SlideShare uma empresa Scribd logo
1 de 24
Baixar para ler offline
Machine Intelligence at Google
Rob Craft, CloudIntel Lead
robcraft@google.com
@KindaRobCraft
2
Google’s mission is to organize the world’s information
and make it universally accessible and useful.
3
one important technology we use is neural networks
OUTPUTINPUT
4
neural net models learn from examples
labeled photos
“cat”
“dog”
“car”
“apple”
“flower”
OUTPUT
5
neural net models learn from examples
Make tiny adjustments to
model so output is closer to
label for a given image
labeled photos
“cat”
“dog”
“car”
“apple”
“flower”
OUTPUT
6
after a model is trained, you can test it
?
unlabeled photo
7
after a model is trained, you can test it
unlabeled photo
“cat”
8
Input Output
“rice”
“restaurants in Seoul”
“hello!”
“A close up of a small child holding a stuffed animal.”
powerful functions that neural nets can learn
안녕하세요
9
Android
Apps
Gmail
Maps
Photos
Speech
Search
Translation
YouTube
and many others ...
Used across products:
2012 2013 2014 2015
1500
1000
500
0
Number of directories containing model description files
rapidly accelerating use of deep learning at Google
DeepMind- AlphaGo
Demis Hassabis
Atari 2600 testbed: 100+ Atari games from the 70/80s
Agents just get raw pixels as inputs (~30K)
Wired up to action buttons but NOT told what they do
Goal is simply to maximize score
Everything learnt from scratch, ZERO pre-programmed knowledge
ONE agent with ONE set of parameters to play ALL the different games
THINK OF A ROBOT in the arcade watching the screen controlling the joystick
DQN Deep Q-network- Atari testbed
Image: Gnome Enterprises
Breakout
General Atari Player
14
Street nameStreet number
Street View
Sign
Business facade
Sign
Business name
Traffic light
Traffic
sign
Street number
15
[glacier]
Google Photos
15
16
Google Translate
16
17
Smart reply
in Inbox by Gmail
10%
of all responses
sent on mobile
18
signal
for Search ranking,
out of hundreds
improvement
to ranking quality
in 2+ years
#3 #1
Search
machine learning for search engines
RankBrain: a deep neural network for search ranking
19
beyond core products, into areas like health and robotics
“Deep Learning for
Robots: Learning from
Large-Scale Interaction”,
Google Research Blog,
posted March, 2016
Sharing our tools with researchers and developers
around the world
repository
for “machine
learning”
category on GitHub
#1
Released in Nov.
2015
2121
Machine Learning Use Cases
Structured Data
Classification/ Regression
● Customer Churn Analysis
● Product Diagnostics
● Forecasting
Recommendation
● Content Personalization
● Product X-Sells/Up-sells
Anomaly Detection
● Fraud Detection
● Asset Sensor Diagnostics
● Log Metric Anomalies
Unstructured Data
Image Analytics
● Identify damaged shipments
● Explicit Content Classification
● Identify “styles” in images
Text Analytics
● Call Center log analysis
● Language Identification
● Topic Classification
● Sentiment Analysis
2222
Ready to use Machine
Learning models
Use your own data to
train models
Cloud
Vision API
Cloud
Speech API
Cloud
Translate API
Cloud Machine Learning
Develop - Model - Test
Google
BigQuery
Stay
Tuned….
Cloud
Storage
Cloud
Datalab
NEW
Alpha
GA BetaGA
Alpha
Beta
GA
2323
beyond core products
for applications like
health and robotics
smart platforms
to accelerate
progress worldwide
Google products
to create great
experiences for users
machine intelligence is already making a huge difference
and there are many, many more opportunities
23
2424
Questions?
(thanks)
@KindaRobCraft

Mais conteúdo relacionado

Semelhante a Google Photos neural network learns from labeled examples to recognize images

Adapt or Die: Keynote with Anant Jhingran
Adapt or Die: Keynote with Anant JhingranAdapt or Die: Keynote with Anant Jhingran
Adapt or Die: Keynote with Anant JhingranApigee | Google Cloud
 
6 Open Source Data Science Projects To Impress Your Interviewer
6 Open Source Data Science Projects To Impress Your Interviewer6 Open Source Data Science Projects To Impress Your Interviewer
6 Open Source Data Science Projects To Impress Your InterviewerPrachiVarshney7
 
Artificial intelligence in android development
Artificial intelligence in android developmentArtificial intelligence in android development
Artificial intelligence in android developmentanikeshkumar11
 
