AMII Conference

AMII

Randy:

• Where does cutting edge AI come from?
  • Carnegie Mellon University
• Federal AI Strategy Development
• Ranking of AI and ML: CMU, Tsinghua, University of Alberta, Cornell University, Technion, MIT, UT Austin, JKUST, Berkeley, UMass, UMich, UCLA, Stanford, etc;
  • China AI Strategy has moved Tsinghua from 10^thto 2^nd
• AMii focuses on efforts of 4 areas
• Pan-Canadian AI Strategy
  • $125 million across Canada to attract and retain academic talent
  • Recognizes Canada’s centers of AI excellence: Mila (Montreal), The Vector Institute (Toronto) and Amii (Edmonton)
  • First cohort of Canada CIFAR AI Chairs includes 29 researchers at institutions across Canada
• Amii has diverse expertise in machine intelligence
  • Algorithmic game theory
  • Algorithms and theory
  • Bio/medical informatics
  • Dana mining & analysis, etc;
• Machine Intelligence Achievements
  • Pioneer of reinforcement learning
  • First group to beat poker pros at heads-up no-limit texas hold’em
  • Solved game of checkers
  • Academic origins of AlphaGo and the Atari Game Project
  • Developed UCT algorithm at the heart of many advancements in games
  • Thailand National Innovation Award for Tuberculosis Diagnosis
  • System capable of passing Japanese Bar exam
  • Open-source community centered around adaptive prosthetic limbs
• Amii explores through world-class research and training
  • Amii Explores: pushes the bounds of scientific knowledge by enabling world-class research and training and facilitating knowledge, technology and talent transfer from academia to industry
  • Amii Educates to boost intelligence literacy: courses, workshops & seminars
  • Amii provides a range of meetups, events and educational opportunities to help drive machine intelligence literacy for Alberta workers and businesses—build, reskill, upskill, and retain machine intelligence expertise in Alberta
    • Management level courses
    • Technical level certification
    • Bespoke curriculum development
  • Amii Innovates: guide businesses on path to machine intelligence adoption by helping them develop strategies, shift processes, and systems and build in-house knowledge and teams
• Connect with amii:
  • [email protected]
  • amii.ca
• Q&A:
  • Will we catch up to China? No
  • China, US, Israel, & Saudi Arabia lead the world on collecting data/information
• Model-free RL Atari Game Learning
  • Google’s DeepMind Deep Q-learning learns how to play the game after watching the screen
• Big Data Challenges
  • Square kilometer array
  • DNA transistor
  • Environmental monitoring fluxnet
  • How are you able to determine what to keep / throw away
• NCAA Football Clusters
  • Challenge: figure out division in NCAA
  • Run a clustering algorithm on the schedule
  • Model visualization changes based on what information you are trying to extract
• Diagnosing ADHD
  • Machine learning classifier uses a participant’s resting state fMRI scan to diagnose the individual into one of 3 categories: healthy control, ADHD combined type, ADHD inattentive type
  • Used participant’s personal characteristic data: collection age, gender, handedness, performance IQ, verbal IQ, and full scare IQ, FMRI data
• Image-based glioma biopsy
  • Compute S-transform frequency spectrum for each pixel, the average over all directions to get 1D spectra for each tumor
  • Build multi-frequency classifier based on these labeled cases
  • In these measurement results, BLUE 30 co-deleted 1p/19q, GREEN 24 intact 1p/19q
  • Results: 95-96%
• Optimizing water treatment control
• Use machine learning to find separators in the data
• Natural language & approximate semantics: use the amount of google searchers also in google translate
• COLIEE Statute Law Challenge
  • Statute: if a person in order to allow a principal to escape imminent danger to the principal’s person, reputation or property, the Manager shall not be liable to compensate for damages…
  • Question: in cases where an individual rescues another person from getting hit by a car by…
• Query case àcase law judgements àhow to pick the ones to be noticed and the lawyer will notice in the count
• Why we need explanatory AI systems
• Funny car cartoon on automation: “Does your car have any idea why my car pulled it over?”
• MYCIN expert knowledge
• Rich Sutton, weak versus strong methods
  • Methods that scale with computation are the future of AI
• Learning a model to classify images
  • classifier to distinguish dogs and cats
• Experiments
  • Word deletion experiments
  • Green words = positive
  • Building a model of language to explain
• Summary
  • World leading research group
  • Highlights of game playing, reinforcement learning, NLP, XAI
  • Expanding capacity, building industrial collaboration function
  • Industrial projects with DeepMind, DeepMind Health, Google Brain, Mitsubishi, Borealis AI, Alberta Treasury Branch, VW Group
  • Eager to build sustainable industrial relationships to continually refresh “line of sight” to potential high impact applications
• Q & A:
  • Generating automatic models instead of analytically building them
  • If you get 2 different explanations and they are contradictions, then what is the model you have to create to compromise the two

 

Edmonton’sArtificial Intelligence Ecosystem: Research & Business:

• Path of how businesses are created revolving AI in Edmonton
• AI is a transformative technology
• AMII:
  • U of A’s chinook challenges
  • IBM’s deep blue defeats Kasparov
  • Reinforcement Learning Author at U of A (Sutton)
  • DeepMind’s AlphaGo beats Sedol
  • Deep Stack Beats Texas Hold’em (Bowling)
  • DeepMind Opens Lab in Edmonton
  • AI Can Pass Japanese Bar Exam
  • csranks.orgàselect AI AND ML
• Government commits $100 million to AI companies
• Albert’s AI/ML Business Strategy Plan: Research & Development, Talent, Infrastructure, Industry Solutions, Market Presence
• Companies that focus on AI in Edmonton: scope AR, AltaML, testfire labs, serious: labs,zept, ATB financial, Promethean Labs, oneBridge, Telus, Jobber, Servus credit union, PCL, ZEPT, Vadu, NTWIST, Drugbank, frettable ,SAM, Health Gauge, Darkhorse Analytics
• Four Key components to build a successful AI/ML company:
  • Technical talent & IP, product development & business skills, industry domain expertise, data = successful AI/ML companies
  • Alberta àarchitecture, engineering, & construction, health, financial services & insurance, energy & utilities, transportation & logistics
• CEO of Japanese company needs to shut down a nuclear plant àsending in robots to disarm it
• Trying to create an AI industry in the Edmonton area
• A lot of those who get successful at AI have a HUGE amount of data
• Why Edmonton?
  • Great place to live
  • Vibrant quality of life àlargest snowball life in the world
  • City of Entrepreneurs
  • Economic leadership
  • Growing innocation community
  • City best place in Canada for youth to work
  • Tops in housing affordability
  • Top ranked research and educational institutions
• Bruce Alton: [email protected]
• Daylin Breen: [email protected]