Did you ever imagine living in a world where playing a perfect game of Pac-Man could lead to innovations that could impact everyday life for everyone on the planet? Well, it’s happening. A team of Microsoft researchers used a system featuring AI to achieve the maximum score on the popular 1980s video game, and the implications are much further reaching than they may appear on the surface.
The Ms. Pac-Man Experiment
The Maluuba team, a recent deep learning focused acquisition by Microsoft, used an AI-based method called reinforcement learning to get the perfect score. And what they learned along the way could help similar solutions complete more complex tasks designed to support human capabilities. The approach took the puzzle that was the Ms. Pac-Man game and broke it into more manageable pieces. These were distributed among multiple AI agents, allowing them to work together towards a common goal.
The distributed approach allowed AI-based operations to mimic how some believe the human brain functions. Instead of tackling the problem in its entirety, the system portioned out the puzzle. Then, each piece was worked in an effort of reaching the primary goal: achieving a perfect score.
Instead of using the more traditional supervised learning approach where AI’s are fed examples of positive and negative behavior to shape their actions, the experiment used reinforcement learning. This method involves only providing high-level feedback and allows the system to improve based on using trial and error and getting input based on the effectiveness of the results.
In the business world, reinforcement learning may allow AIs to participate in growing sales. They could potentially determine which current customers to target with specific actions, like contract renewal marketing, at particular times of the day. Each agent oversees their piece of the puzzle and provides a determination to a top agent, who analyzes the recommendations and makes a final decision.
By using AI in that fashion, sales professionals can focus on tasks that are deemed most likely to succeed by the system. This may increase the chances of actually making a sale since the professional’s efforts are entirely pointed at identified targets that are more inclined to make a purchase.
This approach also has a significant amount of potential value for other complex tasks, like language processing.
Currently, these developments are in the fledgling stages, as the potential has been identified and tested, but real-world applications aren’t fully developed. Over time, these technologies could allow companies to automate more tasks, taking tedious work out of the hands of employees. Workers can then concentrate on high-value activities that actually require their input to succeed.
If you’re interested in learning more about how AI is going to impact the everyday lives of people in the future, the skilled professionals at Solving IT can provide you with valuable information about these emerging technologies. Contact us today to see how our services can help your business succeed.