What happens when artificial intelligence is no longer used simply as a tool, but as a partner in human performance, learning, decision-making, and even physical recovery?
That question became the central theme of my research. Across the articles and videos I reviewed, three ideas stood out most: the growing partnership between people and machines, the changing meaning of expertise, and the possibility of using artificial intelligence to extend human abilities in practical and even life-saving ways.
Human and computer cooperation
One of the most significant ideas came from Edward Feigenbaum, who predicted that advanced human-computer interfaces would allow people and machines to cooperate on tasks with consequences far greater than what was possible at the time.
What makes this idea important is that it moves beyond the assumption that computers simply serve people. Instead, the human and the machine contribute different strengths to the same task. The computer processes information, detects patterns, and evaluates possibilities at a scale that would be difficult for a person to manage alone. The person contributes context, judgment, creativity, values, and an understanding of the larger purpose.
This prediction feels especially relevant now. Systems such as IBM Watson and DeepMind have demonstrated how artificial intelligence can analyze complex information and support decision-making. One of the defining features of artificial intelligence is its ability to learn from data and improve its performance rather than requiring a programmer to create a separate instruction for every possible task.
The result is a shift from automation toward collaboration. The most valuable applications may not be those in which artificial intelligence replaces a person, but those in which it helps a person do something better than either could accomplish independently.
Artificial intelligence and the meaning of expertise
The second major idea involved how we define expertise.
Stamper and Moore explain that expertise varies by field. In chess and music, researchers have often discussed the idea that approximately 10,000 hours of practice may be required to reach an expert level. This reminded me of Malcolm Gladwell’s Outliers, which helped popularize the same general idea.
In learning systems, however, mastery may be defined differently. A system might estimate that a learner has mastered a skill when there is a 90 or 95 percent probability that the learner understands it. This creates an interesting distinction between human expertise, which develops through experience and practice, and machine expertise, which may be measured through patterns, probabilities, and performance.
The research on game-playing agents made this issue even more interesting. Human players demonstrate expertise through recognizable strategies. In some artificial intelligence systems, a human explicitly encodes those strategies. In others, deep neural networks develop policies through repeated experience.
The problem is that these systems can become difficult for humans to interpret. They may perform at an expert level without being able to explain their reasoning in a way that a person can easily understand.
This creates an important tension. If a machine can make better decisions than most people, but those decisions come from a black box, should we describe the machine as an expert? Performance may demonstrate competence, but expertise also involves explanation, judgment, adaptability, and understanding.
Artificial intelligence may therefore change not only how expertise is developed, but how it is recognized.
Using artificial intelligence to build a better human
The third and most fascinating lesson involved the use of artificial intelligence to improve human capability.
The examples I encountered went far beyond software applications. They showed technology becoming integrated with the human body and used to improve performance, mobility, safety, and perception.
One example involved the real-time collection of data in race car driving. A system capable of tracking performance continuously can help a driver make better decisions, identify patterns, and improve reactions. Interestingly, “race car” is also a palindrome, which may be the least important but most memorable detail of that example.
The medical applications were even more striking. Choosing to undergo an experimental procedure involving amputation and an advanced prosthetic requires an extraordinary level of courage. The integration of thought, movement, and bionic technology demonstrates how artificial intelligence can help restore or expand physical ability.
Another example was C-THRU, a helmet designed to help firefighters see in environments with little or no visibility. The system uses infrared light and edge-detection technology to identify shapes that may be hidden by smoke, such as walls, furniture, or a person trapped inside a building.
This is a clear example of technology extending a human sense rather than replacing human judgment. The firefighter still decides where to move and what action to take, but the system provides information that the human eye cannot access under those conditions.
Applications like these show the potential of artificial intelligence to make people safer, more capable, and more independent.
What I want to investigate next
The research also led me to a more personal question: How can artificial intelligence support endurance training and recovery?
As someone interested in ultra-running, I am particularly curious about how artificial intelligence can help athletes make better decisions concerning training, nutrition, pacing, injury prevention, and recovery.
The more useful data an athlete can collect, the better prepared that person may be to recognize patterns and adjust a training plan. However, the challenge is not simply collecting more information. The real value lies in knowing what the data mean and how to act on it.
Several questions stand out:
- Can artificial intelligence identify signs of overtraining before the athlete recognizes them?
- Can it adjust training based on sleep, heart rate, fatigue, weather, elevation, and prior performance?
- Can it help determine nutritional needs during long-distance events?
- Can it improve recovery by identifying when rest is more beneficial than additional training?
- How are experienced ultra-runners currently using artificial intelligence and wearable technology?
I am also interested in whether the greatest value comes from applications alone or from a combination of software, wearables, sensors, and physiological monitoring.
Platforms such as AI Endurance, TrainAsONE, PKRS.AI, and Vert appear to offer different approaches to adaptive training and performance analysis. Exploring how these systems make recommendations, what data they use, and whether their guidance improves outcomes would be a useful next step.
Final thought
The most important conclusion from this research is that artificial intelligence is becoming more than a tool for completing tasks. It is increasingly functioning as a collaborator, coach, interpreter, and extension of human ability.
At its best, artificial intelligence does not make the human unnecessary. It allows the human to see more clearly, make better decisions, and perform beyond previous limitations.
The question is no longer simply what machines can do.
The more meaningful question is what people and machines may be able to accomplish together.