- Artificial Intelligence was first discussed in an academic setting during The Dartmouth Summer Research Project in 1956.
- AI can deal with new situations, whereas traditional software only handles the cases it’s programmed to perform.
- Rules-based programming is limited but precise, but AI can handle much broader capabilities. It can (and will) make mistakes in new situations.
By Greg Cholmondeley
Artificial intelligence (AI) is perhaps one of the greatest buzzwords of all time. This high-tech term is bantered about endlessly in the media, and you can’t help but imagine cyborgs and self-driving cars when you hear it. Even so, AI isn’t science fiction and it’s much more than just a buzzword—it is currently available technology that is poised to transform much of our world, including the printing industry, in the coming decade. Despite its popularity, the term “artificial intelligence” is often misused. This piece explores what AI is and also explains how it differs from more traditional software programs.
Defining Artificial Intelligence: It’s Complicated!
The concept of AI has been around for a long time. In fact, AI was first discussed in an academic setting during The Dartmouth Summer Research Project in 1956. Of course, we didn’t have the computational power, storage capacity, or interconnectivity necessary to realize its potential back then. By the 1980s, however, our computers were powerful enough to support early forms of AI called expert systems. Then, in 1997, IBM’s Deep Blue chess-playing AI beat Grandmaster chess champion Gary Kasparov at his own game. This involved a supercomputer in 1997, but that is no longer required given how our computational infrastructure has advanced over the past quarter-century. Cloud servers now have the power and capacity to apply this technology to a wide range of areas, including weather prediction, virus analysis, or asking questions of your phone’s voice assistant.
With this information as a backdrop…what is AI, and how does it differ from traditional programming? Most AI definitions are less than helpful. Consider these examples:
- According to Oxford Languages, AI is “the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.”
- According to Merriam Webster, AI is “an area of computer science that deals with giving machines the ability to seem like they have human intelligence.”
Isn’t there a fundamental rule not to use part of a phrase (in this case, intelligence) when defining that phrase? The question then becomes, what is intelligence? Merriam Webster defines intelligence as “the ability to learn or understand or to deal with new or trying situations, or the ability to apply knowledge to manipulate one’s environment or to think abstractly as measured by objective criteria (such as tests).”
By combining the offerings above, I produced what I believe to be a much more useful definition of the AI term:
Artificial Intelligence vs. Traditional Programming
The simple definition provided above highlights the key difference between AI systems and traditional software programs. AI can deal with new situations, whereas traditional software only handles the cases it’s programmed to perform. Consider automating a prepress workflow using traditional techniques:
- You could link together strings of prepress functions into unique workflows. Prepress operators would drop jobs into hot folders for each. This automates prepress, but it requires separate workflows for each situation and operator involvement to select the proper workflow.
- You could write software to inspect job tickets and document files and use if/then/else decision trees to process jobs. This reduces operator decision-making, but it would become quite complex if forced to handle every situation. Furthermore, these programs dump anything that is not recognized into manual handling folders.
- Advanced systems like Xerox FreeFlow use rules-based programming with reusable subroutines and modules to dramatically reduce the amount of coding. Still, this requires specific coding of every possibility based on the jobs and characteristics of each unique installation.
By comparison, an AI solution would begin with a core set of capabilities that grows as it learns the types of jobs and how you handle them in your shop. You teach the software to handle work rather than programming it. This is like hiring a knowledgeable prepress operator and training them to work in your shop.
Some of the notable differences between traditional programming and AI programming are:
- Traditional programming only does what it is told to do. AI excels at pattern recognition and learns by correction. Thus, rules-based programming is limited but precise, while AI can handle much broader capabilities—it can (and will) make mistakes in new situations. AI learns from its mistakes, just like people do.
- Traditional programming rapidly grows in programming size and complexity as the environmental complexity increases. Thus, it works well for applications with limited and well-defined possibilities. AI is designed to handle new and unexpected situations and more real-world applications where anything can happen.
The Bottom Line
While AI is gaining footholds in numerous segments, it’s still in its infancy in production printing. That will change in the coming decade, but early examples include Tilia Labs’ products and Ricoh’s Pro Scanner option.
Even though most mainstream AI developers are not specifically focusing on our industry, AI is coming to print production automation. Production printing has many areas that are well-suited for AI systems. Most of them are not as complex as applications like self-driving cars, fraud protection, personalized shopping, or healthcare advice.
If you’re a vendor in this space, think about how your offerings could benefit from AI, and consider acquiring AI experts because they have a very different skill set than traditional programmers. If you’re a print provider, start automating your systems and collecting your printer and MIS data. AI solutions require automated systems and data to work. Transitioning from a fully manual operation to an AI-automated one will be more complicated than simply adding intelligence to your automation.
Greg Cholmondeley is a recognized expert on workflow automation, strategic planning, software solutions, and the printing industry. Before joining Keypoint Intelligence, he was President of PRINTelligence Consulting, where he analyzed and assessed production enhancement software solutions for vendor development and consumer understanding. He is a frequent speaker and panelist at industry events as well as a published author.