Like robotics, artificial intelligence (AI) was long regarded as a “future technology.” But, like robotics, we can now confirm that AI is more than mere sci-fi lore. AI is very much alive in our personal and professional lives, quickly becoming as mainstream as mobile devices. (How many times have you asked Siri, Alexa or Bixby to help you with something today? And how many conversations have you had with colleagues about the role that AI could play in your retail, manufacturing, warehousing, field service or government operations?)
Indeed, very real discussions are happening all around us today, with many asking “how can artificial intelligence can benefit humans?” And AI is generating a great deal of interest from companies of all sizes and across all sectors, with many organizations betting big on AI-powered systems, whether as a developer or end-user beneficiary.
But the truth is that we’re just beginning to understand AI’s potential, and AI’s current IQ may be a bit overhyped. So, we’ve asked Yan Zhang, one of Zebra’s AI experts, to help us understand how artificial intelligence works and whether or not it is a smart investment for your organization right now. You’ll be interested to hear what she has to say:
Your Edge Blog Team: When you hear the term AI, can you tell us the first image that pops into your head?
Yan: The first AI image that immediately gets visualized in my mind is the self-driving cars in complex road scenes. That’s one of the most impressive AI advancements and is still being improved.
Your Edge Blog Team: Is that based on your personal perception of AI or your biases as an engineer/developer?
Yan: This may be due to my “biases” as an engineer. The first AI image I had in mind 10 years ago was a “superman” that can do everything a human can do and more from sci-fi movies and novels. As my daily work in the Computer Vision field intersects more and more with AI, I’ve focused on applicable AI techniques such as those used in the self-driving industry that benefit various projects here at Zebra.
Your Edge Blog Team: How does that that compare to the description – or expectation – that your customers, colleagues or even next-door neighbors might share around AI?
Yan: Each person has a different understanding of AI based on his or her background, experience and profession. Some may view it as a superman-type being as portrayed in sci-fi movies. Those who work in manufacturing may relate it to the robot arms in their facility. Others may see it as the iRobot vacuum that cleans the floor on its own. AI engineers like myself view AI as Artificial General Intelligence (AGI) and Artificial Narrow Intelligence (ANI). While the former describes an intelligent machine that ultimately does everything a human can do, the latter includes more attainable AI techniques and applications such as self-driving, AlphaGo, smart speakers, home assistants, drone delivery and medical data analysis using deep learning techniques, just to name a few.
Your Edge Blog Team: Do you think these are all realistic descriptions of what AI is today or what it could be?
Yan: AGI is far reaching, technically speaking, and it also involves study beyond technology in sociology, ethics and social economics. ANI is already here with the examples above and still developing. There are quite a few home speaker offerings from different tech giants on the market. We also frequently see headlines on self-driving taxi and road tests. More robots and drones are being adopted within manufacturing, transportation and logistics and retail – and especially within warehouses. Down the road, AI-enabled intelligent automation will be more widely adopted in multiple verticals. Applying AI techniques to agriculture, environment protection, drug discovery and disease diagnosis in health care can benefit the entire human society.
Your Edge Blog Team: So, it sounds like AI really isn’t as intelligent as many people may assume. At least not yet. How close are we to artificial intelligence by its true definition?
Yan: AI is a broad term with various definitions, especially from the general intelligence perspective. AI was first used by John McCarthy in 1955 referring to a machine which mimics human cognition and problem solving. Machine learning is a branch of AI that was coined back in 1959 by Arthur Samuel as the field of study that gives computers the ability to learn without explicitly being programmed. A newer and commonly used definition on machine learning is the scientific study of computer algorithms that automatically extract patterns and build models from existing data and make predictions and inference on new data.
Your Edge Blog Team: With that in mind, should we really be using the term “machine learning” instead of AI?
Yan: This is an excellent question that is asked by many executives and other people new to AI. Michael Jordan, a prominent AI pioneer, has an excellent blog post that tackles this common “artificial intelligence vs. machine learning” terminology debate. As he notes:
‘Most of what is being called “AI” today, particularly in the public sphere, is what has been called “Machine Learning” (ML) for the past several decades.’
He goes on to explain that:
“…labeling of researchers aside…the use of this single, ill-defined acronym [i.e. AI] prevents a clear understanding of the range of intellectual and commercial issues at play.”
