Artificial Intelligence = Better Decisions for Industrial Suppliers!
This blog post on Artificial Intelligence (AI) continues our series discussing new technologies that are impacting industrial suppliers and shares our recommendations for manufacturers and distributors. It was prompted by some startling news this week! DARPA announced a $2B campaign to develop the next wave of AI Technologies and their excellent (16 minute) video above summarizes the stages of AI development and where it's going next.
However there are many AI applications already in use by industrial suppliers, see examples below but first, in case you aren’t already familiar with AI, here’s an introduction. Skip the intro if you don't need it...
A New Source of Human Knowledge!
Pedro Domingos is a University of Washington professor, machine learning researcher and author of 'The Master Algorithm' book which offers many insights including that our natural sources of knowledge have been expanded:
- Evolution - a lot of knowledge is encoded in our DNA as a result of a very long process of weeding out the things that don’t work and building on the things that do
- Experience - knowledge acquired throughout our lives and encoded in our brains
- Our culture - knowledge that only human beings have which comes from talking with other people, from reading books and so on
- Computers - after thousands of years we have a new, fourth, source that is discovering knowledge from data and producing greater quantities of knowledge much faster than the previous three sources.
More Knowledge - Better Decisions?
Of course we use our knowledge to make decisions and more knowledge can lead to better decisions. Some decisions:
- We make alone - "do I want to date that person"?
- We delegate entirely to computers - like "is there a wrinkle in this car seat" (a type of automation)
- We make in collaboration with computers - such as a doctor making a diagnosis with the support of an AI system
This is the key point of AI - to enable us to make better decisions - especially in the 2nd and 3rd category :-)
What are Artificial Intelligence and Machine Learning?
AI is an umbrella term for many approaches to creating intelligent machines that can perform a task without human involvement. Machine learning is one of the many approaches under the umbrella. In simple terms machine learning is an analysis of large quantities of sample data to then make statistical predictions often via an algorithm. This gives machines the ability to learn to perform a task without human involvement. Here are a few examples:
- Speech recognition that responds to your voice or automated phone systems providing service at lower cost than human operators
- Generative design
- Face recognition
- Vehicle number plate tracking
- Traffic lights that reduce your wait time
- Industrial part search system that find the right part from a photograph
- Zilliant's analysis of >43.6M manufacturing transactions revealing where B2B industrial manufacturers can capture 10-50% more in annual profit and revenue
- Detection and recognition of cancerous cells
- Self-driving vehicles
- Tracking financial transactions to prevent fraudulent credit card charges
- Targeting online ads
- IT security
- Optimized CPQ product configuration systems
- Optimized routing of grocery shoppers through a store by re-sequencing their shopping list
This McKinsey podcast and transcript talks about how machine learning is already enabling better decision making, how it works and its limitations. But, in summary, with machine learning we train computers by giving them labeled data. To train a computer to recognize an object within an image, or recognize an anomaly within a data stream that signifies a piece of machinery is about to break down, we give it labeled data. We tell it, “in these types of images, the object is present, in these types of images, the object is not” or “in these types of data streams, the machine’s about to break down, and in these types of data streams, the machine’s not.” Having the data is the first requirement but labeling it is essential too - and labeling is actually generating a lot of work for people to do!
3 Ways Industrial Suppliers Could Use Machine learning
- As an analytic or optimization technique to deliver better results for questions you're already asking about your data, for example, the grocery shopper example above delivered a 50% improvement and was built by just three engineers, using Google's open-source tools Keras and Tensorflow
- To ask new questions of the data you already have, for example, a manager reviewing customer emails might search for 'angry’ emails
- To allow analysis of new data types such as audio or video because we’ve taught the computer to ‘see’ (and 'hear') as well as 'read' text and numbers - so cameras (and microphones) become sensors generating streams of (potentially) machine-readable data. For example, an automobile car seat supplier put a neural network on a cheap Digital Signal Processors (DSP) chip with a cheap smartphone image sensor, to detect whether there’s a wrinkle in the fabric. Expect all sorts of similar uses for machine learning in very small, cheap widgets, doing just one thing, as described here.
4 Recommendations for Industrial Suppliers
1. Don’t fear AI, use AI where it can help improve decision making. AI-powered automation will lead to job losses and changes – individuals will need to acquire skills that work with, not compete against, machines. In an opposite to robotics, white collar roles (like engineers, lawyers, doctors, etc.) will be easier to automate with AI because they are the roles that make the most decisions based on knowledge! As Andy McAfee of MIT describes in the video above, company executive's future value will be less about making the tough judgement calls and more about setting the vision while managing company culture to achieve it.
2. Recognize that AI is still developing. DARPA’s funding will likely help make development more rapid and the boundary between what is better done by the machines and what is better done by humans will keep changing. In most jobs, expect a combination of human and computer will lead to the best decision making.
3. Understand AI’s current limitations. In many cases our AI machines have become much better than humans at doing their specific tasks. But knowledge deduced from data (even very large samples) is uncertain because we can never know if it has been generalized correctly or not. Humans are often more certain of their knowledge than they should be and similarly we need to be careful with knowledge from our new 4th source! Can we actually explain what the algorithm is doing? Can we interpret why it’s making the choices and predictions it’s making? Is the data available or do we need to start collecting it? Is it labeled or how will we label it? Is there any bias in our data? Is there any bias in the way the data was collected?
4. Decide whether this technology, as available today, could improve one or more of your company’s processes? From the examples above appreciate that potential applicability is broad. Think about the potential for AI by following the money to where revenue can be increased or costs reduced. For McKinsey’s suggested process see How advanced industrial companies should approach artificial intelligence strategy.
We'll watch DARPA’s “AI Next” campaign with great interest. Of course the industry isn't dependent on DARPA's schedule and it's possible that companies (like Alphabet, Amazon, Microsoft, etc.) will develop third wave AI as, or more, quickly but this program could light a fire under them. Either way AI pioneers should rapidly move beyond today's 'basic' machine learning and closer to AI that actually thinks and explains its thinking. As always, please share your comments below or, if you'd like our opinion on your use of AI and machine learning, please call us or click either button below: