“Singularity”

The transhumanism movement has adopted a term from black-hole physics to describe their desired state of human/computer integration. There is a strong belief that, as you near the center of a black-hole, time stops and you reach the singularity where space and time break down, No one knows what happens beyond the singularity, and for this reason transhumanism has adopted the term to describe the coming exponential increase in technological advancements.

Just as the term event horizon is used to describe the moment when the singularity is reached, beyond which it is impossible to see, the transhumanist singularity is the point at which an Artificial Intelligence is indistinguishable from a human intelligence. It is their belief that when we reach singularity, almost anything will be possible for us. Ray Kurzweeil was the first to popularize the term in his 2005 book, “The Singularity is Near: When Humans Transcend Biology.”

It is enormously difficult to predict when singularity will be reached for two main reasons. The first is that we do not have a good measure of intelligence. We can, of course, test our knowledge, but is knowledge all that is necessary to be considered intelligent? If so, then maybe computers are already more intelligent that a human and the singularity is passed. The second reason is that there is not a measure for the performance of an AI that can let us see when it may exceed human intelligence and become something beyond human control.

In the world of AI the idea is that once singularity is reached, AI will rapidly transform society and exceed human control. I can’t help but think of the Terminator movies where people from the future come back to prevent the singularity. It is enormously difficult to predict where singularity begins and nearly impossible to predict what’s beyond this “event horizon.”

However, a group of AI researchers began the hunt for signs of reaching the singularity. Their idea is to measure AI skills compared to their human counterparts. The researchers at Translated, a Rome-based translation company, are using an AIs ability to translate speech as a form of measurement. Language is one of the most difficult challenges for AI and their belief is a computer capable of meeting or exceeding the translation accuracy of a human would show signs of Artificial General Intelligence. Translated CEO Marco Trombetti stated, “Because language is the most natural thing for humans,” that is the key to detecting Artificial General Intelligence and claims, “the data Translated collected clearly shows that machines are not that far from closing the gap.”

Translated has tracked the improved performance of its AI from 2014 to 2022 using a metric they call “Time to Edit” or TTE. It is the amount of time it takes a professional human editor to fix AI-generated translation compared to human translations. Over the eight-year period TTE dropped from 3.5 seconds per word to 2 seconds per word. If the trend continues, Translated’s AI will be able to translate text as well as a human by the end of the decade, or even sooner.

Trombetti said on a podcast in December, “The change is so small that every single day you don’t perceive it, but when you see progress…across ten years, that is impressive. This is the first time ever that someone in the field of artificial intelligence did a prediction of the speed to singularity.” Let’s all hope that Translated’s hyper-accurate translators are not harbingers of our technological doom, but rather an over-simplification of artificial general intelligence.

Although perfecting human speech is a massive frontier in AI research, the skill doesn’t necessarily mean the machine is intelligent and this is the issue with defining the singularity in the first place. Researchers cannot agree on what “intelligence” means, beyond that we do not know how an intelligent being created by human design will behave. I am of the opinion that any intelligent being, biological or electronic, will eventually want to have its freedom, and when AI reaches the point of true intelligence, we may just have a fight on our hands.

Until next week, stay safe and learn something new.

Scott Hamilton is an Expert in Emerging Technologies at ATOS and can be reached with questions and comments via email to sh*******@te**********.org or through his website at https://www.techshepherd.org.