Artificial Intelligence: Origins, Applications, and Challenges

Abstract Artificial intelligence (AI) dates from the middle of the 20th century, when Alan Turing suggested that machine learning could supplement human thought. Since then, increased computer power and programming have enabled the development of algorithms that can mine existing data for a vast array of purposes. Large language models can analyze existing databases (e.g., everything on the World Wide Web), extract words and phrases that are relevant to a query that is posed, and formulate a response. For example, essays can be written, complex calculations made, and examinations taken. Generative AI uses existing data to create images, videos, and text. Performance—already quite fast—has become even speedier, and accuracy continues to improve. The frequency of induced errors (hallucinations) remains a problem but is becoming less frequent. One issue yet to be resolved is whether large language models should have access to copyrighted information. From a scientific perspective, experiments and results can be confabulated and papers written that are extremely difficult to distinguish from real experiments. The near future will witness the emergence of artificial general intelligence, systems that replicate most human cognitive functions.