 17. Eric Bezzam | @EricBezzam | @snips #OpenVoice AUDREY: automatic digit recognizer (1952)
  • 18. Cloud vs. on-device Eric Bezzam | @EricBezzam | @snips #OpenVoice
  • 19. Eric Bezzam | @EricBezzam | @snips #OpenVoice Cloud based • Rely on remote computing power for general voice assistant (more memory and comp. power) • Always listening, streaming when Wake Word detected
  • 20. Eric Bezzam | @EricBezzam | @snips #OpenVoice Cloud based • Rely on remote computing power for general voice assistant (more memory and comp. power) • Always listening, streaming when Wake Word detected Source: 2017 Speak Easy Report
  • 21. Eric Bezzam | @EricBezzam | @snips #OpenVoice Cloud based • Rely on remote computing power for general voice assistant (more memory and comp. power) • Always listening, streaming when Wake Word detected Source: 2017 Speak Easy Report Is it necessary to send voice data?
  • 22. Eric Bezzam | @EricBezzam | @snips #OpenVoice Your voice is fundamental to your identity myactivity.google.com/myactivity
  • 23. Eric Bezzam | @EricBezzam | @snips #OpenVoice Your data is safe with us
  • 24. Eric Bezzam | @EricBezzam | @snips #OpenVoice On-device (Snips) • Rethinking solutions → specialized models instead of general AI • Specialized models allow smaller model and need less data
  • 25. Eric Bezzam | @EricBezzam | @snips #OpenVoice Cloud-level performance on a Raspberry Pi 3 (Medium post)
  • 26. Eric Bezzam | @EricBezzam | @snips #OpenVoice Free as in beer! Snips Console console.snips.ai docs.snips.ai NLU Training Maker Kit hackster.io/contests/building-for-voice
  • 27. Eric Bezzam | @EricBezzam | @snips #OpenVoice
  • 28. Eric Bezzam | @EricBezzam | @snips #OpenVoice data hungry ≠ data greedy