ARTIFICIAL GENERAL SUPER-INTELLIGENCE COULD ARRIVE BY 2027, SCIENTIST PREDICTS

(← click) Unless you've been asleep for the past few years, you've likely heard lots of buzz about "AI" – Artificial Intelligence. A little less familiar to the general public is the term "AGI" – Artificial General Intelligence. The first, AI, is collecting huge amounts of data on a given subject, such as programming, or physics, or medicine, then loading up a large database with this data in a structured way according to rules of correct grammar and syntax. Such databases have been around for several years for the legal and medical professions, and now, with voice input and output, you can simply ask a question in ordinary English and hear or read the answer, usually quite correct and exact... but it can "hallucinate" or put the data together incorrectly so that the output may sound plausible but is just plain wrong.

But the second term, AGI, is newer, and means collecting even huger amounts of data on a multitude of fields and train the machine with more refined rules so that the output is very precise on any of these multitude of fields and matches the best of expert knowledge of any one human individual in any of these fields, "general" knowledge on many, many topics. But the newest term that's just now popping up is "AGSI" – Artificial General Super-Intelligence. This is such a vast collection of data on every conceivable field of knowledge – virtually all that has been recorded over the course of human history – combined with rules of logic, grammar, and syntax in a multitude of languages, so that you can ask it anything in almost any language and it will reply in that language and in a way superior to any human being in any field of expertise whatsoever. Many AI experts have stated this is decades away, some have predicted it within a decade or so, but the above headline says it's just around the corner, in about 18 to 30 months. Here's a summary of the above article:

"A leading AI expert, Ben Goertzel, predicts artificial general intelligence (AGI) could arrive as early as 2027, evolving rapidly into artificial superintelligence (AGSI or ASI). Goertzel believes human-level AGI is plausible within the next 3-8 years, potentially leading to a rapid intelligence explosion. Others, like Shane Legg and Geoffrey Hinton, have made similar predictions about the timeline for AGI, albeit with varying degrees of confidence. Goertzel emphasizes that recent advancements in large language models (LLMs) like ChatGPT have increased the possibility of AGI, but LLMs alone won't create it. Goertzel speculates that once AGI reaches human level, it could quickly become radically more intelligent, potentially leading to a singularity or intelligence explosion. The prediction comes with caveats, noting that AI's 'mind' might not be similar to a human mind, and the technology's development may not follow a linear, isolated path."

Another article, "AGI TO ARRIVE 'MUCH SOONER' THAN IN 2-3 YEARS: SOFTBANK’S MASAYOSHI SON," by the head of Softbank, a leading AI firm, basically implies that AGI may be already here, kept in wraps somewhere. During my time in military intelligence 60 years ago, we were using computers that the outside world didn't even know existed until several years later when IBM introduced them to the business world. So I can easily surmise that right now the government and the intelligence community already have AGSI, they already know almost everything about everything and everyone in the world. What's going on here?

Get the full article on our ARC-News blog. And create your own blog for less than $1.99 per month!