Patrick Schwerdtfeger is the author of “Anarchy, Inc.: Profiting in a Decentralized World with Artificial Intelligence and Blockchain” and a regular speaker for Bloomberg TV. He’s a leading authority on technology trends including big data, artificial intelligence and machine learning. After a number of years of limited progress, artificial intelligence has seen astonishing developments in recent years and machine learning is part of that process. Patrick covered “Monetizing Big Data” and “Will Artificial Intelligence Destroy Us?” at the 2016 NASSCOM conference in Hyderabad, India, as well as dozens of other events. He described the difference between narrow or weak AI and how it differs from general purpose AI, with AlphaGo and the recent recreation of the Bose-Einstein Condensate (BEC) by an AI as examples. These platforms are “general purpose” because they do not start with a set of rules. Instead, they are given access to data and are given an objective to achieve, that’s all. They can then process the data as they wish in an effort to achieve the objective. In all of these cases, the developers of the platform were stunned at the path chosen by the AI. It turns out that machines “learn” very differently than humans, and Patrick can present that reality at your next event.
Past speaking clients include:
Recent speaking destinations include:
The five (5) emerging use cases for artificial intelligence and machine learning include:
- Sensors and the IoT (Digital Transformation)
- Chatbots and Socialbots replacing Call Centers
- Chatbots and Socialbots as Companion Robots
- Facial, Object, and Image Recognition (Security Applications)
- Autonomous Vehicles (Cities, Freeways, Farms)
- Automated Retail Checkout (Amazon Go Stores)
The State of Artificial Intelligence (AI)
The recent years have seen tremendous progress in the area of deep learning and neural networks. Various AIs have surpassed expectations and demonstrated the potential of machine learning technologies. Examples include AlphaGo (via Deep Mind, acquired by Google in 2014) and an AI designed to recreate the Bose-Einstein Condensate (BEC) in less than an hour, developed by Australian researchers in 2016. And of course, there are countless other applications (by IBM Watson and Vicarious among others) that have not made the headlines but have dramatically improved the performance of everyday tasks (like spam filtering and image recognition). Some have referred to these recent developments as an AI Spring, following an AI Winter which saw little progress over recent years.
The strength of these AIs grows with the additional layers of analysis, adding context to simpler versions. Each additional layer dramatically improves the results, allowing the overall capabilities to compound on top of each other. The inevitable result will be an increasingly complex and integrated digital infrastructure with modules interpreting different aspects but then contributing to a larger analysis. The requirement to host such an infrastructure is enormous computing power. In 2016, the two fastest supercomputers in the world were both located in China. They are investing heavily in the enabling technologies behind artificial intelligence. Google is also investing heavily. Naturally, we can expect significant innovations coming from those two places.
The possible dangers of artificial intelligence emerge when there is one single monolithic AI that is far more advanced than any other platform. If such an AI were allowed to develop in isolation (without any material competition), later AIs would never be able to catch up. In fact, their relative capabilities would only grow further apart. Once a truly “general purpose” machine learning infrastructure is in place, it could easily be replicated, allowing the core capabilities to grow exponentially. Particular areas of future development include virtual and augmented reality interfaces.
Patrick builds his keynote programs by accumulating and studying use cases and success stories. In new fields like Artificial Intelligence, the best way to do that is to follow the companies on the forefront of innovation. Of course, machine learning emerged from the exploding fields of big data and the internet of things. Some of the leading AI and Machine Learning innovators include:
- Perceptio: Advanced AI for smartphones (acquired by Apple)
- Emotient: Emotion-detection technology to improve understanding of customer sentiment (acquired by Apple)
- Converto: Marketing Intelligence (acquired by AOL)
- Dark Blue Labs: Deep learning-based technology for understanding natural language (acquired by Google)
- DeepMind: Solve intelligence. Use it to make the world a better place (acquired by Google)
- DNNresearch: Use of deep learning and neural networks for image search (acquired by Google)
- Indisys: Natural-language processing (acquired by Intel)
- IQ Engines: Image Recognition Software (acquired by Yahoo!)
- SkyPhrase: Natural-language processing technology (acquired by Yahoo!)
