Artificial Intelligence is a transformational technology. Similar to the industrialization of the 19th century sweeping across nations and changing societies systematically, AI has been slowly meta-morphing all fields and even social orders.
As an observer and recorder of the change sweeping the world, we got in contact with Patrick Bangert, who is an outrider in this field. Patrick is the Vice President of the Artificial Intelligence Division at Samsung SDS. He commenced his journey as an academic but quickly pivoted to the industry due to a greater interest in finding solutions to real-life problems.
After an entrepreneurial stint where he established his company that brought AI solutions with a vision to tackle industrial problems, Patrick headed towards the frontline divisions of the AI industry in the world. At Samsung SDS, he continues to drive innovative solutions that bring about real change in people’s lives.
In an exclusive with CIOLook, Patrick shares his expert insights on the rapidly accelerating and disruptive field of AI:
Brief our audience about your journey as a business leader until your current position in your company. What challenges have you had to overcome to reach where you are today?
After getting a master’s degree in Physics and a Ph.D. in Applied Mathematics, I first started out in research as a professor of mathematics in German y. Seeing that I could not apply the craft to real-life challenges from academia, I eventually formed a company, Algorithmica technologies, that brought machine learning and artificial intelligence to the chemicals, power, and oil-and-gas industries.
We focused on two major problems: predicting machine failures in advance and calibrating physical processes to function more economically. That experience turned me from a mathematician into an entrepreneur. I learned about marketing, sales, legal issues, consulting, and many other aspects necessary to running and growing a business.
As the company grew, I became a manager of people as well, which involved many softer skill sets. To this day, I think there is no degree or course that prepares you for the breadth of running a company, and so everyone must figure this out on their own with occasional advice and help from friends along the way. These friends are the secret sauce of the journey!
After an exit from Algorithmica, I joined Samsung SDS – the IT company in Samsung Group – as their leader of the Artificial Intelligence (AI) team. We do AI work from time series, over natural languages, to computer vision. The work I’m most proud of is our work on medical images. It’s gratifying to create mathematical models that save people’s lives by providing fast and reliable diagnoses.
Tell us something more about your company and its mission and vision.
Samsung SDS is the IT company of Samsung Group, which is an international conglomerate of companies spanning consumer electronics, computer memory, and other parts, as well as diverse other businesses.
Samsung SDS is driven by a mission to boost South Korea’s IT competitiveness, constantly developing the country’s digital solutions and services industry for over 35 years, all while meeting the needs of global consumers and leading the latest technology trends.
Enlighten us on how you have impacted your niche through your expertise in the market.
When you look at AI today, the conversation is almost always about models and algorithms. There is a lot of hype, enthusiasm, and also fear. In fact, the making of models is something that is fairly mature for most simple purposes. The difficulties lie in two major places – preparing the data for training and embedding the model in the right software ecosystem for it to provide tangible value.
These two aspects take time, require people, and cost money. Unfortunately, they are often neglected or forgotten, and this is the main reason that nine out of ten AI projects in commercial practice fail to yield economic value.
A pivotal activity in most AI projects is the combination of raw data with the domain expertise of those people who are knowledgeable about the data, and whatever process gave rise to it. This process is known as “labeling” or “annotation” and could be quite complex. An example is outlining a traffic sign and identifying it as a stop sign.
One of the main contributions of my group is the building of a complex toolkit that essentially automates this workflow and reduces human labor by over 90%. Combined with other automation methods of AI, this ultimately allows the creation of a good dataset and the training to occur in a handful of weeks which used to take many months.
Describe in detail the values and the work culture that drives your organization.
Samsung has five core values: people first, pursuing excellence, leading change, upholding integrity, and ensuring co-prosperity. One could go on at length about what these mean and how they are lived in daily work life. For the AI team, pursuing excellence, for instance, means that we aim to produce very accurate and performant tools and models that hold par with the state-of-the-art (SoTA). In fact, we aim to advance SoTA when we are able and thus lead the change for the better.
