AI and automation is all the rage nowadays - but what’s the history of these technologies, innovations and ideas?
AI and automation is all the rage nowadays - but what’s the history of these technologies, innovations and ideas? This slides will discuss the brief history of the current interesting technologies and their development to society and mankind.
1. Art of AI and Automation: primer
The history of AI technologies
2. Table of Contents
1. General History of AI and Automation
2. History of Natural Language Processing
3. History of Robotic Vision
4. Panel Discussion
3. Andrew
Liew Cool!Let’s
rock!
10,000++
2,000++
Name: Andrew Liew Weida
Title: Analytics + Innovation +
Technology
Role: Subject matter expert
Brief Profile:
Andrew is a experienced technologist with over 10 years of experience,
having cofounded 4 startups and being the early employee executives
of 2 startups, having work over 9 international cities.. He specialized in
Technology industry, particularly within the HR, SaaS, Analytics and
On Demand Marketplace segment. Andrew has worked extensively
across 8 countries and assisted numerous clients in advising how
technology can scale their operations, how to reduce risk in
implementing technology and the potential impact of tech to companies
top line.
• Startup experience – over 5 years of CXO work for tech startups & tech
enabled companies in areas of financial management, HR management, digital
product management
• Financial Analytics- creating data sourcing map to build data driven simulations
for MNCs in the banking, Private and Public service sectors for increasing
revenue , reducing cost and reducing risk management.
• Reward Design –working with Remuneration committee of MNCs to ensure
financial feasibility of compensation design to ensure market competitiveness
and internal harmony.
Education:
• PhD candidate [Econometrics]@ ANU
• Master (Economics) (Admission with
scholarship)@ ANU
• Master (Finance) (1/400, Valedictorian)@
Usyd
• BSc (2.5/4 years, fastest graduating student in
faculty history) @NUS
• AB leadership@ Korea University
• UIUX HBF
• Externalities
assessment
• Analytics
• Benchmarking
• Reward
management
• Tech based Cash
flow Modelling
• Fundraising
• Digital Product
Management
• Salesforce
Effectiveness
Key Competencies:
Relevant Work experience:
Cofounder
About the moderator
4. We will be superhuman within the next 10-20 years
14. History of Planning and Scheduling
Joseph-Louis (Giuseppe Luigi)
comte de Lagrange
Isaac Newton
Iterative Method
George Bernard Dantzig
Linear Programming
16. Five of the attendees of the 1956 Dartmouth Summer Research Project on Artificial Intelligence reunited at the July
AI@50 conference. From left: Trenchard More, John McCarthy, Marvin Minsky, Oliver Selfridge, and Ray Solomonoff.
(Photo by Joseph Mehling '69)
30. A sense of Deep Learning
Link: Google Tensorflow
31.
32.
33. What’s the caveat?
Black Box
Algorithmic model
Unknown DGP
Obtain predictive accuracy
Prediction
White Box
Econometric model
Known Mechanics
Validate hypothesis +
interpretability
Causality
Suitable for HR
applications:
reward, training,
recruiting, talent
management
Suitable for HR
applications:
benefits,
employee
records, 1st
degree of pre-
defined screening
for mature jobs
37. Dr. Vaisagh (VT)
Education
• PhD in Computer Science - NTU, Singapore
• B Eng in Computer Engineering - NTU, Singapore
Relevant Experience
• Built the impress.ai platform which is being used by top banks, telecom and consulting
companies
• Published several papers in top tier international journals and conferences on AI and
complex systems (bit.ly/vt-publications)
• Lead a research team working on city-scale traffic simulations at NRF-funded institute,
TUM CREATE while supervising 5 PhD students and leading multi-entity collaborations with
A-Star and Continental Automotive Pte Ltd.
• Completed his PhD on developing agent based models for understanding human crowds in
2014 from Nanyang Technological University
Co-founder and CTO
About me
Superpowers for recruiters
Key Competencies:
• Product Development
• Scientific Research
• Artificial Intelligence
• Software Architecture
• Simulation and Modelling
• Complex Systems
What does impress.ai do?
• Intelligent productivity enhancement tool for recruiting
• Bots screen, interview and shortlist talent in real time, at scale
• Using Artificial Intelligence, Machine Learning and NLP to augment HR
38. Natural language processing (NLP)
is concerned with the interactions
between computers and human
(natural) languages
NLP – What is it?
39. • NLP deals with a lot of core
issues of what we define as
machine intelligence
• Some applications –
• Natural Language Understanding,
Natural Language Generation,
Sentiment Analysis, Translation,
Text Classification, etc. etc.
The Turing
test
NLP – What is it used for?
40. 1954 - The Georgetown-IBM
Experiment
Statistical machine translation
Technically simple. Helped get
funds into computational
linguistics.