- Madbits: Deep learning-based visual intelligence platform to identify contents of images (acquired by Twitter)
- TellApart: Predictive advertising for e-commerce and retail (acquired by Twitter)
- Explorys: Predictive healthcare data analytics (acquired by IBM)
- AlchemyAPI: Cloud platform with natural-language capabilities including keyword extraction and categorization (acquired by IBM)
- Cogenea: AI-based virtual assistant (acquired by IBM)
- MetaMind: AI-based personalization and customer support solutions for companies (acquired by Salesforce)
- PredictionIO: Open-source machine learning server (acquired by Salesforce)
- Arria: Natural Language Generation
- SkyTree: Enterprise-Grade Machine Learning for Big Data
- Botanic: Building Humane Machines
- Banjo: Discover every event around the world as it happens.
- Infer: Predictive Sales and Marketing
- MindMeld: Advanced AI to Power Conversational Interfaces
- Automated Insights: Natural Language Generation
Patrick’s Machine Learning Keynote Speech
Patrick regularly customizes his keynote program to best complement the event objectives of his clients. As such, his standard artificial intelligence and machine learning keynote is only a basic framework. As a matter or process, he builds his keynotes in modules, where each module is between 5 and 8 minutes long. The opening module discusses the concept of leverage and how technology offers leverage to those who capitalize on emerging capabilities. After that, Patrick provides a strategic overview of the topic, covering recent developments in neural networks and deep learning as well as implications for business executives and investors.
The keynote then moves into a series of case histories and success stories. These use cases are incredibly important which explains why the Harvard Business School uses the case history method in their MBA program. Patrick has accumulated dozens of case histories and examples that are both entertaining to hear and insightful to analyze. Depending on your event objectives, he can select the use cases that are most appropriate for the occasion. Finally, after demonstrating the breadth of the field, Patrick usually closes with his signature ‘think bigger’ message, encouraging attendees to expand their thinking and consider possibilities that might seem implausible at first.
Patrick Schwerdtfeger recently held two virtual events for SAP to review emerging use cases for artificial intelligence and machine learning. At the end of that session, participants were invited to submit questions. We have included those questions as well as his answers below:
Interested to hear how jobs will really change with adoption of ML and AI
Most process-oriented tasks will be automated. The remaining jobs will involve complex human relationships, empathic communication, creativity, and agility.
What skills are going to compliment a future under AI and ML?
Valued skills will fall into two buckets. First people with STEM educations will be valuable in the development of further technologies. Second, skills including empathic communication, emotional intelligence, creativity, and agility will be necessary to bridge the gap between machines and humans.
What data will Machines get their information from and how will they determine what behavior is right and wrong? Examples are spelling mistakes or ineffective Leadership…..
This is a problem already today. The biases reflected in the data are creating biases in the AI / ML platforms. We still need to develop a system to solve the problem. But who will have the authority to determine “good” and “bad” behavior? This will be debated for years to come.
what is an agile human?
I just used the “agile human” label to refer to people with physical dexterity and agility. Manual tasks that cannot be automated will require capabilities that machines cannot achieve.
Job displacement is terrible for the displaced employees and good for the companies in terms of cost savings, loyalty etc, this seem apparently ok but, if up to 40% of people is going to be unemployed who will have the money to buy services or products from the companies who are firing people to “save” some money? this is very very concerning, how is SAP preparing the enterprises for that?
Humans are creative, and humans WANT to work. Those whose jobs are displaced by automation will inevitably find something to do. They may struggle and earn far less than they did in previous jobs, but they will end up doing something to earn money along the way. Many people trumpet “universal basic income” (UBI) as a solution, but that will fail over the long term. Regions that adopt UBI will be at a competitive disadvantage over those who don’t. On the other hand, technology is demonetizing countless industries, so less income in the future may be sufficient to maintain a sustainable lifestyle.
For which use cases do we have the most compelling business cases and mature solutions.
Recommendation engines are already well established. The other four are emerging quickly right now. They include (1) facial, image, and object recognition, (2) natural language processing, (3) automated retail checkout, and (4) autonomous vehicles.
