The AI industry is getting 'scary good' - here are four ways AI can outperform humans

Historically, humans have spent decades or even centuries developing perfect strategies for strategy games using artificial intelligence.

Scientific American reported IBM researcher Murray Campbell described chess as 'a game that requires strategy, foresight, logic-all the qualities that make up human intelligence.'As part of the Deep Blue project, Campbell and his team developed a six-foot supercomputer that defeated Garry Kasparov in the 1997 World Chess Championship.

During the pivotal final game, Kasparov claimed IBM cheated, which the company denied. An early move that caused Kasparov to lose confidence in Deep Blue was the result of a bug that caused the computer to choose a move at random.

Kasparov may have lost the tournament because he overanalyzed a 'last-resort fail-safe' move, according to Nate Silver's book. As in chess, both players have access to all of the action unfolding on the board - one player could mislead the other into a mistake, but both can see and assess every move.

An artificial intelligence model can benefit greatly from the complexity of Texas Hold 'Em, which is a card game with random draws, hidden information, and deception.

A team of Carnegie Mellon University researchers developed Libratus, a poker robot that defeated four of the best over 120,000 hands. Engineers improved their self-improving poker-playing AI, Pluribus, in 2019. Forbes reported that Pluribus can learn from its previous moves and act accordingly.

The first time an AI program beat multiple human players, Pluribus beat five other humans at one table. Professionals will study the calculations made by AI models and the poker community has to keep up with banning players from using AI in online games.

The AI robot dog Morti learned to walk quicker than a human. 'Our robots are practically born with no understanding of their leg anatomy or function,' Felix Ruppert, co-author of the study, told The Independent.

Central Pattern Generators (CPGs) in the spinal cord control the movements of limbs and muscles as people and animals become independent. To simulate a naturally occurring CPG, Ruppert and Badri-Sprowitz built their 'walking intelligence' system on a computer.

Morti's 'virtual spinal cord' would measure a step's pressure and compare it with CPG predictions. Morti's legs would swing like a pendulum if it fell, or it would touch the ground slower or faster if it fell.

Since humans are less advanced in developing their intellect at birth, it takes about a year for newborns to learn how to walk. As a demonstration of an artificial intelligence program powering motion, Morti differs from game-playing bots, potentially resolving robots' history of clumsiness.

In American schools, reading comprehension is a staple assessment. It is estimated that reading is nearly twice as efficient as listening, and it could be argued that understanding what you read and being able to recall it is the most important trait in learning.

A Microsoft Research Asia AI bot that reads and understands as well as humans were developed in 2018. Microsoft's model was tested against the Stanford Question Answering Dataset, which consists of reading comprehension questions based on Wikipedia articles.

In a ScienceDaily study, researchers found that humans are biologically designed to interpret letters and numbers, regardless of whether they play chess or poker.

If you want to test the model's understanding of your text, you can plug in your text on Microsoft's site. These are all examples of Artificial Narrow Intelligence - systems that are programmed to excel at one task.

Eventually, artificial general intelligence will provide computers with the ability to perform any task as well as a human, including reasoning, deception, and abstract human abilities not purely computational.