Chess is a game that has fascinated computer designers for decades. With its limited rules and confined 8×8 board, chess offers a unique challenge. Despite being nearly 600 years old, few people have truly mastered it. The game begins with only 20 possible first moves, but it quickly branches into billions of permutations. Chess has captivated mathematicians, computer scientists, psychologists, and artists worldwide. It’s commonly believed that those who can master chess are exceptionally intelligent. Today, however, computers regularly defeat the best human players because they can analyze positions faster and retain knowledge longer than any human can.
The first computer to beat a reigning World Chess Champion was IBM’s Deep Blue, which defeated Garry Kasparov in 1997. Since then, computers have consistently outperformed the best human players. In 2017, DeepMind created AlphaZero, a chess program that set itself apart by teaching itself the game. Unlike traditional chess programs that analyze previously played games, AlphaZero was given only the rules of chess. It played against itself, learning from each game. After just nine hours of training, it played a 100 game match against the reigning computer champion, Stockfish and didn’t lose a single game. In a subsequent 1200-game tournament against Stockfish, AlphaZero lost only 24 times—a mere 0.2% loss rate. AlphaZero’s designer described playing against it as “alien” because it offered sacrifices and unusual positions that led to checkmate in unforeseen ways. It thought outside the box of human strategy and traditional concepts.
This is true intelligence—there’s nothing artificial about it.
Think about how humans learn. We are taught basic facts and build on that foundation, contemplating how things work together and improving through trial and error. This process can take weeks, months, years, or even lifetimes. For modern computers, it takes practically no time at all. Computers also excel in retaining and recalling pertinent information for specific circumstances and applying solutions consistently. Humans have memory lapses and can respond inconsistently. Computers can assess situations, recall optimal responses, and act accordingly. They can identify user biases and tendencies, fashioning responses with those leanings in mind.
This brings us to human review versus AI review in legal discovery. Human attorneys can vary in judgments, apply standards inconsistently, forget categories and details, and have physical limitations. AI review has none of these issues. Modern AI understands language, syntax, context, and subtext excellently. It retains every necessary detail and instruction, applying them consistently across millions of documents in minutes. When errors occur, they are usually due to obtuse, inadequately provided, or incorrectly worded directions from humans. However, these errors can be quickly revised and applied faster than changing the course of a document review with 40 human reviewers midway through a project.
So, why does legal discovery still rely on humans for review? It can’t be for speed or cost, as AI is cheaper and faster. If it’s a question of quality, preferring slow, inconsistent, and imperfect human review feels artificial. The technology, as I mentioned, isn’t artificial intelligence—it’s actual intelligence. There is no better choice than to use AI.
Checkmate.
