Here is something we all love: tech buzz words and the grand promise of new technologies. You know its right, because me, you and the person pitching the sales deck, we use them all the time. “The Internet Of Things will connect everything,” “blockchain will democratize everything,” and “AI will solve all of our problems.”
AI specifically is high on the buzz word list: from traffic jams to climate change, there is a solution, and it is the new breed of machines that can think and act like us.
Most of everything that was once plain “digital” is becoming “AI-enabled.” AI is all around us from Instagram filters to Targetted marketing. If your company is not AI-enabled, are you missing out?
Let’s pause for a minute and challenge our selves a bit, is AI here at all? Are all the claims correct, or is it just a lot of marketing buzz? And when will true AI arrive?
1. First Off, Why You Should Even Care If AI Is Here Or Not
Before diving into the question if AI is, and will it ever be a reality, lets first establish why we should care about the answer.
If you are a business leader, a corporate or startup entrepreneur, you want to take advantage of the best of what technology has to offer. With new technology capabilities, such as AI, we can revisit old problems and potentially find a better way of solving them. We can also take a stab at some unsolved problems. Maybe with this new tech, we can resolve a big hairy issue that had no solution because the necessary tech was not available. Either way, if with new technological capabilities, we solve old problems in a better way, or if we solve problems that previously had no tech solution, we create new business value. Leveraging tech to create new business value is good news for business; there is, however, one requirement. The tech has to be able to deliver the goods.
If you are in a position to make decisions on how problems will be solved using new tech, and how new business value will increase by leveraging new tech, you need to know what is hiding under the hood. Can the buzzed tech of the day, which is on everybody’s lips, do what you expect of it? The answer to this question can come in three ways, two of which are bad for your business:
- If you undershoot tech’s capabilities, then you are leaving business value creation opportunities on the table.
- If you overshoot, you will inject your operations with risk and will not be able to deliver a growth generating product to the market.
- If you get it right, you can launch a product in the near term and leverage new tech to increase your business value
Back to AI. Here is a reality check. If AI is prominent in your company website, in your investor decks, and your product description, as of 2/2020, you are overshooting. Let’s dive into the details of why that is.
2. What Is Wrong With How We Understand AI
The concept of AI has been around for a while. Scientists and engineers have been fascinated with the possibility of creating a machine that can think like us. Those early AI entrepreneurs believed they understood the mechanics of how our brains work, and based on this understanding, all they had to do is create a digital replica of this process.
To examine how much we progressed on the path of making machines that think like us, take the famous “is this a cat” case. An early proof point showing AI can “think like us” was to train a machine to identify if there is a cat in an image.
Identifying an object in an image is a complex task that until “AI” showed up, only people could do. Cats were not a random choice for this exercise. There are an endless supply fo cat images online, most of which contain a useful text tag such as “look at my beautiful cat” or some other sentence of a similar nature. For training a machine to identify cats, these tags are priceless. For a machine to be able to determine that an image contains a cat, many thousands of tagged images had to be processed. Using Machine Learning, the machine can identify a pattern (pattern recognition), which it can use to determine if a picture contains a cat. This “self-learning” process is fascinating, useful, and has multiple business use cases, but is it intelligence?
Going back to the initial intent of AI developers, AI was supposed to enable machines to think like humans. For comparison, a child needs to experience two interactions with a cat to be able to identify a cat the third time they meet. This is a significant difference in capabilities, and it highlights that how people and machines think, learn, and understand is very different.
3. Machines Can Compute Better
Machines are good at repetitive tasks. A factory robot punching a hole in a piece of sheet metal, or a sorting robot in a distribution center, will never tire, complain or strike. Coupled with Computer Vision and Machine Learning capabilities, these machines can even improve how they perform tasks. But that is precisely the point. We design machines to perform a specific task, and at a best-case scenario with current tech capabilities, can improve their performance on the intended task. They will not learn how to all of a sudden, read a book. They will not turn their optical sensors to the sky and wonder what those shining lights that show up only at night, and they will not develop any feelings of friendship and comradery with their neighboring robot. Machines have no emotions. People, on the other hand, are driven by emotions. Without this missing component, can machines ever think like us?
