It is clear that artificial intelligence is developing at an alarming rate. Even as we do this, algorithms around the world are writing code to improve themselves. And while we marvel at the speed of machine learning innovation today, the technological singularity may be closer than we expected.

This is according to AI translation startup Translated, which is studying how artificial intelligence approaches language models and predicts how fast it is accelerating toward a technological singularity (Translated (opens in new tab), via Popular Mechanics (opens in new tab))

The technological singularity is the moment when technology begins to reproduce itself out of control; it is at this point that both science fiction writers and scientists conclude that humanity will essentially become obsolete. Just as the event horizon of a black hole obscures the singularity in physics, it is extremely difficult to predict exactly when, where, and in what form the technological singularity will occur.

Watching the evolution of AI (open in new tab) over the past year has been a roller coaster ride. When it comes to translation, the challenges are immense, even for the best machine learning algorithms.

Recognizing this fact, Translated set out to track and measure how much the Time to Edit metric (TTE) of its translation algorithm had improved over the past eight years TTE is essentially the time it takes a human editor to correct an AI translation The TTE is the time it takes for a human editor to correct the AI's translation. Comparing the TTE of its algorithm to the TTE of human translations, the company noticed that the gap was slowly but surely closing.

With this trend in mind, the company concludes that its AI will likely match the level of human translators within a decade, with a technological singularity likely to follow.

"We all understand that we are approaching a singularity in AI," says Marco Trombetti, CEO of Translated.

"For the first time we can quantify the speed at which we are moving toward singularity."

While it is true that modern technology is stepping into serious Torment Nexus (open in new tab) territory, there are thousands of aspects of the technological singularity to consider. Therefore, it is hard to say whether improvements in language modeling can be predicted to exceed the event horizon of AI. [Whether the ability to understand human language is an indicator of intelligence is another matter entirely.