Product Management Event Held at the Product Conference in Silicon Valley.
Aarthi Srinivasan, Director of Product at Target, shared her information on tech singularity. She gave an introduction to Artificial Intelligence and Blockchain, and talked about the different types of AI and blockchain. She also discussed the intersection between AI and Blockchain.
How to Effectively Monitor SD-WAN and SASE Environments with ThousandEyes
Machine Learning and Blockchain by Director of Product at Target
1. Machine Learning and Blockchain by Director
of Product at Target
www.productschool.com
2. Who said this?
2
Mark my words, AI is far more dangerous than nukes
On what he’s afraid of
“I’m really quite close to the cutting edge in AI and it scares the hell out of me. It’s capable of
vastly more than almost anyone knows. And the rate of improvement is exponential. ... We
have to figure out some way to ensure that the advent of digital superintelligence is one
which is symbiotic with humanity. I think that’s the single biggest existential crisis that we
face, and the most pressing one. ... Mark my words, AI is far more dangerous than nukes.”
5. Successful if we discuss
1. Introduction to AI
2. Types of AI
3. Introduction to Blockchain
4. Types of Blockchain
5
Goal: Provide a view on the intersection of AI & Blockchain
6. AI is not new - Why now?
6
Ref – McKinsey Co, MIT Lex Fridman, HBR,
1. Computing scale: CPU, GPU,
ASICs
1. Datasets and infrastructure to
handle big data
1. Amazon, Google, FB, MSFT
investing in platforms
8. 8
ARTIFICIAL INTELLIGENCE
The capability of a machine to imitate
intelligent human behavior
MACHINE LEARNING
1.Getting computers to learn or
recognize something without being
explicitly programmed
DEEP LEARNING
Type of ML that can process a wider
range of data resources, requires less
data preprocessing by humans
Let’s get on the same page
9. 9
Types of AI
MACHINE LEARNING
Supervised Learning
Unsupervised
Learning
Reinforcement
Learning
DEEP
LEARNING
Convolutional Neural
Network - CNN
Recurrent Neural Network -
RNN
10. 10
Types of ML: Unsupervised Learning Algorithms
DESCRIPTIVE
Look into the Past
PREDICTIVE
Understand the future
PRESCRIPTIVE
Advise on outcome
Unsupervised Algorithms Hierarchical clustering • K-Means
• Gaussian
• Hierarchical clustering
Recommender Systems
Use cases • Segment customers in
groups based on
characteristics such as
age, income, interests
• Cluster loyalty-card
customers into
microsegments
• Segment customers based on
latent preferences for
campaigns and promotions
• Segmentation based on
likelihood of employee
attrition
• Social media keyword based
clustering
• Movie, Items, News
recommendations
based on preferences
& interests
Ref – McKinse.comy, https://halobi.com/blog/descriptive-predictive-and-prescriptive-analytics-explained/
11. 11
DESCRIPTIVE AND PREDICTIVE ALGORITMS EXAMPLE USE CASES
• Linear regression • Price of a diamond by its shape, size, clarity etc.
• Understand factors contributing to product sales such as prices, advertisements..
• Optimize price points and estimate price elasticity.
• Logistic regression (Classification) • Classify diagnosis as benign or malignant
• Classify customers as will payback loan or default
• Linear quadratic / discriminant analysis • Client churn prediction
• Sale closing probability
• Decision tree (can be regression of classification
model)
• Understand salient product attributes contributing to purchase
• Provide decisioning framework for hiring / health questions
• Naïve Bayes (Classification based on probability) • Analyze sentiment to analyze product perception
• Create classifiers to filter spam
• Support Vector Machine (SVM) • Predict patients per hour
• Predict likelihood of ad clicks
• Random Forest • Predict call volume for staffing decisions
• Predict power usage in an electrical-distribution grid
• Adaboost • Detect credit card fraudulent activity
• Simple low cost image classification e.g. sat images for climate change)
• Gradient-boosting Trees • Forecast product demand and inventory levels
• Predict price of cards based on characteristics such as mileage
• Simple Neural Network • Predict a patient’s likelihood to join a healthcare program
• Predict conversion of trial users to paid users
Types of ML: Supervised Learning
Reference: https://www.mckinsey.com, https://halobi.com/blog/descriptive-predictive-and-prescriptive-analytics-explained/, wiki, Papers
12. 12
Types of ML: Reinforcement Learning
PRESCRIPTIVE USE CASES
Advise on outcome
Reinforcement Learning - algorithm receives
reinforcing rewards for its positive actions (e.g.