As of today and the near future, machines can perform specific tasks very well; in some cases, they can even learn and improve their performance of a particular job, but is this intelligence?
4. Artificial Inteligence or Inteligent Automation?
Definitely, for the present, but also the foreseeable future, the business opportunities of leveraging what we call “AI” are more IA (Intelligent Automation) than AI (Artificial Intelligence).
Using Machine Learning, Pattern Recognition, Neural Networks, and host of other market-ready advancements in the field of AI (note, in the field of AI, not yet AI), we can increase our ability to automate processes. More tasks, which in the past demanded a human operator, can now be shared or ultimately handed over to machines.
With all of these beautiful advancements, these “machines” (in the broad sense of sensors, actuators, edge computing, and cloud capabilities) do not have Artificial Intelligence. They do have a growing ability in which to make sense of data, handle more significant sets of data, interact with the physical world, infer data learn, and improve how they perform over time.
But only for the task for which we designed them.
The person that once stood where now a sorting robot is standing could do much more than just sort items on a conveyor belt. It could read, it could get angry, fall in love, and figure out what is the best way to ask for a raise. The human is intelligent, machines are still just faster at computing. What is rapidly changing is the ability of computers to compute, the complexity of tasks these machines can handle, and the ability to learn in the sense of improving on past actions. We can inject machines, processes, environments, and supply chains with Intelligent Automation, but is this intelligence?
5. How We Think
Should AI aim to teach machines to think like us? Let us consider for a moment how we, as humans think. We base our thoughts on past experiences and feelings. We feel faster then we think. Culture molds our thinking. When an occurrence happens, we first set an emotion in motion, and that feeling triggers a thought. Machines have neither past experiences (if they do, they are limited to narrow fields such as identifying a pattern of a cat, driving in a lane, or picking an object), and they most certainly do not have feelings. Missing these components, how can machines think like us? What is becoming clear is that the holy grail of AI should not be to teach machines to think like humans, but to teach machines to think like machines.
6. How To Sew A Shirt
We have fallen into this trap before. We have tried to turn machines into automatons that mimic human behavior; these attempts failed. Machines can do incredible things, but we need to design them in a way that fits their strengths. Today we struggle with building autonomous cars, intelligent analysis tools, complex computer vision systems, but we are not the first to try and automate our world. Rewind to the early 1800s and the challenge of the day was how to build a sawing machine. The world population was increasing, and walking naked was not an option. There were just not enough free hands to make the clothes to meet demand. We had to come up with a better solution for making garments.
Sewing two pieces of cloth together is a highly complex task. Eye and hand coordination are needed. There are patterns to consider. Skill in deciding how much force is necessary to pull on the thin thread before in tears. How on earth can a machine be taught to do all of that?
The technology of the day was mechanics and electricity. Many entrepreneurs tried to build machines that mimic the moving hand, the motions of threading a needle, and all failed. Machines just do not move like humans. Finally, after many failed attempts, the challenge was answered when Elias Howe in the 1840s looked at the process of sewing and engineered a machine version of this task, hence the “sewing machine.”
While the human hand has multiple degrees of motion, sensing capabilities, and the benefit of a set of eyes to serve as a guide, all the machines of the time could do was rotate or move linearly. For a sawing machine to work, it had to operate within these constraints.
Elias Howe solved the puzzle not by making a machine that imitates humans but by understanding the constrains of how a mechanical machine can operate.
The outcome of a human sewing a dress or a machine sawing a dress may look similar, but how they go about completing this task is entirely different.
7. AI Will Be a Myth Until…
If we continue and pursue AI with an intent to mimic human thinking, we stand to fail, because humans and machines think in very different ways. To increase machine computation capabilities to the point in which the end outcome is similar to a decision a person makes will be achievable when we understand the best way for machines to think like machines. Until that day, the safest bet is to take out AI from your marketing materials and replace them with ML (machine learning). To make the best use of what tech has to offer today, focus on which processes, behaviors, environments, and supply chains can be made more intelligently automated and leave intelligence out of the equation for the time being.
Credit: Google News