1950s - Descriptive Linguistics Movement
NLP through the years
1950 1950s 1954 1957 1970s 1980s 1990s
1957 - Chomsky’s Syntactic Structures
Lead to big focus on developing Universal
grammars
Away from statistical approaches
1950 - Turing test
Defined what we expect from
machine intelligence
1970s - Conceptual Ontologies and the Semantic
Web
Limited data, rules and inference systems
1980s - the rise of statistical linguistics
Machine learning based approaches gain
traction
Focus on probabilistic analysis over rules
1990s onwards
The internet, big data, personal
computingStructural Linguistics (1916 ) – prepare corpora
of nouns, verbs, phonemes etc.
Shannon Probabilistic theory of computation
(1947)
41. 1950s
There was limited data available
A lot of effort started to be put into
digitization of records
NLP through the years – Data perspective
1950 1950s 1954 1957 1970s 1980s 1990s
1960s and 70s
NLP models developed worked with limited data
1980s onwards
Availability of corpora for statistical ML
models to thrive
2000s onwards
Human computation and automatic data
annotation
Facebook, Twitter, etc.
Captcha, Amazon Mechanical
Turk
2000s
42. NLP through the years – The computational
perspective
1950 1960 1970 1980 1990
1980s onwards
Moore’s law
Personal
Computers
1990s onwards
The
internet
2000
Late 2000s onwards
GP-GPUs to speed up neural
network based processing
Today
Cloud Computing
Specialized chips for
AI
43. Natural Language Annotation – The food for
NLP
Data preparation amounts to 80% of the
time spent on a typical data analysis project
44. Natural Language Annotation – The food for
NLP
• Generating corpora for NLP research is hard
work
• Till the 2000s: Graduate student man hours
helped generate lots of corpora in universities
• Human Computation – Captcha, Amazon
Mechanical Turk
• Data-centered design – Facebook, Twitter,
impress.ai ☺
Captcha – An example of
human computation
Hashtags – An example of
data-centered design
45. NLP through the years – Tools – a personal
perspective
Tools
From using
academic libraries
like NLTK, numpy
and scipy to using
Tensorflow, spacy
and keras.
Writing complex
neural networks and
parsers in fewer
than 10 lines of
code.
Techniques
Bag of words and
expert-based
feature extraction to
word2vec
Single layered
neural networks to
Recurrent Neural
Networks and
LSTMs
Example based
approaches
Cloud
computing
Difficulty of getting
lab hours on a
powerful enough PC
to spinning up
Cloud instances
with GP-GPUs in an
instant
Deployment
techniques and tools
Containerization,
CI/CD tools, Stack
Overflow, Github
Makes it super
easy to get started
and deploy code
46. Where we are today?
Translation
• Arms race between
Technology giants to create
the best translation engines
• Excellent Wired Article last
year on the same
Voice assistants and the bot
revolution
• Siri, Google Assistant, Alexa,
Customer Service bots,
Interviewing bots
47. Concluding thoughts
• Biggest development – Availability of tools, computing and data to
the masses
• A fundamental breakthrough in NLP along the lines of what has
happened in computer vision in recent years is still missing
• The answer to this may lie in going back to the roots of NLP and
exploring paths that were not viable earlier
49. Cool!Let’s
rock!
Name: Abhishek Gupta
Role: CEO
Brief Profile:
.
• Startup experience – over 3 years of CEO work for Robotics Startups- Edgebotix and Movel AI
• Hardware experience - Designed , prototyped and tested ARM MCU based educational robots
Education:
• Masters in Embedded Systems, NTU
• Bachelors in Electronics and
Instrumentation, VIT
• Embedded System
• Project Management
• Hardware Design
• Software
Development
• Artificial Intelligence
• Fundraising
Key Competencies:
Relevant Work experience:
Founder
About the speaker 2
Abhishek is a serial entrepreneur. Abhishek co-founded MOVEL AI. Movel AI is the next generation robot
navigation software platform. The AI software uses computer vision, deep learning, and sensor fusion for
robot navigation. The technology makes robots work in many places that is not possible before: crowded
places like a hospital with a lot of human traffic, or very large space like an airport where even human can
easily get lost.
He was a leading researcher at SUTD, where he worked on self-driving bicycles and solar-powered robots.
Before starting Movel AI, he founded EdgeBotix, a hardware robotics company, where he designed and sold
hundreds of educational robots to a Singapore University.
52. History
●1950’s – Two dimensional imaging for statistical pattern recognition developed
●1960’s – Roberts begins studying 3D machine vision
●1970’s – MIT’s Artificial Intelligence Lab opens a “Machine Vision” course –
Researchers begin tackling “real world” objects and “low-level” vision tasks (i.e.
edge detection and segmentation:
●1980’s – Machine vision starts to take off in the world of research, with new
theories and concepts emerging
●1990’s – Machine vision starts becoming more common in manufacturing
environments leading to creation of machine vision industry
53. Future of Robotic Vision
●Letting robot decide based on Robotic Vision
●Collaborative learning
●Working together with Humans
54. Panel Discussion
Andrew Liew Vaisagh (VT)
Co-founder / CTO Co-founder / CEO
Abhishek Gupta
Co-founder / Analytics