portfolio optimization)
• Optimization of trading strategy for options trading
• Stock and pick inventory – robotics
• Optimize real time pricing for an online auction
• Balance electricity load in grids based on demand cycles
• Optimize self-driving car behavior
• Optimize driving routes in cars
Ref – McKinse.comy, https://halobi.com/blog/descriptive-predictive-and-prescriptive-analytics-explained/
13. 13
Types of Deep Learning
USE CASES
Convolutional Neural Network
A multilayered neural network designed to extract
increasingly complex features of the data at each layer to
determine the output
Used to infer data from unstructured data e.g. images
Complex Image recognition for:
1. Medical Scans
2. Manufacturing defects
3. Website & video game image monitoring
4. Human sentiment / communication through images
5. Ariel image surveillance
Recurrent Neural Network
A multilayered neural network that can store information
in context nodes, allowing it to learn data sequences and
output a number or another sequence
Used for time series or sequences such as audio or text
Sequenced data uses such as:
1. Language translations
2. Chat bots
3. Aerial surveillance sequences with CNN
4. Narratives for reports (Narrative Sciences)
5. Communication tips & Captions
Ref – McKinse.comy, https://halobi.com/blog/descriptive-predictive-and-prescriptive-analytics-explained/
16. What is a Blockchain?
16
• A blockchain is a growing list of digital records or blocks that are secured and linked.
Each block contains
• Hash value link to the previous block
• timestamp and
• Data
• Genesis block by Satoshi Nakamoto (Bitcoin paper) – 3rd Jan 2009
“The Times 03/Jan/2009 Chancellor on brink of second bailout for banks“
• Eliminate the intermediaries
• Creates a decentralized system
17. 17
ABC’s TX
VERIFIED BLOCK
John TX
& more new Txs
Nonce
Prev block reference
/ Previous hash
Timestamp
Example Bitcoin Blocks
New TX 1
NEW BLOCK
New TX 2
& more new Txs
Nonce
Prev block reference
/ Previous hash
Timestamp
…...
18. Types of Blockchain
18
PERMISSIONLESS
PERMISSIONED
• Anyone can participate and validate a block
• Common validation method proof of work
• Restricted actors can validate a block
• Various methods of consensus are used e.g.
Byzanthine fault tolerance
19. 19
Blockchain Technology Platforms
BITCOIN ETHEREUM HYPERLEDGER R3 CORDA
Verification Proof of work
Data format:
Merkle tree (20
txs per sec)
Proof of work
Data format: Patricia
tree
Consensus based –
Modular & Extensible
Consensus with
Financial sector as
focus area
Permission State Permissionless
with basic
contracts
Permissionless with
smart contracts (e.g.
solidity)
Permissioned with
smart contracts (e.g.
Golang, Java)
Permissioned with
Smart contracts
(Kotlin, Java)
Smart Legal prose
Distributed Distributed
system with all
accounts equal
access
Distributed system
with all accounts
equal access
Distributed system
with role-based
restricted access
Microledgers semi-
distributed systems
with restricted access
Block creators External account External Account,
Contract account
Multiple roles such as
Validator or Transactor
Multiple roles
including Notary
Cryptocurrency Bitcoin currency Ether or other tokens
via smart contract
No currency (chaincode
tokens if required)
No currency
Ref – McKinse.comy, https://halobi.com/blog/descriptive-predictive-and-prescriptive-analytics-explained/
20. AI $37B market by 2025
20
* - 2012 – 2017 ; Ref: Venture beat,
$15 B* AI investments with $15 Trillion impact on GDP by 2030
Images: Intershala
Start ups
~$8B (2012-2016)
Auto Tech
Core AI
(Training data)
Healthcare
2 31
Big Corporations
~$6B (2013 - 2016)
Voice is the
new Text
AI Platform
Cloud
1
2
Vision & Image
recognition
3
Ford invested $1B in Argo self-driving AI tech
21. Blockchain Future
21
Identity, Hardware &
Platforms
• Identify Platforms
• Blockchain
standardization
Crypto devices & Apps (Crowd data)
• Crypto wallets, phones / ipads
• Open source apps for
crowdsourcing
• All your phone apps using
Blockchain such as Airbnb,
Facebook, Searches, Twitter type
applications in Blockchain with
token rewards.