Seminar report on Google Glass, Blu-ray & Green IT
Seminar report on Google Glass, Blu-ray & Green ITSeminar report on Google Glass, Blu-ray & Green IT
Seminar report on Google Glass, Blu-ray & Green ITAnjali Agrawal
 
10 Things Every PHP Developer Should Know About Machine Learning
10 Things Every PHP Developer Should Know About Machine Learning10 Things Every PHP Developer Should Know About Machine Learning
10 Things Every PHP Developer Should Know About Machine Learning10x Nation
 
10 Things Every PHP Developer Should Know About Machine Learning
10 Things Every PHP Developer Should Know About Machine Learning10 Things Every PHP Developer Should Know About Machine Learning
10 Things Every PHP Developer Should Know About Machine LearningChristopher Mohritz
 
Google Analytics Konferenz 2018_Machine Learning / AI mit Google_Lukman Ramse...
Google Analytics Konferenz 2018_Machine Learning / AI mit Google_Lukman Ramse...Google Analytics Konferenz 2018_Machine Learning / AI mit Google_Lukman Ramse...
Google Analytics Konferenz 2018_Machine Learning / AI mit Google_Lukman Ramse...e-dialog GmbH
 
Scaling Up Machine Learning: How to Benchmark GraphLab Create on Huge Datasets
Scaling Up Machine Learning: How to Benchmark GraphLab Create on Huge DatasetsScaling Up Machine Learning: How to Benchmark GraphLab Create on Huge Datasets
Scaling Up Machine Learning: How to Benchmark GraphLab Create on Huge DatasetsTuri, Inc.
 
Google report 01
Google report 01Google report 01
Google report 01Sunny Gupta
 
Top no-code AI Platforms
Top no-code AI PlatformsTop no-code AI Platforms
Top no-code AI Platformsaitech1
 
Connector Corner: Extending LLM automation use cases with UiPath GenAI connec...
Connector Corner: Extending LLM automation use cases with UiPath GenAI connec...Connector Corner: Extending LLM automation use cases with UiPath GenAI connec...
Connector Corner: Extending LLM automation use cases with UiPath GenAI connec...DianaGray10
 
The AI Spring - Innovations for the next decade
The AI Spring - Innovations for the next decadeThe AI Spring - Innovations for the next decade
The AI Spring - Innovations for the next decadeZinnov
 
Latent Panelists Affinities: a Helixa case study
Latent Panelists Affinities: a Helixa case studyLatent Panelists Affinities: a Helixa case study
Latent Panelists Affinities: a Helixa case studyGianmario Spacagna
 
Jeff Dean at AI Frontiers: Trends and Developments in Deep Learning Research
Jeff Dean at AI Frontiers: Trends and Developments in Deep Learning ResearchJeff Dean at AI Frontiers: Trends and Developments in Deep Learning Research
Jeff Dean at AI Frontiers: Trends and Developments in Deep Learning ResearchAI Frontiers
 

Semelhante a Google Photos neural network learns from labeled examples to recognize images (20)

Adapt or Die: Keynote with Anant Jhingran
Adapt or Die: Keynote with Anant JhingranAdapt or Die: Keynote with Anant Jhingran
Adapt or Die: Keynote with Anant Jhingran
 
GOOGLE
GOOGLEGOOGLE
GOOGLE
 
Presentation1
Presentation1Presentation1
Presentation1
 
6 Open Source Data Science Projects To Impress Your Interviewer
6 Open Source Data Science Projects To Impress Your Interviewer6 Open Source Data Science Projects To Impress Your Interviewer
6 Open Source Data Science Projects To Impress Your Interviewer
 
Stock_price _prediction.pptx
Stock_price _prediction.pptxStock_price _prediction.pptx
Stock_price _prediction.pptx
 
Google Tools & services
Google Tools & servicesGoogle Tools & services
Google Tools & services
 
OpenAI.pptx
OpenAI.pptxOpenAI.pptx
OpenAI.pptx
 
Artificial intelligence in android development
Artificial intelligence in android developmentArtificial intelligence in android development
Artificial intelligence in android development
 
Seminar report on Google Glass, Blu-ray & Green IT
Seminar report on Google Glass, Blu-ray & Green ITSeminar report on Google Glass, Blu-ray & Green IT
Seminar report on Google Glass, Blu-ray & Green IT
 