How long do you think it will take before AR is commonplace? Will it follow historical adoption timelines of mobile phones or be quicker?
Right now, we’re just waiting for some VR / AR headset to gain broad market adoption. So far, nobody has cracked the code, but it’s inevitable that someone eventually will. As soon as a headset gains broad market adoption, propagation will go quickly, although I’m not sure if it will be as quick as mobile phones or even quicker.
How will security officers be replaced? If someone comes in that isn’t supposed to be there, how will they be removed?
Security guards will probably never go away entirely, but imagine all the TSA employees at airport security checkpoints. The vast majority of those employees can be replaced with new security tunnels (and other technologies). The number of guards remaining will only be there to handle unusual situations.
I have a knowledge about ML being a branch of Artificial Intelligence theoretically. However, can you explain in practical way, what is the “tine line” , where Machine Learning ends and Artificial Intelligence starts? Why Digital Twin is more associated to AI rather then ML? Machine learning is one possible technique to create artificial intelligence. Artificial intelligence is a thing. Machine learning is a process.
How can early adopters identify the right technology to invest in? If there are competitive offers, and one is going to be adopted as standard in the future, there is a risk to invest an hige amount of money in a technology that will be superseded in the long run.
This is very true. Inevitably, “innovation” requires budgeting failure. You have to try new things, which implies that you don’t know if it will work or not. Every time an organization adopts a new technology, they’re placing a bet as to its future utility. The best thing to do is sort market competitors by gross profit, and then follow the technologies implemented by the highest gross profit players.
Would love to hear your thoughts on timeline for singularity and ethical aspect of AI.
Ray Kurzeil expects the singularity to occur by about 2045. I don’t see the sigularity the same way he does, but certainly, the computational power of computers will supersede humans before then. I’m not sure machines and humans will ever actually “merge” together. Instead, I worry about a single monolithic AI that begins operating independently from humans. It’s impossible to anticipate how that will play out.
Where will we partner vs. compete in these new spaces? In order to stay relevant (and employable) in an AI future, humans need to partner with technology as much as possible. As soon as we’re competiting, we invite problems. The problems will begin when machines adopt objectives that contradict our own, but I do not know how or where that will play out.
What is the person less Chines Bank called?
State-owned Beijing-based China Construction Bank (CCB)
How SAP can support customers to transform their employes jobs?
There is a significant opportunity for SAP (or a variety of other large enterprise companies) to help organizations maintain a healthy corporate culture amidst technology-driven layoffs. They can also help organizations train their remaining employees to work WITH technology rather than AGAINST technology.
How does the Amazon Go stores distinguish between who of the customers have taken what from the shelves?… so that they know, who to charge
The stores have cameras everywhere, allowing them to track individual shoppers inside. Also, each shopper has a phone in his/her pocket, providing additional GPS data to the store. Even the shelves have cameras, so the system knows who is in front of the shelf when an item is picked up. And finally, these systems are learning and improving along the way, possibly incorporating other variables in the future.
What SAP APIs or full-function routines exist as solutions for clients to consume quickly and easily versus a longer programmatic approach to deliver AI and ML? It would seem that quickly consumable utilities would lower the barrier to adoption.
I’m not familiar with SAP’s APIs and full-function routines, but your point is absolutely correct. The UI for new technologies should be as simple and intuitive as possible, accelerating adoption.
Regarding Autonomous }driving: Is there a way to minimize the accidents. or Will it take more time for it to happen since it is still a fresh thing, due to the accidents it may cause?
Already today, autonomous driving results in far fewer accidents than human drivers. The problem is that the media talks about EVERY accident caused by autonomous vehicles and barely mentions that hundreds and thousands of accidents happening all the time with regular vehicles.
If passengers face each other, how do airbags work?
This is a good question, and I don’t know the answer. But I’m quite sure the designers are addressing those concerns.
It seems as if Amazon, Google and Microsoft are quite far ahead with the technologies. Does SAP offer something similar or what do we do to stay competitive in this area?
