Barely a day goes by without AI being in the headlines. Yesterday, Vanity Fair did a fairly lengthy piece on Elon Musk’s view on the perils of AI (Also worth reading is AI vs MD by Siddhartha Mukerjee in New Yorker Magazine). On the other spectrum was Treasury Secretary Mnuchin not worrying about AI for another 50-100 years (Counter view from former Treasury Secretary Larry Summers). There is a lot of digest and ponder upon. That’s why I was really excited to dive into AI and are kicking off our series in Seattle last week. We will continue the series on AI in Vancouver and San Francisco later this year. More on it a bit later.

It was truly one of the most intellectually stimulating panel I have ever hosted (and I have done hundreds). It was because of the panelists and their willingness to explore the nascent topic in different dimensions. The diversity of the panelists added to the color and we were truly able to explore the topic in detail even though AI is a fairly broad topic.

The panelists were:

Sridhar Solur, SVP of Xfinity, IoT, and Data services for Comcast

Dr. Margaret Mitchell, Senior Research Scientist on AI at Google

Ken Denman, former CEO, Emotient (acquired by Apple)

Dr. Sam Browd, Neurosurgeon, University of Washington

Chetan Sharma, CEO, Chetan Sharma Consulting (moderator)

Before we begin discussing AI, we have to pay homage to brilliant men and women who have thought about machines, automation, software, and AI for many generations. The idea of artificial intelligence was embedded in the human consciousness centuries ago since the days of Aristotle and the ancient civilizations of India and China. Fast forward a few centuries and the Dartmouth Summer Research project in 1956 by John McCarthy and friends laid the foundation of the Artificial Intelligence evolution. Since then, the industry has gone through several cycles of ups and nuclear winters but finally AI has come of age and some amazing achievements and progress is ahead of us.

While there are numerous books and papers on AI, there are a couple of reports/papers that came out last year that might be useful for folks just getting up to speed on AI – 100 years of AI report from Stanford and AI, Automation, and Economy by the Obama Administration.

We have talked about AI in the context of the “Connected Intelligence” technology Wave. It is a key building block of the industry architecture.

The salient points discussed during the panel were:

• The primary reason AI is hot again is that this time it is better prepared. The enabling technologies for AI – the hardware that enables tremendous amount of processing in ms, the software algorithms that have been refined over decades have been good at using the data, and then finally the data itself is becoming plentiful. There is more data being generated every day that can be used by algorithms to better understand and more importantly train the algorithms to address a specific problem or the opportunity.
• The processing power available at the ambient edge is allowing new and interesting use cases.
• Another area of confusion is the very definition of AI with people using AI, AGI, ML, and DL interchangeably. Artificial General Intelligence is the most basic form of AI that researchers use while as Deep Learning is the technique of AI that uses data to continuously train and perfect the response like Google did with Deepmind to win against humans. AI algorithms are becoming good at detecting patterns and applying that knowledge to refine algorithms.
• While current AI is good at detecting patterns, and reporting back the probabilities, it is missing context awareness as an input. By understanding the context in which the question is being asked,
• The knowledge gained from AI is being applied to real-life scenarios: Comcast applies it for home security – using algorithms to detect loitering or attacks on the routers, discovering patterns in health at an aggregate level as well at an individual level, understanding and treating patients, understanding consumer behavior and response, tasking customer service applications to the software agents, autonomous vehicles, UAV, and much more.
• The applications are in pretty much all verticals – public safety, education, entertainment, robotics – it will indeed be a fascinating ride for the next couple of decades as we sort all of this out.
• For Comcast, the evolution of AI was straight forward. They already a machine learning team working on voice applications (TV remote) and they slowly expanded the team to deal with security and image recognition.
• The application of AI is much pronounced in health because of the complexity of the task of picking out anomalies. For example, concussion data from athletes can be analyzed and along with their past history, the diagnosis and treatment can be far more precise.
• Similarly, the measurement of pain is more art than science and is based on patient feedback rather than any scientific measurement especially amongst the kids. By reading the facial muscles, tone of voice, one can get a more precise reading of what the patient might be feeling than just asking them to grade from a scale of 1 to 10.
• Some functions like radiologists can be fully eliminated once AI gets good enough to reduce the false positives to a negligible level.
• One of the most important aspect of practicing AI is the role of the project leader or the project manager who needs to have the foresight of how some of the decision making in AI is likely to impact the users or how the responses will effect how AI responds. This was evident in the ill-fated Microsoft chatbot that leaned all the bad aspects of human behavior in a particular community (it was quite popular in China and Japan and a disaster in the US) .. but these are all learnings that help improve AI. Having foresight is more important that hindsight.
• Alexa from Amazon has done a great job of bringing voice back on the table. One of the interesting applications discussed was using “coughing data” heard by Alexa to very accurately diagnose and advise the asthma patients if they need to use the inhaler or not. Similar techniques can be applied to diabetes, sleep apnea, suicide prevention, autism, and much more.
• One of the challenges of the AI in medicine is that data is not readily available – HIPPA doesn’t allow easy access to trainable set that can be used to understand and make breakthroughs. It will remain a challenge until regulators and consumers come together to figure out ways to make this dataset available to the researchers.
• Many a times, the dataset just doesn’t represent the real world so we have to resort to compensating for the holes and it is a very delicate process for if you end up training the engine wrong, it will give spurious results back.
• There is a severe disconnect between the tech world that is going full in on AI vs. the political class that seems so oblivious to the changes that are coming. Treasury secretary’s ill-informed opinion that AI is not going to be something to worry about for the next 50-100 years is a classic example of the indifference and lack of understanding at the highest levels of the government. The last administration started to broach the subject of automation and AI late into the game but at least some opinions were striating to form on the “ground truth” which is essential before any policy work can start.
• Automation will keep marching ahead and WILL replace routine jobs and major economies are just not ready. It might hurt US and developed economies more in the short-term but countries like China and India are not immune to the impacts – it is likely to have an even more devastating impact on the economy and politics in the long-term in the emerging markets.
• Because of AI, there are some fascinating uses cases to ponder on – for e.g. what happens when there is all this free time in the car cabin – how will that change society? How will AV redefine how we have structured cities? Employment? All these will have enormous implications on how we live and work.
• There are several issues around ethics and diversity that one needs to work diligently towards.
• Gates, Musk, and Hawkings have sounded the alarm on the perils of AI, Automation, and Robots. While I perhaps have a bit more faith in human ingenuity in dealing with threats, their warnings shouldn’t be taken lightly because of profound implications.
• Privacy is a big concern in the age of AI. With all the data being produced by individuals, the opportunity for abuse is enormous. How will the industry rise to be the stewards on consumer privacy rather than just the opportunists to exploit the vulnerable.
• Obviously, the holy grail of AI is mimicking human brain to a point where you can have thought-to-text capabilities.
• While there are several areas AI helps, it is not suited for everything – especially when you don’t have much data and patterns can be deciphered by humans or simple tools like Excel and BI.
• Eventually, AI will become the underlying layer of the Connected Intelligence framework as we envisioned in our Connected Intelligence series of papers.

 

My thanks to all panelists and the attendees who made it lively. 90 minutes flew by and we had scores of questions still to address. We know it is a deep topic with broad implications so we are continuing our journey into the AI world with our Mobile Breakfast Series event in Vancouver on 4/26 and then an Evening with AI at Google Launchpad in San Francisco on 5/8.

Hope to see you then.