AI application with smart contracts
• Finance protection contracts
• Document fraud detection
• iOT safety & security (Pollution
monitoring, Museum art)
• Medical AI contracts
• Autonomous cars
• Decentralized Organizations
(Common Education Standards, Bot
shares, Refugee crisis, weaponry)
5 10
Company Token Description
Singularity Net AGI Connect Siloed AI algorithms & decentralize (Ben Goertzal OpenCog)
Effects.ai Mar 24 - EFX 1st : Mechanical Turk, 2nd :AI Marketplace, 3rd: Compute share
Medrec Private Blockchain authenticated by medical researchers to store medical records
Loomia - Clothing panel tile that does Lighting, Heating, Sensing data to collect you money
22. 22
We will achieve technology singularity with ethics
Ref – http://blog.crisman.com image with edits, Business times, Guardian
The "Kuratas" robot in Tokyo, Nov. 2012. The military
robot can be controlled by a pilot or via a smartphone. It is
armed with a futuristic weapons system, including a multi-
rocket launcher.
Elon Musk & Deepmind’s Mustafa Suleyman leading a
group of 116 specialists from across 26 countries who are
calling for the ban on autonomous weapons. - 2014
27. Bitcoin protocol steps
Aarthi Srinivasan
27
Start: Broadcast new
transaction
Verification: User
Signature & funds
Proof-of-work:
Prevent double
spending
Mining: Earn
bitcoin rewards
Recheck
transactions &
start new block
1
2
34
5
Search Engine Account e.g. ABC wants users to use its search engine and will pay them 1 crypto unit(or some fraction of a bitcoin) as a
reward in return for using the search engine.
28. Key Terms
28
• A set of data used to predict relationships. Data and answers for each
sample.
• E.g. A diamond’s size, cut, color and clarity helps predicts the price.
Training Set
• Uses training set to make a prediction.
• E.g. Model predicts diamond prices based on past prices.Supervised Learning
• Provide data without suggesting anything so computer can identify patterns
or groupings.
• E.g. Customer segmentation, DNA groupings.
Unsupervised Learning
• Each distinct measurable data value you select in the training data set.
• E.g. A diamonds’ size is one of the feature’s for predicting price.
Features/ Variables /
Attributes
• Using the features provided in the training set make a prediction. Fit a curve
using the data provided.
• E.g. Price of diamond = X*Cut + Y*Clarity + Z*Size + other features…
Supervised: Regression
• A defined set of categories that can be labeled for placing new observations.
• E.g. Presence of absence of cancer; Types of diabetesSupervised: Classification
• Process of assigning observations into subsets.
• E.g. Customer segment creationsUnsupervised: Clustering
30. Voice will be the new Text
30
Google
• 11+ acquisitions
• ML Platform creation
• Vision / Image and Speech
recognition
• Business Process improvements
Apple
• 7+ acquisitions
• Vision / Image and Speech
recognition
• Catch up with Google on
platform creation (Turi)
Facebook
• Vision / Image and Speech
recognition
• Voice activation SDKs
Microsoft
• Voice enabled assistant
• Type ahead predictor
• Voice activation SDKs (AI
Fund)
31. 31
These companies market cap surpass the GDP of
India (previously Russia and Canada)
Reference – Scott Gallowa
32. Part-time Product Management Courses in
San Francisco, Silicon Valley, Los Angeles, New
York, Austin, Boston, Seattle, Chicago, Denver,
London, Toronto
www.productschool.com