10 Things Every PHP Developer Should Know About Machine Learning
10 Things Every PHP Developer Should Know About Machine Learning10 Things Every PHP Developer Should Know About Machine Learning
10 Things Every PHP Developer Should Know About Machine Learning
 
10 Things Every PHP Developer Should Know About Machine Learning
10 Things Every PHP Developer Should Know About Machine Learning10 Things Every PHP Developer Should Know About Machine Learning
10 Things Every PHP Developer Should Know About Machine Learning
 
Alphago Tech talk
Alphago Tech talkAlphago Tech talk
Alphago Tech talk
 
Google Analytics Konferenz 2018_Machine Learning / AI mit Google_Lukman Ramse...
Google Analytics Konferenz 2018_Machine Learning / AI mit Google_Lukman Ramse...Google Analytics Konferenz 2018_Machine Learning / AI mit Google_Lukman Ramse...
Google Analytics Konferenz 2018_Machine Learning / AI mit Google_Lukman Ramse...
 
Scaling Up Machine Learning: How to Benchmark GraphLab Create on Huge Datasets
Scaling Up Machine Learning: How to Benchmark GraphLab Create on Huge DatasetsScaling Up Machine Learning: How to Benchmark GraphLab Create on Huge Datasets
Scaling Up Machine Learning: How to Benchmark GraphLab Create on Huge Datasets
 
Google report 01
Google report 01Google report 01
Google report 01
 
Top no-code AI Platforms
Top no-code AI PlatformsTop no-code AI Platforms
Top no-code AI Platforms
 
Connector Corner: Extending LLM automation use cases with UiPath GenAI connec...
Connector Corner: Extending LLM automation use cases with UiPath GenAI connec...Connector Corner: Extending LLM automation use cases with UiPath GenAI connec...
Connector Corner: Extending LLM automation use cases with UiPath GenAI connec...
 
The AI Spring - Innovations for the next decade
The AI Spring - Innovations for the next decadeThe AI Spring - Innovations for the next decade
The AI Spring - Innovations for the next decade
 
Latent Panelists Affinities: a Helixa case study
Latent Panelists Affinities: a Helixa case studyLatent Panelists Affinities: a Helixa case study
Latent Panelists Affinities: a Helixa case study
 
Jeff Dean at AI Frontiers: Trends and Developments in Deep Learning Research
Jeff Dean at AI Frontiers: Trends and Developments in Deep Learning ResearchJeff Dean at AI Frontiers: Trends and Developments in Deep Learning Research
Jeff Dean at AI Frontiers: Trends and Developments in Deep Learning Research
 

Mais de TheFamily

Building a design culture from day one
Building a design culture from day oneBuilding a design culture from day one
Building a design culture from day oneTheFamily
 
Individual Contributors vs Managers
Individual Contributors vs ManagersIndividual Contributors vs Managers
Individual Contributors vs ManagersTheFamily
 
Build the decentralized team you ever dreamed of
Build the decentralized team you ever dreamed ofBuild the decentralized team you ever dreamed of
Build the decentralized team you ever dreamed ofTheFamily
 
CEOs best practices to win time back & focus on what matters
CEOs best practices to win time back & focus on what mattersCEOs best practices to win time back & focus on what matters
CEOs best practices to win time back & focus on what mattersTheFamily
 
Managing fully remote teams
Managing fully remote teamsManaging fully remote teams
Managing fully remote teamsTheFamily
 
State of European Tech by Atomico
State of European Tech by AtomicoState of European Tech by Atomico
State of European Tech by AtomicoTheFamily
 
Building a real estate startup
Building a real estate startupBuilding a real estate startup
Building a real estate startupTheFamily
 
A VC view on Enterprise Sales
A VC view on Enterprise SalesA VC view on Enterprise Sales
A VC view on Enterprise SalesTheFamily
 
Find your style and create emotions
Find your style and create emotionsFind your style and create emotions
Find your style and create emotionsTheFamily
 
From product to ecosystem
From product to ecosystemFrom product to ecosystem
From product to ecosystemTheFamily
 
Demystifying the product black box
Demystifying the product black boxDemystifying the product black box
Demystifying the product black boxTheFamily
 
The secrets to create bank brand love
The secrets to create bank brand loveThe secrets to create bank brand love
The secrets to create bank brand loveTheFamily
 
Building an insurance startup with Alan, Luko, Coverd & Balderton
Building an insurance startup with Alan, Luko, Coverd & BaldertonBuilding an insurance startup with Alan, Luko, Coverd & Balderton
Building an insurance startup with Alan, Luko, Coverd & BaldertonTheFamily
 
