Why can't your computer get frustrated, too?

Why can't your computer get frustrated, too?

Brookfield, Wis. — Anyone working with a computer has those moments when things go wrong and then just go wrong again. It could be an everyday operating system feature, a difficult level in a video game or a diagnostic tool in an industrial setting. The user gets annoyed, angry and frustrated — and then realizes how to do things differently and make the machine work.
Should the computer feel and act the same way?
A Wisconsin company is working toward that end, trying to enliven computers with emotions and thinking capabilities.
“It’s actually enacting what the human is doing,” said Helena Keeley, who, along with her husband Tom, own and run Brookfield-based Compsim. The company, founded in 1998, is a software technology developer working on artificial intelligence (AI) technology.
The company’s KEEL (Knowledge Enhanced Electronic Logic) engine, a graphical programming environment in which the computer can balance different criteria and relationships against each other, aims to create artificial intelligence that can make decisions and alter its course based on errors it encounters.
Helena contrasts that kind of dynamic reasoning with traditional programming, in which a program knows only the right way to do something.
But what if a machine could go through failure over and over at some task at hand and adjust its whole approach?
“It’s just embedding human reason reasoning, and that’s what we believe humans do, so that’s what we believe devices should do in some cases,” said Helena, who serves as Compsin CEO.

Artificial intelligence models

While computers have been getting faster, and people have used them for an ever-expanding number of purposes, most uses of the machines require a human being as the active decision-maker: whether to go to a certain website, whether to rip a CD, type a document, etc.
Video games, a field dependent on the cutting-edge of computing technology, still set up formulas for game flow and for characters to behave in certain ways in reaction to the player’s decisions.
“One of the core things about creating a simulation or a model is that you need to know for what purpose you’re creating your simulation or your model,” said Doug Whatley, CEO of Breakaway Games, a developer of both commercial games and professional’s simulations such as the medical game Pulse. “Part of what holds AI back right now is that artificial intelligence has simply been focused on pattern recognition and the basic steps of decision-making as opposed to the bigger picture of why decisions are made and when they need to be made.”
The key, according to Tom Keeley, who serves as Compsim president and chief programmer, is how the system works. By treating matters of judgment as a right-brain analog process, the programmer’s job is to create an environment in which criteria can be taken in and measured mathematically, essentially converting a right-brain process into a digital process fit for the pure left-brain calculations of a computer.
“What we have, then, is this graphical language, which allows you to define the importance of information,” Tom said. “And it allows you to essentially, with those lines, define how each piece of information interacts with other pieces of information.”
Tom said his technology can offer access to the inner workings of decision-making not only in a human-like way programmers would like to see machines behave, but also in others ways that people cannot.
“One of the key aspects of [KEEL] is that it gives you 100 percent explainable decisions and actions,” Tom said, comparing it to the less thorough self-analysis that a human actor is capable of. “We would suggest that a human, even if he does something and he makes a mistake — he couldn’t explain how the brain worked and how he made the mistake.”

Simulations and automation

Tom sees various uses for the technology in automation. A car could keep track of its own internal setup and recommend repairs; an unmanned car in a dangerous territory could navigate a road and negotiate its way around boundaries; biosensors in a patient could automatically recommend changes to drug dosages and other aspects of healthcare.
The goal is to create artificial intelligence that, like a person, can be made to take into account new information and adjust course accordingly.
In addition to physical devices, AI could also be expanding in enterprise IT systems in years to come as the business environment keeps changing.
“The demands on business are changing so rapidly that the systems that support the company’s goal have to be incredibly flexible,” said Daeron Meyer, a project manager with nVisia, an enterprise architecture firm with offices in Chicago and Milwaukee. Instead of single mainframe systems that could last 10-20 years, Meyer stressed the importance of rules engines that can be monitored by people in charge to ensure corporate uniformity and compliance with shifting regulations.
“There’s certainly a lot of hubbub around self-repairing, self-healing systems,” Meyer added, but, nevertheless, systems can still be prone to hardware or software failure and would continually need human supervision to make them work properly.
AI “has not adapted as fast as people had hoped,” said University of Wisconsin-Madison professor James Gee, an expert in video games. “And some of the stuff that you would want to do with AI — for example, mimicking a real conversation — has turned out to be close to impossible to do.”
Gee points to the popularity of massively multi-player online games as an indicator of what people really want: computers that respond to them on a human level, altering to adjust to their own actions and their own thoughts and input.
If current efforts in improving AI technology do pan out, Gee says, the potential for fully functional simulations in learning would be enormous, spanning everything from medicine to politics to everyday interaction.
“Simulations that were able to do this could transform a lot of our lives because people could, in a very complex world, try stuff out before they ruin the world we’re in.”
Click here for past WTN coverage of Compsim

Eric Kleefeld is a writer for WTN in Madison. He can be reached at eric@wistechnology.com.