ISE Magazine Volume : 50 Number: 6
By Joseph Byrum
An overview of what artificial intelligence can bring to society and businesses
The first Industrial Revolution of the 18th century introduced new means of mechanization, looms and steam-powered devices, which freed laborers from the most onerous tasks and liberated factories from the limitations of human and animal power. In the 19th century, the second revolution brought mass production techniques that allowed products to be efficiently manufactured at an unprecedented scale. The transformation to digital technology, led by the development of the computer, is the third revolution that dominated the late 20th century. Now we are on the verge of a societal transformation driven by the smart technology of the 21st century.
Each industrial revolution ushers in leaps in the speed, efficiency and utility of industrial processes that have a multiplier effect on growth. There’s good reason to think the fourth of these revolutions will be the greatest of all – which is why every manager needs to start planning to take advantage of it sooner rather than later.
What is smart technology? The term refers to any system or device that uses a combination of technologies that include machine learning, artificial intelligence, robotics and data analytics to accomplish more with fewer resources.
In the past, each industrial revolution allowed businesses to optimize their processes in a way that overcame certain human limitations, but the smart technology revolution promises to go one step further.
Smart automation may very well drive machines beyond human capabilities in every domain.
Leaping beyond human capabilities
Smart technology is an artificial creation that, by its nature, lacks many of the things that can hold back human performance. We are biological, living creatures with emotions that often get in the way and cloud judgment. Robots and machines do not share biological limitations.
Does that mean smart technology will necessarily outperform a human in all tasks?
Perhaps at some point. For now, there is evidence humans are still on top when it comes to highly complex tasks. For activities with clearly defined rules, such as chess, a computer can win if its programmers are up to the challenge of beating the best human. Examples of the machines’ success in increasingly complex tasks include:
- University of Alberta’s Chinook program beat Marion Tinsley, the world’s best checker player, in 1994.
- IBM’s Deep Blue beat chess master Garry Kasparov in 1997.
- IBM’s Watson beat the all-time top “Jeopardy” players in 2011.
- Google/DeepMind’s AlphaGo beat the champion Go player in 2017.
- Carnegie Mellon University’s Libratus and Lengpudashi beat human poker champions in 2017.
These accomplishments demonstrate the ascending scale of complexity that smart technology has overcome in a very short time. Checkers has 5 x 1,020 possible moves, while the number of possible games of chess is estimated to be 10,120. The complexity of the ancient strategy board game Go, with the potential for 2 x 10,170 legal moves in a single game, seems beyond comprehension.
To play poker, the machines needed more than just knowledge, they had to acquire the skill of bluffing. The AI had to mislead professional human players without them catching on to its strategy. Carnegie Mellon researchers estimated the complexity of heads-up, no-limit Texas Hold ’em at 10,160.
What might happen next
So smart technology has conquered ever more complex tasks from 1994 to 2017. It’s reasonable to say that over the next 23 years, smart technology will drive many more tasks far beyond human limits.
Once machines can learn and understand as we do, they will be able to become fully independent. They won’t depend on programmers and data scientists. That raises a number of questions. How will humans and artificial intelligence get along in future? Can we improve the chances of harmonious coexistence? Or will things get ugly?
Google’s AI research division, Google Brain, last year announced the PAIR initiative, which stands for “People + AI Research Initiative.” Its mission is to conduct the fundamental research needed to make future AI systems people-centric. It is part of a philosophy the Google community calls “human centered” in which machine learning algorithms solve problems with human needs and behaviors in mind.
This method is based on the principle that the simplest solution to a given problem won’t always involve machine learning. The initiative also stresses the importance of developing AI systems with human users interacting up front with prototypes, so that designers can understand the mental models that people form when they interact with the machine. The initiative encourages developers to anticipate failure, as algorithms will miscategorize input data a certain percentage of the time, for example, misidentifying one out of four cat pictures as a dog picture. System designers need to anticipate such errors and consider their consequences.
Keeping these principles will go a long way toward reducing conflict and ensuring AI adheres to Google’s former motto “Don’t be evil.”
Making the most of the fourth revolution
So how can managers make the most of nonevil smart technology that can surpass human capabilities in roles traditionally reserved for humans?
In many industrial systems, the human element can be the most difficult factor to plan for. We have memory lapses and rely on guesswork. We grow tired and make mistakes. AI excels at eliminating human factors to make decisions based on statistical probability. Just science and the best available data. Smart technology gathers available data from a business process and optimizes the data to provide the best outcome through superior decision-making.
Thus, harnessing smart technology in an enterprise will one day be as simple as going to a website and downloading the latest “AI” software (after paying a hefty licensing fee, of course). Unfortunately, such sophisticated learning capabilities are far from being off-the-shelf products for most industries. That means implementing AI in an organization at this early stage tends to be more involved, requiring AI solutions that have been custom built to address a specific problem using machine learning algorithms.
To create specific AI tools, managers must assemble a team that can identify the elements needed to address the business problem at hand. This will include gathering data and formulating the algorithms needed to make appropriate judgments based on statistical analysis. With these, the AI is able to conduct what are essentially an endless series of trial-and-error experiments to zero in on what works. Thanks to modern processing power, this happens in a fraction of the time it would take a human to complete the same task with less certain results.
Building an AI project requires subject-matter experts, software engineers and mathematicians who can work well together to create these needed elements. Such talent is often not available in-house. Existing staff has subject-matter expertise, but not every organization will have the specialists needed to create the algorithms that make the most of machine learning. It’s certainly possible to hire additional, full-time staff to fill these roles, but there could be cheaper alternatives. Thanks to open innovation, managers can build AI tools using crowdsourcing to tap into the desired skillsets on a per-project basis.
Patience is critical to the success in developing smart tools, because this process does not deliver results overnight. Rather, it takes substantial time and management resources to efficiently divide up the tasks before sending them out for crowdsourced solutions. Open innovation is a hands-on process that is guaranteed to disappoint anyone seeking a quick and easy fix. Quick and easy will have to wait until general AI is closer to reality.
Is it worth devoting the time to AI now, rather than waiting? To achieve smart technology’s benefit – resource optimization through improved decision-making – the answer is yes.
Those that fail to take advantage of smart technology are at risk of being overtaken by more forward-looking competitors.
Published at :