Mixing Product & Tech by Jean Lebrument, CTO & CPO at Brigad
Mixing Product & Tech by Jean Lebrument, CTO & CPO at BrigadMixing Product & Tech by Jean Lebrument, CTO & CPO at Brigad
Mixing Product & Tech by Jean Lebrument, CTO & CPO at BrigadTheFamily
 
A new breed of CTO - Philippe Vimard, CTO & COO at Doctolib
A new breed of CTO - Philippe Vimard, CTO & COO at DoctolibA new breed of CTO - Philippe Vimard, CTO & COO at Doctolib
A new breed of CTO - Philippe Vimard, CTO & COO at DoctolibTheFamily
 
Building a logistics startup  with Trusk, Totem & SpaceFill
Building a logistics startup  with Trusk, Totem & SpaceFillBuilding a logistics startup  with Trusk, Totem & SpaceFill
Building a logistics startup  with Trusk, Totem & SpaceFillTheFamily
 
Building an accounting startup with Fred de la compta, Acasi & Chaintrust
Building an accounting startup with Fred de la compta, Acasi & ChaintrustBuilding an accounting startup with Fred de la compta, Acasi & Chaintrust
Building an accounting startup with Fred de la compta, Acasi & ChaintrustTheFamily
 
Scale your tech team from 0 to Series A
Scale your tech team from 0 to Series A Scale your tech team from 0 to Series A
Scale your tech team from 0 to Series A TheFamily
 
Onboarding developers and setting them up for success
Onboarding developers and setting them up for successOnboarding developers and setting them up for success
Onboarding developers and setting them up for successTheFamily
 
Apprendre à penser comme un journaliste
Apprendre à penser comme un journalisteApprendre à penser comme un journaliste
Apprendre à penser comme un journalisteTheFamily
 

Mais de TheFamily (20)

Building a design culture from day one
Building a design culture from day oneBuilding a design culture from day one
Building a design culture from day one
 
Individual Contributors vs Managers
Individual Contributors vs ManagersIndividual Contributors vs Managers
Individual Contributors vs Managers
 
Build the decentralized team you ever dreamed of
Build the decentralized team you ever dreamed ofBuild the decentralized team you ever dreamed of
Build the decentralized team you ever dreamed of
 
CEOs best practices to win time back & focus on what matters
CEOs best practices to win time back & focus on what mattersCEOs best practices to win time back & focus on what matters
CEOs best practices to win time back & focus on what matters
 
Managing fully remote teams
Managing fully remote teamsManaging fully remote teams
Managing fully remote teams
 
State of European Tech by Atomico
State of European Tech by AtomicoState of European Tech by Atomico
State of European Tech by Atomico
 
Building a real estate startup
Building a real estate startupBuilding a real estate startup
Building a real estate startup
 
A VC view on Enterprise Sales
A VC view on Enterprise SalesA VC view on Enterprise Sales
A VC view on Enterprise Sales
 
Find your style and create emotions
Find your style and create emotionsFind your style and create emotions
Find your style and create emotions
 
From product to ecosystem
From product to ecosystemFrom product to ecosystem
From product to ecosystem
 
Demystifying the product black box
Demystifying the product black boxDemystifying the product black box
Demystifying the product black box
 
The secrets to create bank brand love
The secrets to create bank brand loveThe secrets to create bank brand love
The secrets to create bank brand love
 
Building an insurance startup with Alan, Luko, Coverd & Balderton
Building an insurance startup with Alan, Luko, Coverd & BaldertonBuilding an insurance startup with Alan, Luko, Coverd & Balderton
Building an insurance startup with Alan, Luko, Coverd & Balderton
 
Mixing Product & Tech by Jean Lebrument, CTO & CPO at Brigad
Mixing Product & Tech by Jean Lebrument, CTO & CPO at BrigadMixing Product & Tech by Jean Lebrument, CTO & CPO at Brigad
Mixing Product & Tech by Jean Lebrument, CTO & CPO at Brigad
 
A new breed of CTO - Philippe Vimard, CTO & COO at Doctolib
A new breed of CTO - Philippe Vimard, CTO & COO at DoctolibA new breed of CTO - Philippe Vimard, CTO & COO at Doctolib
A new breed of CTO - Philippe Vimard, CTO & COO at Doctolib
 
