- With just one measure, we could approach technological singularity by the end of this decade, if not sooner.
- A translation company developed a metric, Time to Edit (TTE), to calculate the time it takes professional human editors to correct AI-generated translations compared to human translations. This can help quantify the speed towards the singularity.
- An AI capable of translating speech as well as a human could change society.
In the world of artificial intelligence, the notion of “singularity” occupies an important place. This slippery concept describes the moment when AI surpasses human control and rapidly transforms society. The difficulty about the singularity of AI (and why it borrows 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 technological “event horizon.”
However, some AI researchers are looking for signs of achieving the singularity, measured by advances in AI approaching human-like skills and abilities.
One such metric, defined by Translated, a Rome-based translation company, is an AI’s ability to translate speech with the accuracy of a human. Language is one of AI’s toughest challenges, but a computer capable of bridging this gap could theoretically show signs of artificial general intelligence (AGI).
“This is because language is the most natural thing for humans,” Marco Trombetti, CEO of Translated said at a conference in Orlando, Florida, in December 2022. “Nevertheless, the data collected by Translated clearly shows that machines are not that far from closing the gap. »
The company tracked its AI performance from 2014 to 2022 using a metric called “Time to Edit” or TTE, which calculates the time it takes for professional human editors to correct AI-generated translations by compared to human translations. Over this period of 8 years and analyzing over 2 billion After the changes, Translated’s AI showed slow, but undeniable improvement, slowly closing the gap towards human-level translation quality.
On average, it takes about a second for a human translator to edit each word of another human translator, according to Translated. In 2015, it took professional editors approximately 3.5 seconds per word to verify a machine-translated (MT) suggestion. Today, that number is only 2 seconds. If the trend continues, Translated’s AI will be as good as human-produced translation by the end of the decade (or even sooner).
“The change is so small that you don’t see it every day, but when you see progress…over 10 years, it’s impressive,” Trombetti said. on a podcast. “This is the first time anyone in the field of artificial intelligence has predicted the speed towards the singularity.”
While this is a new approach to quantifying how close humanity is to the singularity, this definition of singularity runs into similar problems. to identify AGI more broadly. And while perfecting human speech is certainly a frontier in AI research, this impressive skill does not necessarily make a machine intelligent (not to mention the number of researchers I don’t even agree on what “intelligence” is).
Whether or not these hyper-accurate translators are harbingers of our technological catastrophe, that doesn’t take away from Translated’s achievements in AI. An AI capable of translating speech as well as a human could very well change society, although the true “technological singularity” still remains elusive.
Darren lives in Portland, has a cat, and writes/edits about science fiction and how our world works. You can find his previous work on Gizmodo and Paste if you look hard enough.