What Google Translate Tells Us About Where AI Is Headed Next
1 min read
Summary
AI chatbots are incredibly popular but their fundamental limitation is data, it has exhausted most of what is available on the internet already.
The chatbots work best with a handful of languages like English and Spanish, but even then, they can struggle with idioms, technical terms and nuances, leaving room for error.
Today’s AI chatbots are trained on large volumes of data that teach them to “reason” by predicting the next word in a sentence using probabilities.
This technique was popularised in the 70s and 80s, when it was used for speech recognition and translation, the latter which has struggled to overcome errors.
Despite shortcomings, machine translation use is high as shown by the Google Translate app reaching one billion installs in 2021.
The future of AI likely involves making the best of limited data, being excellent at a few tasks, mediocre in others, and unreliable elsewhere.