Building a logistics startup  with Trusk, Totem & SpaceFill
Building a logistics startup  with Trusk, Totem & SpaceFillBuilding a logistics startup  with Trusk, Totem & SpaceFill
Building a logistics startup  with Trusk, Totem & SpaceFill
 
Building an accounting startup with Fred de la compta, Acasi & Chaintrust
Building an accounting startup with Fred de la compta, Acasi & ChaintrustBuilding an accounting startup with Fred de la compta, Acasi & Chaintrust
Building an accounting startup with Fred de la compta, Acasi & Chaintrust
 
Scale your tech team from 0 to Series A
Scale your tech team from 0 to Series A Scale your tech team from 0 to Series A
Scale your tech team from 0 to Series A
 
Onboarding developers and setting them up for success
Onboarding developers and setting them up for successOnboarding developers and setting them up for success
Onboarding developers and setting them up for success
 
Apprendre à penser comme un journaliste
Apprendre à penser comme un journalisteApprendre à penser comme un journaliste
Apprendre à penser comme un journaliste
 

Google Photos neural network learns from labeled examples to recognize images

  • 1. Machine Intelligence at Google Rob Craft, CloudIntel Lead robcraft@google.com @KindaRobCraft
  • 2. 2 Google’s mission is to organize the world’s information and make it universally accessible and useful.
  • 3. 3 one important technology we use is neural networks OUTPUTINPUT
  • 4. 4 neural net models learn from examples labeled photos “cat” “dog” “car” “apple” “flower” OUTPUT
  • 5. 5 neural net models learn from examples Make tiny adjustments to model so output is closer to label for a given image labeled photos “cat” “dog” “car” “apple” “flower” OUTPUT
  • 6. 6 after a model is trained, you can test it ? unlabeled photo
  • 7. 7 after a model is trained, you can test it unlabeled photo “cat”
  • 8. 8 Input Output “rice” “restaurants in Seoul” “hello!” “A close up of a small child holding a stuffed animal.” powerful functions that neural nets can learn 안녕하세요
  • 9. 9 Android Apps Gmail Maps Photos Speech Search Translation YouTube and many others ... Used across products: 2012 2013 2014 2015 1500 1000 500 0 Number of directories containing model description files rapidly accelerating use of deep learning at Google
  • 11. Atari 2600 testbed: 100+ Atari games from the 70/80s Agents just get raw pixels as inputs (~30K) Wired up to action buttons but NOT told what they do Goal is simply to maximize score Everything learnt from scratch, ZERO pre-programmed knowledge ONE agent with ONE set of parameters to play ALL the different games THINK OF A ROBOT in the arcade watching the screen controlling the joystick DQN Deep Q-network- Atari testbed Image: Gnome Enterprises
  • 14. 14 Street nameStreet number Street View Sign Business facade Sign Business name Traffic light Traffic sign Street number
  • 17. 17 Smart reply in Inbox by Gmail 10% of all responses sent on mobile
  • 18. 18 signal for Search ranking, out of hundreds improvement to ranking quality in 2+ years #3 #1 Search machine learning for search engines RankBrain: a deep neural network for search ranking
  • 19. 19 beyond core products, into areas like health and robotics “Deep Learning for Robots: Learning from Large-Scale Interaction”, Google Research Blog, posted March, 2016
  • 20. Sharing our tools with researchers and developers around the world repository for “machine learning” category on GitHub #1 Released in Nov. 2015
  • 21. 2121 Machine Learning Use Cases Structured Data Classification/ Regression ● Customer Churn Analysis ● Product Diagnostics ● Forecasting Recommendation ● Content Personalization ● Product X-Sells/Up-sells Anomaly Detection ● Fraud Detection ● Asset Sensor Diagnostics ● Log Metric Anomalies Unstructured Data Image Analytics ● Identify damaged shipments ● Explicit Content Classification ● Identify “styles” in images Text Analytics ● Call Center log analysis ● Language Identification ● Topic Classification ● Sentiment Analysis
  • 22. 2222 Ready to use Machine Learning models Use your own data to train models Cloud Vision API Cloud Speech API Cloud Translate API Cloud Machine Learning Develop - Model - Test Google BigQuery Stay Tuned…. Cloud Storage Cloud Datalab NEW Alpha GA BetaGA Alpha Beta GA
  • 23. 2323 beyond core products for applications like health and robotics smart platforms to accelerate progress worldwide Google products to create great experiences for users machine intelligence is already making a huge difference and there are many, many more opportunities 23