As Artificial Intelligence continues its impressive journey into outperforming humans, video games offer an interesting and unexpected challenge. Recently, a project aimed to train an AI to master the classic game Pokémon Red. Despite a whopping 50,000 hours of training, the digital contender was left stumbling in its digital boots.
Peter Whidden’s endeavor to train an AI in the art of Pokémon gameplay employed reinforcement learning, a branch of AI where agents are rewarded for making the right decisions. By internalizing these rewards, the AI, in theory, learns the best strategies over time. This AI, operating under the name Albert, was trained by interpreting images from the game’s screen, making decisions on button presses accordingly. However, instead of a swift and seamless learning curve, Albert’s journey proved to be a humorously bumpy one.
A quirky outcome of the reinforcement learning process was Albert’s propensity to get distracted, not unlike the typical human player. In one instance, instead of navigating through the game’s challenges, Albert found solace in admiring the digital scenery, an activity the algorithm deemed rewarding. It took tweaking the rules to pull the AI out of its self-imposed vacation.
But distractions weren’t the only challenge. Despite 50,000 hours of dedicated training, Albert struggled significantly to even understand the foundational elements of the game. For instance, the AI took over 7,000 hours just to overcome its first gym leader, Brock. To put this into perspective, human players, even novices, would typically take a fraction of that time to achieve the same feat.
The hurdles didn’t stop there. Albert, still locked in combat with the second gym leader, Misty, continues to be ensnared in a web of poor decisions, some of which seem to defy logic. One perplexing behavior exhibited by Albert was that it kept on depositing Pokémon into the PC, thereby reducing its available roster and shortchanging itself in battles.
Peter Whidden incorporated numerous incentives to guide Albert’s progress. For instance, by rewarding Albert for the discovery of novel scenes (distinct pixel changes), he aimed to instill an exploratory spirit. Unfortunately, this had unintended repercussions: Albert’s aforementioned habit of staring at the water.
The reinforcement system also faltered when it came to battles and catching Pokémon. Originally, Albert evaded every combat scenario, a strategy far from conducive to progression in a game like Pokémon Red. To counter this, Whidden tied rewards to the cumulative level of Albert’s Pokémon party. However, this change led to another quirky behavior. Upon depositing Pokémon into the PC during a healing session, Albert’s total party level plummeted, effectively traumatizing the AI and making it averse to Pokémon Centers.
So, yeah, you can only imagine just how far you can get playing any Pokémon game if you didn’t level up your party nor avoiding going to Pokémon Centers.
To make matters worse, elemental strengths and weaknesses, which is key to success in Pokémon battles – except, you know, maybe for that time Ash’s Charizard beat Gary’s Blastoise in the anime – posed yet another challenge. It was only by a chance sequence of moves that the AI managed to leverage Squirtle’s Bubblebeam attack effectively, a strategy any seasoned human player would recognize and exploit almost intuitively.
Interestingly, amidst its many missteps, Albert did showcase some remarkable abilities. The AI quickly mastered a set of moves that guaranteed the capture of a wild Pokémon with a singular Pokéball toss. This technique, known as RNG manipulation, is a complex strategy even seasoned human speedrunners spend years perfecting.
But the game wasn’t without its humorous moments. Albert exhibited an inexplicable fondness for Magikarp, arguably the least effective Pokémon. Unbeknownst to Albert, purchasing the Magikarp, again and again, which is a a recurring joke among fans.
Albert’s foray into Pokémon Red serves as a poignant reminder that gaming remains a domain where human intuition, for now, holds a distinct edge. Sometimes, even in the age of technological marvels, the simplest joys come the act of catching a machine name its captured Rattata, AI and one can’t help but marvel at the unpredictability of it all.
Speaking of Pokemon, The Pokémon Company is hinting at releasing fewer games going forward, focusing on quality over quantity.
