Know My Company
Tell us about your interaction with AI and other intelligent technologies that you work with, in your daily life.
AI is not only my daily business, but it’s also my hobby. The relevant part for me in AI/ML is “automated scientific research” by analyzing existing complex data to get statistically significant insights and individualized predictions from data, and “decision automation” by building large scale decision systems to optimize for a given goal on the basis of the insights and predictions.
As a physicist and data scientist, I am constantly thinking about how algorithms can be developed, improved and advanced to benefit the enterprise and more importantly, create a better world, on the basis of available data.
How did you start in this space?
I was a professor at the Karlsruhe Institute of Technology and participating at experiments at CERN where I invented the NeuroBayes algorithm. This neural network can predict complete probability distributions for single observations, trained on many observations. These probabilities are conditional on many input variables and thus allow very individualized predictions. I realized this algorithm was very powerful not only in physics research but also for businesses. In 2002 I began working towards my dream and founded my first ML (project) company. In 2008, I founded Blue Yonder, today a leading provider of AI and ML solutions that enable retailers, consumer packaged goods and other companies to intelligently transform their operations and make more profitable, automated business decisions that deliver higher profits and at the same time optimized customer experiences. In 2018, JDA acquired Blue Yonder and now together, we deliver artificial intelligence and machine learning (AI/ML)-driven supply chain and retail solutions for 4,000 of the world’s leading retail, manufacturing and logistics companies.
How do you differentiate JDA from other similar service providers?
JDA is unique in that we can deliver AI- and ML-driven solutions from end-to-end for the complete supply chain from manufacturing to the store and up to the end customer, for the first time with real, tangible results for customers. Many companies talk about AI and ML, but few are actually doing it, or have customers finding success with it. Some have AI/ML Labs, but not a scalable and reliable product that a large enterprise trusts to take tens of millions of their key decisions worth millions of dollars per day fully automatically.
With a supply chain cloud platform storing data for the complete supply chain and with AI and ML, companies can connect data and insights to enable more profitable operations, as to overcome the silo optimizations of the past. All these are keys to the Autonomous Supply Chain™ decision-making capacity needed to simultaneously optimize profit and customer experiences.
How do you see the raging trend of including ‘AI in everything’ impacting businesses?
There is a need to differentiate AI’s hype from its real value. The beauty of this technology is that it can be applied to nearly every domain, but that doesn’t mean that it needs to be or even should. AI’s true value lies in solving specific problems where it is proven to deliver better and more constant quality results than humans could do. As such, it should be applied for improvement; whether it is automating mundane tasks or using data to make smarter decisions about everyday business operations. Businesses that have embedded AI into their decision making for these reasons will not only survive but thrive.
What are the biggest challenges and opportunities for AI companies in dealing with rising technology prices?
Like all companies, AI companies should work economically. The constant growth in CPU and storage described by Moore’s law has led many to forget that and just plug larger software frameworks together to solve their problems. Some computing resource providers make tools exactly for that but are not necessarily economic for every task. This might work for some time, but limits will be reached soon, especially as companies continue to scale.
From our historic heritage in Elementary Particle Physics, we always had too much data to be handled comfortably with the then present technology, and as such, have concentrated in algorithms that are powerful but economic. Also, data structures and storage should be optimized, taking into account quality, speed and cost, which in the cloud actually becomes more visible.
How should young technology professionals train themselves to work better with AI?
AI requires life-long learning. This should start with a deep education and tedious study of math, statistics, computing and data science. Don’t just learn one tool or one perspective, however. Continue to explore multiple ways of solving problems through hands-on experience. Learn relevant programming languages (I prefer Python) and play with ML algorithms in your spare time. Go to data scientist meetups and join a group or active community that facilitates knowledge transfer, common learning, and experimenting.
How do you consume information on AI/ML and related topics?
I’ve found that you can learn a lot from scientific publications and books, but you learn more and faster from your peers. I look to my colleagues, an entire network of data scientists and engineers, for new information on AI and ML. I also enjoy participating in talks and seminars at colleges, universities and industry conferences.
What makes deploying AI so hard?
It’s really a culture and mindset change that needs to occur. AI technologies make decisions that revolve around a level of certainty that humans simply can’t reach. No matter how old or traditional the sector, how experienced the human or how consistent you think your company’s operations may be, it will always be the case of “opportunities missed” for companies who decide not to embrace this technology.
And while it’s understandably a big psychological hurdle for many to allow machines to make decisions for them, it is a must for companies to maintain a competitive advantage. Using AI needs new processes, a new understanding of the human role and a new definition of leadership.
Which is harder – choosing AI or working with it?
Choosing AI. Companies often hold onto familiar tools for a long time and it is difficult to overcome that resistance. As previously stated, they must accept the idea that a machine can make far better predictions and decisions than a human can. However, once the investment is made and they begin to see specific improvements, it’s easier to make the case for a wider rollout of AI in the organization.