In other words, AI is often misused, which leads its current and potential capabilities to be misconstrued in the larger context.
Technically, we could use AI as a broad term when referring to an intelligent machine and then use machine learning to describe techniques and computer algorithms that extract patterns and build prediction models from existing data to complete prediction/inference actions on other data streams. But, in general, machine learning should be used when we refer to ANI—Artificial Narrow Intelligence.
Your Edge Blog Team: Understanding now that AI – and even machine learning – are still maturing, is it risky for companies to invest too heavily in such technologies today?
Yan: Not necessarily. There has been tremendous progress made around AI and machine learning and successful applications deployed over the past five years, especially in the Deep Learning area. A lot of performance breakthroughs in Computer Vision, Natural Language Processing, robotics and self-driving applications are attributed to Deep Learning technological advancements along with computer hardware, such as the Graphics Processing Units (GPU) required for AI and large-scale machine learning computing. Perhaps that’s why Gartner is reporting that nearly 60 percent of companies have already started to implement artificial intelligence and machine learning technologies in some capacity, with several companies planning to expand their utilization of AI and machine learning quite dramatically by 2022.
However, I caution these companies – and anyone else looking at AI and machine learning – to be realistic about the promises and limits of this technology in its current state.
Don’t get me wrong: I am not saying that you should wait to embrace AI or machine learning solutions until it matures more. On the contrary, in fact. AI is one of the “three A’s” that Zebra Ventures’ Senior Director Tony Palcheck talks about often when asked about strategic technology investments that forward-looking companies may want to consider right now and I agree. (Automation and analytics are the other two.) Just understand that AI and machine learning won’t apply to all tasks all the time.
Your Edge Blog Team: Which sectors stand to gain the greatest return on investment from AI and machine learning technologies right now?
Yan: Although not prominent in headlines, the financial industry widely adopted machine learning quite a while ago and has seen returns in portfolio allocation optimization as well as in the prediction of future financial and economic events. Tech giants such as Google, Facebook and Amazon have used and benefited from machine learning in a wide spectrum of their products ranging from online search, to recommendation, cloud APIs, home speakers and assistants, etc. AI and machine learning are also being adopted in retail and logistics to automate repetitive tasks and optimize workflow efficiencies through dynamic analytics.
Your Edge Blog Team: Would AI be a smart investment for an organization that seeks to partially augment and/or automate their workforce in the next 5 years?
Yan: I positively believe, at least personally, that AI would be a great investment, based on our engagement with customers in various verticals such as manufacturing, retail, and logistics and the study of pain points. Automating the workforce would increase operational capacity and create more value for customers. By relieving human workers from repetitive and labor-intensive tasks and instead assigning those tasks to machines or robots, the overall productivity and morale of the human workforce will be increased, which has numerous intangible benefits to organizations.
Your Edge Blog Team: How will AI systems rely on – or interface with – the mobile, Internet of Things (IoT), blockchain, augmented reality, automation and robotics technologies that are in play today?
Yan: All of these technologies can benefit each other and create efficient solutions side by side. In the case of mobile and IoT, various AI techniques have been developed and customized to mobile platforms with limited computation resources, e.g. optimized Deep Learning models for mobile devices. At the same time, mobile and IoT providers such as Qualcomm have developed powerful GPUs and tools to support more sophisticated AI techniques. AI supports automation and robotics techniques in many aspects from robot arm grasping and manipulation to autonomous navigation and path planning where real-time decisions need to be made on the spot on the edge processor. Edge AI has been the core technology for these solutions and is crucial to Zebra’s ongoing innovation considering our large device portfolio.
Your Edge Blog Team: Any last advice for those planning to leverage AI in some capacity to support their business?
Yan: AI can’t work on its own for an organization. AI experts and domain experts need to work closely with each other and apply AI to suitable solutions in order to create value for organizations and their customers. A successful AI project requires careful planning and expertise from multiple disciplines. While many successful AI applications have emerged, it’s also important to be realistic about AI’s promises and limits and keep in mind that AI or machine learning has overlaps with statistics. It shouldn’t be measured against the perfect accuracy. Lastly, understand that there will be doubts and uncertainty during the path of applying AI – just as there has been when adopting other new technologies in the past from horse carts to cars, personal computers and the internet. It’s critical to remain strategically focused with perseverance and tactically flexible with AI techniques.