It’s true that Amazon, Google, Facebook, and Microsoft are moving quickly. Although I am not an employee of SAP and am unfamiliar with your full menu of products, I believe the most valuable angles SAP has are (1) existing data and (2) access to enterprise customers. SAP should partner with some of these other companies to leverage available assets and contribute to future solutions.
As self-driving cars are deployed more widely, who should be liable when accidents happen? Should it be the company that made the car, the engineer who made a mistake in the code, the operator who should’ve been watching? This is a fascinating area, and many people are debating precisely these issues. Inevitably, it will depend on the courts. Once a few of these cases get prosecuted, the courts will decide where the blame should be allocated, and these decisions will become the precedent for future cases.
Can you comment on the usage of AI from international investment banks? Is it already a reality? Definitely. Wall Street was the first sector to hire AI and ML developers en masse (back in 2000 to 2005). Those developers initially programmed “black box” high frequency trading platforms, but I guarantee that they’re now feeding live trading data into ML platforms to identify profitable trading opportunities. Check out the AIEQ and AIIQ ETFs for early examples of similar technologies for regular investors.
Didn’t Facebook shut down an AI program when they learned that it was creating its own language?
Yes, this did happen. The media distorts these sorts of stories. Yes, it happened. Yes, it was disturbing to say the least. But FB obviously learned from that experience, and it will inevitably result in better development in the future. But yes, it happened, and I expect many more similarly creepy things will happen in the years to come. Of course, if the dystopian AI outcomes materialize, the progressions will be littered with these types of stories, before the machines finally take over entirely.
What technologies do you think will contribute to automated warehouses? when?
Large warehouses (Amazon distributions centers, for example) are highly automated already. Shelving units are navigated around the facilities, and pick-pack robotics are being developed right now. There are also other concepts being developed for sorting products in high-volume warehouses, so innovation continues in that area.
what would be the impact on human behavior with the blur between human interaction and machine interaction
Many people will learn to HATE machines. Already today, autonomous vehicles are being keyed and vandalized much more than most vehicles. People will soon learn how autonomous vehicles, for example, react to certain situations. It’s likely that people will intentionally create those situations, instigating those reactions. I wouldn’t be surprised if augmented reality games were created, allowing people to actually win points by screwing with the drone cars. It will get ugly, that’s for sure.
What do you believe to be the next big emerging technology? How do you se AI, ML and big data technologies evolving?
AI will be added to everything. Ubiquitous Internet access will be followed by ubiquitous processing power, probably accessed through any electrical outlet. Processing power will increasingly migrate to “the edge”, closer to individual endusers. VR / AR / MR is coming soon. As soon as a headset gains broad market adoption, B2C companies will be in a mad scramble to create “experiences” for customers, all available through their websites. People will begin voluntarily abandoning the “real” world in favor of a much more exciting “virtual” world. Over time, it will be impossible for the real world to compete with the virtual world.
You gave a good Retail example of how AI impacts loss of jobs but may intro more locations which benefits more people.
I’m not sure if I understand your question. It is true that reduced salaries lowers the breakeven for new retail locations. Apple Stores, for example, are being opened in increasingly low-density communities, driven by efficiencies in their operating procedures.
We are clear that AIs will need to be regulated with ethic rules, but do you think that AIs will also need some form of fundamental rights like the right of free speach ? ANd if so, what would These Rights look like and how would they be different from fundamental human Rights ?
Last year, Arizona attempted to pass legislation allowing their citizens to pay taxes in Bitcoin. Although the legislation stumbled, it demonstrated that one state – Arizona in this case – was willing to be a leader in the development of new legislation. Inevitably, this will play out for AI legislation too. Right now, Estonia is a leader in digital legislation, but there are others. We will have to wait and see what these different jurisdictions include in their laws.
what about Healh Care / Health Mgt? I guess this will be a huge area for AI, focusing on diagnostics, on serving people in their homes instead of going to hospitals, remote services and so on
Yes, healthcare is poised to change quickly over the next few years. Watch non-healthcare providers like Walmart for developments in this area. There are also a lot of VC-funded startups addressing this area. It’s safe to say that we’ll migrate to an on-demand clinic-based model in the future, with a variety of very different a la carte “insurance” options.
