AI: Rewriting the Human Story Conference

Key notes and take-aways from the AI Conference

Maurice Chang
  • AI has been here before
  • What can AI do?
  • Case in point: AI in camera that has face detection (yawn) that can make a high confidence prediction on whether this person is about to cause an act of terrorism, and then trigger the appropriate actions
  • Another case: Dynamic pricing in retail using real-time weather condition data to instantly change pricing
  • Case: AI in lead generation – using AI to look at existing customer list and then analyze which new companies all over NA to call, at the personal level
  • There is a profound economic impact by AI

Deloitte’ s Chief Innovation Officer speaks out: Why AI is important to companies? Where are the opportunities? Where are the risks?

  • U of Waterloo engineering graduate – AI was there back in the 80s, its not new at all
  • We are living, now, in the 4th industrial revolution (1. steam, 2. mass production, 3. computers, 4. AI)
  • Canada is vulnerable to disruption – risk aversion, fewer high growth mature firms than US, 1 in 3 Canadian business invest below average for their industry without knowing
  • Good news is, Canada is responding. We are seeing an explosion in AI startup companies.
  • Case: IT Help Desk – IPsoft. Virtual agent with semantic, episodic memory and cognitive capabilities. Self-generates process improvements and engages human supervisor for process change approvals. Fluent in natural language generation. A Canadian bank did an assessment on its 8,000 strong contact centre and determined 3-4,000 positions can be replaced.
  • Olli – 12-15 passenger capacity, self-driving electric vehicle targeted at campuses. Powered by IBM Watson technology, overseen by a central human operator. Entire vehicle 3-d printed. Doing pilots now. Envision 17 micro-plants where Olli vehicles can be 3D printed. What does this mean for MetroLinx?
  • Also applications in cancer treatment using IBM Watson with very high degrees of accuracy in diagnosis (85%)
  • Qualcomm Tricorder XPrize – create a device under $1,000 hand held that can do 80% of what a doctor can do – you win the prize. Canadian team Cloud X. Got a device now costing $300.
  • Canada is advantaged: Godfathers of AI at UoT – Geoff Hinton (Chief Data Scientist at Google). Richard Sutton in Alberta. MILA in Montreal. These are 3 out of the Top 10 AI players in the world.
  • Vector Institute – 32 companies committed $250k for 10 years.
  • Challenges: Privacy, independence, job retention & transformation
  • Key questions:
    • How prepared is my organization to invest in AI?
    • Do I have a data strategy that will enable AI?
    • How might AI disrupt my industry, adjacent industries, and industries of partners and business?
  • Download the presentation: AgeOfDisruption.xyz Code: DEC-14

Nvidia Country Manager – What is AI, ML, DL?

  • We are in the 4th industrial revolution. Previous revolutions focused on materials and information. This one is on cognitive capabilities.
  • ImageNet was the milestone point when we reached ‘superhuman’ capabilities in vision recognition.
  • GPUs can replace traditional CPUs by an order of 1,000X.
  • More recently Google Go beat a human world champion.
  • The way it works is you feed the model labelled data and apply training algorithms. The learning model develops a multi-layer neural network.
  • The rise of GPU-enabled computing – The end of Moore’s Law – number of transistors on chip doubles every 24 months. No longer true. GPUs changed the picture.
  • Deep learning has emerged as a major disruptor. The networks are trained one layer at a time, and can be used across very noisy sets of data.
  • Who’s involved? Researchers, applied DL/data scientists, application developers
  • Applications: image classification, object detection, voice recognition, language translation, recommendation engines, sentiment analysis
  • Platforms: Tensorflow, Pytorch, Caffe2, mXnet, Theano, Baidu Paddle, Chainer, MatLab

Panel Discussion – How can companies get ready and launch AI driven projects?

  • What is the journey like? Origins overlapped a lot with applied statistics. 7-8 years ago adopted AI-first strategy, started collaborating with U of Waterloo. You need to look at AI as a strategic direction. We’ve seen the movie before. Not unlike the transistor and the image processor. AI is the same, we are at this inflection point. Cost of hardware GPUs coming down. Software is open source and freely available. Job of the leader is to translate opportunities into reality. You need to be inquisitive. Get on Coursera, Udacity. Ask the technical questions to the technical experts. Don’t think of this as a way to cut costs. The ethical human still needs to be there. This is not about getting rid of the human, it is about attaining higher levels of human capacity. Teach your team research methods, teach them how to research. Find academic collaborators.
  • The art of the possible. Brad – founder of the Electric Brain. We implement AI. Process automation. Believe we are in this first wave of AI – the narrow definition. The next wave is general purpose AI. Right now is all about narrow intelligence – topics that are unambiguous and definable. Take filing of tax returns – there are millions of people doing a very definable task. But now for F500 tax preparation – only 500 companies, not enough. We need at least 1,000 data points as a bare minimum. Everything revolves around how many examples of the data you have – the more you have the better. The other requirement is the number of relationships that you expect the AI to maintain.
  • Why should we focus on AI, right now? Well, we are already in the 4th wave of AI. Several ‘winters’. Difference now: AI is way more accessible, way cheaper. Cheaper hardware, cloud infrastructure. Volume of data and complexity of decisions is forcing us to go beyond what we can humanly do. We have to tap into AI.
  • How do we get staff to get used to AI co-workers? Well we all have some experience already – remember Cortana? You need a really diverse team to come up with the AI strategy. The IT ops people need to all get used to what’s required. Your business ops needs to be on board. Your legal hr finance teams. Cannot overstate the importance of your UX team. Finally you need someone to help your team calibrate the right level of expectations.
  • Establishing the trust factor is critical – getting humans to become accustomed to the AI is going to be hard. Traditionally trust is established because a human can explain the reasoning to us. If the AI can reason with us and explain its thinking, we can gain trust.
  • How do you start technology projects that are maybe a bit non-traditional? Imagine I am 50% of the problem. This can profoundly change the dynamics with leadership.

Leave a comment