The idea of the “singularity” looms large in the world of artificial intelligence (AI). This elusive concept describes the moment when AI surpasses human control and rapidly transforms society. The tricky thing about the AI singularity, and the reason for this is to borrow terminology from black hole physics, is that it is extremely difficult to predict where it starts and almost impossible to know what lies beyond this technical “event horizon”. Is.
However, some AI researchers are looking for signs that the singularity has been reached, as measured by AI progress approaching human-comparable abilities and capabilities. One such measure, defined by Translated, a Rome-based translation company, is the ability of AI to translate speech with the accuracy of a human. Language is one of AI’s toughest challenges, but a computer capable of bridging that gap could theoretically show traits of Artificial General Intelligence (IAG).
“That’s because language is the most natural thing for humans,” said Translated CEO Marco Trombetti at a conference in Orlando, Florida, in December. “However, the data collected by Translate clearly shows that the machines are not far from closing the gap.”
The company tracked the performance of its AI from 2014 to 2022 using a metric called “time to edit,” or TTE, which calculates the time it takes professional human editors to proofread AI-generated translations, compared to humans. Is. Over that 8-year period and analyzing over 2 billion post-edits, Translated’s AI showed slow but undeniable improvement as it closed the gap on human translation quality.
According to Translated, on average it takes about one second for a human translator to edit each word of another human translator. In 2015, it took professional editors about 3.5 seconds per word to review a machine-translated (MT) suggestion; Today that figure is only 2 seconds. If the trend continues, Translated’s AI will be as good as human-generated translation by the end of the decade (or even earlier).
The Singularity in Artificial Intelligence: Is It Going to Be Real?
“The change is so small that every day you don’t see it, but when you see progress … over 10 years, it’s amazing,” Trombetti explained in a podcast in December. “This is the first time that anyone in the field of artificial intelligence has predicted the motion toward the singularity.”
Although this is a new way of measuring how close humanity is to a singularity, this definition of singularity runs into similar problems when it comes to recognizing AGI more broadly. While the refinement of human speech is certainly a frontier in AI research, this impressive ability doesn’t necessarily make a machine intelligent (not to mention that many researchers don’t agree on what “intelligence” is). ,
Whether or not these ultra-precise translators are harbingers of our technological doom doesn’t diminish Translate’s AI achievement. An artificial intelligence capable of translating speech as well as a human could transform society, even if true “technical excellence” remains elusive.