Can we use this content in our customer presentations?
I purchase a “use license” for most of the images in my presentations. As such, I can’t provide the images in full resolution, but you could presumably recreate them using slightly different images (or even the same ones with a more inclusive use license). Let me know if you’d like my help when presenting to customers.
How can I get more information on Fleet Learning?
If you do a Google search on this keyword phrase, most of the listings refer to Tesla’s experience with it. You can also search for the “network effect” to get closely related discussions on the topic.
Do Androids dream of electric sheep?
Lol. You might be surprised at the emerging use cases in prostitution circles. Search for “lumidolls” if you’re curious.
what happend with 3er world countries. There are millions of people outside of digitalization. That do not have credit cards or bank accounts. how are they going to jump into this technology?
Blockchain is addressing this reality head-on. In many ways, blockchain is the “anti-business-model”. It’s essentially a public utility, and it is developing quickly in countries like the Philippines. Also, the percentage of NOT connected people is shrinking quickly. If Facebook and/or Google deploy their ISP services (solar powered planes by FB and balloons by Google), Internet connectivity will cover almost everyone in short order.
While I’ve heard about a number of job cuts, I’ve also heard that new opportunities will result from AI. Will the jobs figures you shared be balanced out by other emerging work?
Yes, definitely. There will be countless new jobs in the future; jobs we can’t even imagine today. VR / AR / MR developers are a case in point. Once a headset gets broad market adoption, B2C companies will be in a mad scramble to develop “experiences” for customers, and the demand for developers will explode. And there will be many others. Having said that, however, it is unlikely that the new jobs will emerge as fast as old jobs disappear, so there will be some ‘absorption’ problems during the transition.
How do you think AI can help solve real world problems like poverty, unemployment, terrorism etc. Looks like AI would aggravate some of these problem.
I agree that AI will both help and hurt these causes, especially when the “2nd wave” of users begin deploying new applications. However, I also believe AI will do more good than harm. We are optimizing our planet; company by company, region by region, country by country. It’s an exciting time and I’m absolutely optimistic for the problem-solving capacity of our planet in the years to come.
We are working in this area to continue to learn how jobs will change industry by industry…would it be possible to talk about this further beyond the Q&A – we just held a session at MIT on Future of Work two weeks ago – it would be wonderful to speak with you as well.
Thanks for reaching out, Kerry. I enjoyed our conversation on Monday would love to work with SAP in this area. The first step would be to flesh out the problem and proposed solutions. After that, SAP should develop some sort of program to monetize the growing corporate fear in this area.
How AI is going to handle “General Data Protection Regulation” rules ?
The impact will probably be in the other direction. In other words, AI will not affect GDPR, but GDPR will affect AI, primarily in the availability of data to “learn” from. The most important ingredient for ML is data, and GDPR makes it harder to access and use customer data.
Do you think there will be a time were Neural networks and AI will exhaust as the data feeding process is not continuous. This is an excellent question.
Yes, I do think that AI benefits can be exhausted in certain areas. Once we’ve learned from the data, we don’t need to learn from it again. The job is done. That’s an over-simplification, but data is the necessary ingredient for ML, so if the data runs out, the ML will have nothing to chew on.
What is the single best way to control AI that you have been exposed to? Is it truly able to be controlled?
I am not aware of any successful attempts to “control” AI, aside from unplugging the machine. Perhaps that’s the only viable stop-gap. Remove the power source. Organizations like Open AI (co-founded by Elon Musk) are also looking at stop-gap measures to limit abuse as well as other negative outcomes. Their approach (in conjunction with Neuralink) is to make it easier for humans to “participate” in the AI directly. It will be interesting to see what they come up with.
AI running wild (second wave,etc.) is not a technology conversation, it is more a sociological. We saw so many good things turning bad. Can be AI used for avioding AI running wild? (do you understand what I mean?)
Yes, I understand. And yes, just like computer viruses and anti-virus software battle each other every day, so too will good AI and bad AI battle each other in the future. In fact, I suspect some fascinating clashes could emerge when we get to that stage.
