The state of A.I. Is it as understanding as human intelligence?
Language translation using Artificial Intelligence (A.I.)

Artificial Intelligence (A.I.) is a hot topic these days. Microsoft, Google and Amazon have invested heavily in this technology and continues with efforts to democratise it. This would make A.I. accessible to small businesses who don’t have funding for data scientists, or the computing power needed for intensive algorithmic processes.

As a technology A.I. requires large amounts of data to provide real value; the storage of which these technology giants have made available through its cloud datacentres. This cloud infrastructure is what provides the scale, security and computing power on a relatively inexpensive platform we as consumers need to run A.I. algorithms.

We’re seeing businesses make smarter decisions faster and increasing their profitability as a result. A.I. has also made incredible leaps in the healthcare industry, environmental sciences and anti-terrorism initiatives.

Is it intelligent enough?

Unfortunately, not yet! A.I. currently lacks contextual empathy and this makes it hard to compare to Human Intelligence (H.I.). If we want to achieve this, the focus needs to shift from ‘decoding’ to ‘understanding’. Professor Douglas Hofstadter described this very well in his article on language translation using A.I. We put this into practise and ran a similar test using Google translate.

Give it a try

Using English to Italian translation, we entered the following sentence: “In their house, everything comes in pairs. There’s his car and her car, his towels and her towels, and his library and hers.” This seems like a simple task for ‘deep learning’ A.I. to process. The results proved to be rather disappointing.

The state of A.I. Is it as understanding as human intelligence? image

For those familiar with the Italian language, the translation is grammatically correct and ‘decoded’ correctly. The deep learning algorithm however, failed to understand the context of the sentence. We wanted to communicate that both male and female counterparts have similar items. In Germanic languages like English, Dutch or German this would be simple. In Romance languages like Italian, French or Spanish the possessive pronoun does not agree in gender with the possessor but with the object.

The translator did a good job of translating the original intention to Dutch due to the nature of Germanic languages. Despite its big vocabulary, the translator lacks empathy and context. As a human I can translate the sentence correctly both grammatically and with clear intention.

The state of A.I. Is it as understanding as human intelligence? image

Here you can see how the Italian sentence should have been phrased in order to clearly communicate intent and context.

The state of A.I. Is it as understanding as human intelligence? image

Similar test scenarios were run with Microsoft Bing Translator and whilst it’s evident that some A.I. features might still be on the horizon, it's already has the capabilities for businesses to leverage today. If you’d like to learn more about the potential of A.I. in your business and where to apply it, contact us here.