⇧ [VIDÉO] You May Also Like This Partner Content (After Ad)
Scientists working at a company specializing in translation say that the famous phenomenon known as the “Singularity” will happen soon. A somewhat daring statement, since exact criteria have never been defined on the subject.
Have you heard of the concept of the “technological singularity”, or more simply “the Singularity”? This is a hypothesis, long discussed, according to which artificial intelligence could reach a point of self-improvement beyond which we would be facing unpredictable upheavals that would affect the entire human society.
The term was mentioned as early as the 1950s, but Irving John Good, a British statistician, gave a definition in 1965 that might come close to its current meaning: “ Let’s say that a superintelligent machine is a machine capable in all areas of intellectual activity to far surpass a human being, however brilliant. Since the design of such machines is one such intellectual activity, a superintelligent machine could design even better machines; there would then undoubtedly be an ‘intelligence explosion’, and human intelligence would be superseded very quickly. Thus, the invention of the first superintelligent machine is [théoriquement] the last invention man needs to make “.
In other words, artificial intelligences, once they become smart enough, would be able to improve and update themselves to constantly improve their capabilities. From this critical point, the role of the human species can be questioned.
A group of Italian scientists recently claimed that we were not far from that tipping point. To be exact, we still have seven years to go before we reach it. His statement is based, however, on facts that are quite specific to his field: translation. The company is also called Translated (“Translated”).
“ Many artificial intelligence researchers believe that solving the language translation problem is the closest thing to producing artificial general intelligence (AGI). “, we can read in a press release from this company. Their argument: language is a very natural skill for humans and, conversely, difficult for machines to assimilate. For them, the ability of machines to catch up, or even exceed, humans’ ability to translate language. could therefore be a reason to determine the status of “Artificial General Intelligence (AGI)”. In other words, a program capable of performing any task that can be performed by a human being.
“Time to edit” as a measure of AI “intelligence”
To announce this famous arrival of the singularity, the scientists of this company chose to base themselves on a particular measure, the Time to Edit (TTE). This unit of measurement represents the time it takes professional human reviewers to review AI-generated translations compared to human translations. In theory, if that time becomes more advantageous in an AI-generated translation compared to a human translation, we will reach a tipping point, according to Translated. However, the required correction time has been falling for years. If we follow this graph, in 2015 it took a human translator around 3.5 seconds per word to get an ideal translation. In 2022, it’s more like 2 seconds. And it goes downhill: hence this idea that in seven years the turning point will really be reached.
This deduction is certainly interesting in relation to the field of translation and says a lot about the evolution of the sector. However, it is still part of a very specific field, and is based on a hypothesis still to be debated: that of human language as an absolute reference to determine the “intelligence level” of an AI…