Working with AI also requires good change management to ensure you’re not destroying historically smart human behavior.
How potent is the Human-Machine intelligence for businesses and society? Who owns machine learning results?
Machine intelligence is a sharp weapon and can be used for good and bad. However, the good outweighs the bad, especially in retail. ML and AI can automatically generate smarter business decisions in situations such as improving pricing, promotions, markdowns, and replenishment capabilities. It can also create benefits with fast and short lifecycle products, leading to less waste, optimized labor, and improved sustainability in the fresh food chain. Our technology helps retailers achieve this by enabling them to drive decisions and predictions from their data, and as such, they own the results.
Where do you see AI/Machine learning and other smart technologies heading beyond 2020?
I think the largest value generated by AI in the next years will be in supply chain and marketing.
Why supply chain? Because of the value generated by data-driven algorithmic mass decision systems that already exist now and that will help the front-runners in retail, CPG and manufacturing be superior to their competitors in millions of daily repeated decisions. AI can take many more factors and complicated nonlinear and correlated dependencies of the massively abundant data into account – much better than a human can do – and predict the near future without human bias but with a proper risk assessment. This yields optimal decisions, something the human brain was not developed for, and with improvements in every single decision, that adds up to a huge number on an aggregated level.
Why marketing? Because individualization and progress in causal algorithms can make the (now usually very low) efficiency of marketing measures much higher and will reduce today’s enormous marketing budgets.
Then, I am convinced that progress in public research in natural sciences and medicine will become much faster once ML/AI is commonly applied.
The Good, Bad and Ugly about AI that you have heard or predict –
That AI will replace humans or even fight against humans. This simply is not true. AI should not be viewed as a danger, but as a new opportunity to improve our society and sustainability in exploiting our planet’s resources. AI helps humans by enabling them to make smarter, more intelligent, predictive and prescriptive decisions that will impact their company’s bottom line.
What is your opinion on “Weaponization of AI”? How do you deal with the challenge here?
I think it is overstated. Every company needs an AI plan. Especially repetitive mass decisions for very specific tasks that can be automated, and at the same time individually optimized by AI/ML algorithms, based on experience in terms of collected data. This leads to an alignment of a huge number of everyday decisions with the global company strategy and thus better KPIs at a lower cost. To combat this challenge and hype, I recommend looking for CEOs and CFOs reporting about real successes achieved using AI in competitive low margin business.
On a more general level, AI/ML can be a sharp weapon, and as with each new technology, society must find rules about what is allowed and what not.
What AI start-ups and labs are you keenly following?
No specific ones, there is so much going on out there that I concentrate on acting, not reacting.
What technologies within AI and computing are you interested in?
Our in-house developments: NeuroBayes, Cyclic Boosting, and BY-Causality. BY-Causality is finding and isolating causal relations instead of just statistical correlations from historical data — this is important whenever one wants to change action policies for e.g. better customer targeting and couponing for better. OR-by-AI (solving complex operation research tasks by artificial intelligence).
In general, economic algorithms, scalable algorithms, human explainable algorithms and algorithms predicting complete risk profile (conditional probability densities).
What’s your smartest work related shortcut or productivity hack?
There is no real shortcut. I firmly believe in life-long learning, constant deep and broad interest, curiosity, fearless trying, simply doing, deep understanding, transfer thinking, exploiting mathematics and simple physics principles (the world is simple in its basics), simple but powerful programming languages like Python, and exchange with good peers. To create artificial intelligence algorithms and products you need natural intelligence. And eliciting emotion can create a cognitive response to the overall learning experience, making you much more fascinated with the task at hand. And this is contagious.
Probably most important for productivity is to follow your own vision and not asking for an allowance. Acting, not reacting. And not getting stuck in personal fights, instead of achieving the next goal independent of the opinion of others.
Tag the one person in the industry whose answers to these questions you would love to read.
Thank you, Michael! That was fun and hope to see you back on AiThority soon.
Prof. Dr. Michael Feindt is the brain behind Blue Yonder, a JDA Software company and the market leaders in AI in retail. Blue Yonder is powered by Michael’s NeuroBayes algorithm, developed during his many years of scientific research at CERN, which enables retailers to automate complex decisions across the entire value chain. With AI embedded into their supply chain and merchandising processes, retailers can respond quickly to changing market conditions and customer dynamics, boosting revenues and increasing margins.
Blue Yonder, a JDA company, is a leading provider of artificial intelligence (AI) and machine learning (ML) solutions that enable retailers, consumer products and other companies to intelligently transform their operations and make more profitable, automated business decisions that deliver higher profits and optimized customer experiences. With AI/ML learnings embedded into their core supply chain and merchandising processes, companies can respond quickly to dynamic market conditions and customer preferences, resulting in increased revenues and margins. Blue Yonder provides its solutions through Microsoft Azure and was named one of Microsoft’s retail partners of the year in 2018.