We uphold integrity in part by being honest about what AI can and cannot provide. For instance, we cannot be sure if AI will ever be sentient, but this will certainly not happen any time soon, and so narrow AI is all we will have for the foreseeable future. Sorry, Hollywood. We put our people first by providing maximum work-life balance flexibility and a long-term workplace. We ensure the co-prosperity of society at large and the ecosystem of suppliers and customers by being a reliable business partner and being seriously involved in the entire workflow, or supply chain, of the process we are involved in.
Our efforts in medical imaging speak to all five of these core values in particular.
Undeniably, technology is playing a significant role in almost every sector. How are you leveraging technological advancements to make your solutions resourceful?
We stand on the proverbial shoulders of giants, to be sure. The academic and open-source community in AI is exceptionally vibrant. Just keeping up with what is happening is a full-time job for a small team. In recent years, under the heading of reproducibility, it’s become very common to publish not only the insights in the form of a paper but also the full data and source code of an advancement. We leverage these resources in our work and integrate the best methods that we find relevant for our business problems.
We also collaborate with other commercial companies and universities in integrating or building proprietary solutions wherever this makes business sense. These partnerships may be concerning AI, but often they examine a variety of software and hardware aspects of the workflow before or after the model itself. A special emphasis is on the computational hardware, such as the graphical processing units (GPUs) and data storage and retrieval, which are both crucial to doing AI in practice.
What change would you like to bring to your industry if given a chance?
Honesty. There is so much hype in AI that we experts regularly spend a lot of time and effort talking people off ledges. Much of what people think about AI is not realistic, both the good and the bad. AI does not deliver fantastic results at the press of a button but requires a substantive investment of time and effort. The Terminator is not real and will not be real for a very long time indeed, if ever.
Many commercial companies, often startups, over-sell their capabilities and under-deliver on tangible results. This does not help them and, in fact, makes it worse for everyone because users and customers have become suspicious, and rightly so. The well-known effect of pilot purgatory is the result – interminable testing with very little or no payment. At the same time, the startup kills itself in attempts to make a customer happy, all the while forgetting that “happiness” is not a tractable metric of project success.
What, according to you, could be the following significant change in your sector? How is your company preparing to be a part of that change?
Everyone uses AI multiple times a day and does not necessarily realize it. The next piece of AI that will visibly change the lives of the general population will be autonomous driving vehicles. From an AI perspective, this is a solved problem (at least for all those cases one is likely to encounter in regular traffic). So this is now “merely” a question of manufacturing, sales, legal, and compliance issues to scale up the market. AI in healthcare is the next extensive application of AI – one that is not yet solved scientifically. When it comes in 10 years, it’ll be comparable to the discovery of penicillin, a step-change for the better. AI can diagnose diseases instantly and more reliably than any single human doctor. It can transcribe and file patient-doctor interactions and automate paperwork such as prescriptions and invoicing so that doctors no longer must be data entry clerks but can transition to being caregivers.
Where do you envision yourself to be in the long run, and what are your future goals for your company?
AI is at an exciting moment in its evolution. Its use is widespread in a multitude of diverse narrow applications and continues to be built out daily. The main challenges scientifically are to provide explanations for AI model outputs and to ensure ethical standards for models, their ecosystem, and their uses. The principal obstacles commercially are defining the business problem well and setting the right expectations, including the investment and returns outlook. I see myself leading a major AI division or company at the intersection of these technical and business challenges.
Applications that make a real difference and make the world a better place excite me. Healthcare or climate change are such areas that could benefit Immensely from serious applications of AI in various specific use cases.
What would be your advice to budding entrepreneurs who aspire to venture into your sector?
Much advice has been given by my betters. Some of this bears repeating here. The prime directive for startups is focus, focus, focus. Solve one specific problem really well. Make sure that you really do solve it – from the end user’s point of view. That typically involves many things that are not AI, such as software and user interfaces, good support, and a sensible business model. Maintain a good network of business contacts that help you along the way by being early trial customers or giving advice. Be careful in who you select as your co-founders, as this is often more binding than a marriage.