Are creative and crafts-based jobs safer? If so, why just promote STEM education and not the arts?
I agree with you 100%. There are two categories of “safe” jobs in the future. The first involves STEM fields, as those areas are necessary to build out these future technologies. But the second involves (1) complex human relationships, (2) empathic communication, (3) creativity (including the arts), and (4) agility and dexterity.
After AI what’s left for humans to figure out? Maybe just how the brain works, which will probably be done with AI anyway. Thoughts?
Mapping the brain and neural activity is definitely an area of focus these days, and it will take a long time before we understand it fully. Also, the exploration of space and the origins of the universe will remain a major area of interest. Quantum computing has the potential of explaining thus-far-unexplainable phenomena as well. Anyway, yes, I believe AI will play a role in all of those areas.
Being part of the support team in SAP, what direct should we be working towards, in order to be safe from these changes.
My best advice is to run towards the change, not away from it. Learn everything you can about AI, not only within SAP solutions, but also from other providers. Google Tensorflow is a good place to start. If you know more than the next guy/gal, you’ll be much safer from potential negative outcomes.
What are your thoughts about Psychology and AI?
We are already seeing fascinating applications of AI to understand facial experssions and “micro-expressions” as a proxy for mental health, anxiety, happiness, and relationship potential. In one case, the analysis of micro-expressions was over 90% accurate when predicting relationship success. With that type of success rate, it’s inevitable that people (including me, by the way) will be interested in using these analyses in their own relationships. It should be noted that this case was NOT done using AI or ML, but it’s likely that that will change in the future.
What’s a threat to B2B sales job? Will those jobs get affected ?
Yes, they will be impacted in two ways. Once NLP passes the Turing Test, socialbot technology will become ubiquitous, and companies will split test and optimize all sorts of conversations, including B2B sales conversations. But this is also a generational question. Younger workers don’t want to speak with human sales people. Older workers feel the opposite way. So as the older workers retire, and younger workers dominate corporate roles, sales people will become less desirable in the sales process.
AI will create new jobs do you have some examples of new jobs that will appear in the next years?
I don’t think I want to place bets on this one. It’s really hard to anticipate the specific jobs that will emerge. Having said that, read the answer in row #46 above for a few relevant thoughts.
Are you familiar with the so called four creative super powers we need to be successful in the future as a human: Hacking, Making, Teaching, Thieving, and what is your take on the necessity of these skills?
I had not heard of these four super powers before, so thanks for including them here. I will dig into it. For now, please read the answer in row #3 above for a few initial thoughts.
Do you think AI will create even more income disparity?
Yes, absolutely. Technology is a form of leverage. You’re either on the right side of that equation (leveraging the technology) or on the wrong side (being leveraged BY the technology). Also, the falling birthrate mathematically guarantees a widening disparity between rich and poor (read the book “Capital in the 21st Century” for more on that). So yes, for multiple reasons, the division between rich and poor will continue to widen in the years ahead.
The late Dr Sarno figured out that 80% of ALL medical conditions are actually caused by our psyche, NOT the physical cause that medical science currently attributes them too. Will AI be able to do psychology and psychiatry too?
The placebo effect is incredibly powerful, both in “curing” medical conditions and also “causing” those same conditions.
How will new innovations in AI and ML be challenged or helped by laws, lawsuits, and regulations? And, how will those changes lead to adoption by large corporations?
Lawsuits will definitely impede progress when financial liability is established, but we live in a big world, so progress will continue in different jurisdictions where liability can be avoided. One way or another, development will continue, but it may accelerate or decelerate in different jurisdictions around the world.
Hi Patrick, I agree empathy is important, but in Old age homes they are already starting to have robots at night to look after the elderly. I think they really need empathy, imagine being up at night and not feeling well….so I wonder where we are heading….AI has so many good parts as you mentioned but also scary things.
I agree that the development of AI is scary. I also agree that companion robots (or socialbots) will become commonplace within the next 5 years, not only for the elderly, but also for lonely middle-aged men. Keep in mind that there’s a 30-million female deficit in China as a result of their one-child policy (recently changed), and those 30 million men will need companionship too.