In March 2023, more than 1,000 experts in the field signed an open letter asking AI laboratories to pause the development of artificial intelligence systems more advanced than GPT-4 due to the implications they could have on human activities. Apparently without much success.
Artificial intelligence continues to evolve rapidly in recent months, to the point that experts like Demis Hassabis believe that in a few years we could develop electronic brains with intelligence similar to that of humans. There is talk about Strong artificial intelligence or artificial general intelligence (AGI) with the same understanding and communication capabilities as human beings. But, from here to the appearance of superintelligent brains there is only one step.
What is artificial superintelligence
Superintelligence refers to a hypothetical artificial intelligence that would far surpass human intelligence in virtually all fields of knowledge. This would include abilities such as processing speed, memory, learning, language, problem solving, creativity, among many others. In essence, it would be an artificial mind of the highest possible level, with the autonomy to make its own decisions, including the ability to improve itself.
Some researchers believe that artificial superintelligence (ASI) could be achieved this century thanks to the high level of scalability of the fundamental ingredients of artificial intelligence such as deep neural networks, machine learning and cloud computing. . This artificial superbrain would lead our civilization down paths that are difficult to predict.
On the one hand, a charitable ASI could help humanity solve problems such as climate change, hunger or disease. But there is also the fear that it is impossible to control an intelligence thousands of times greater and that it escapes our understanding. There is no shortage of those who warn of catastrophic consequences if the ASI’s objectives are not aligned with human interests.
Artificial superintelligence is probably the technological concept with the most unknowns and ethical challenges. But it is a debate that must remain open. Close monitoring of this field is the best strategy to ensure we develop safe and beneficial AI.