What AI can -and cannot- do? | Theory

Tags

Top Highlights

• Among the many capabilities of AI, we can list the ability to make predictions and inference, recognizing patterns in data, optimizing and automating arduous processes.

• Although they are similar tasks, predictions are about forecasting something yet to happen, such as a high-precision weather forecast.

• Whereas inferences focus on determining a target output based on a set of inputs known as predictors, for instance, suggesting a book you may like based on your preferences as a customer.

• Other pattern recognition tasks made possible by AI include: Clustering: used in applications like customer segmentation for the discovery of groups of data with similar characteristics. Anomaly detection: used in security and finance domains to identify transactions or other data occurrences that deviate from the normal. And data generation or Generative AI,

• AI can also solve plenty of optimization problems, consisting in finding the best and most efficient possible solution given a number of constraints

• Examples of optimization problems solvable by AI include: finding the optimal route in logistics and delivery services; efficiently operating and controlling energy grids; building a dynamic pricing strategy in travel bookings that maximizes revenue; and launching discount campaigns to increment product or brand sales.

• Automation itself is not AI, as it does not involve mimicking human intelligent processes, but AI can help improve automation, improving the efficiency of processes that would be time-consuming when performed by humans,