Two recent articles caught my eye this week. One article focused on the Fourth Generation of artificial intelligence, calling it artificial intuition. The other article explores the shift from artificial narrow intelligence to Artificial General Intelligence. In the case of artificial intuition, author Mark Gazit describes how helpful AI has become, and its ongoing limitations. Machine learning is still fully dependent on historic data. New and unknown scenarios leave data scientists helpless. Mr. Gazit suggests that in order to have true artificial intelligence, we need machines that can think on their own.
The article goes on to describe the notion of a thinking machine as artificial intuition. This fourth generation of AI enables computers to identify threats and opportunities without being told what to look for, just as human intuition allows us to make decisions without specifically being instructed on how to do so. Considered impossible not too long ago, companies like Google, Amazon and IBM are working to develop solutions, and a few companies have already managed to operationalize it. The article goes on to describe how it works, and how it can be applied. For example, artificial intuition can be used by large global banks to detect sophisticated new financial cybercrime schemes, including money laundering, fraud and ATM hacking.
The second article explores the transition from artificial narrow intelligence (ANI) to artificial general intelligence (AGI). ANI is the current iteration of AI, where applications are use case specific. AlphaGo can win at the game Go but cannot play other games. AGI mimics human intelligence and learning; learning in one domain is applied to another. The visual below (click to enlarge) uses a view of the exponential growth of computing to create a timeline for the realization of AGI, and ultimately the realization of artificial super intelligence. Per this view articulated by famous futurist Ray Kurzweil, AGI is realized by 2040.
The referenced article, authored by Gary Grossman, describes the potential for another AI winter. In this scenario, compute power does not follow the path envisioned by Kurzweil, leading to another AI winter, where expectations of the technology fail to live up to the hype, thus lowering implementation and future investment. Mr. Grossman explains that this has happened twice in the history of AI – in the 1980s and 1990s – and required many years each time to overcome, waiting for advances in technique or computing capabilities.
The article explores where we are on the AGI journey. Mr. Grossman explains that a consensus of several surveys of AI experts suggests AGI is still decades into the future. However, he describes a transitional AI, that just may exist today.