11 february 2019
Neural networks will easily lie to readers by posing as human beings

Neural networks will easily lie to readers by posing as human beings

The Life ASAPA language forecasting models have every chance of being the greatest thing since Bitcoin was invented.

In 2019, neural networks will be able to write a meaningful article using only one introductory sentence, as well as translate texts, answer questions and understand speech. And we would not be particularly surprised if, in the near future, new versions would be able to write poems and codes, create music and interact as equals partners with readers in chat rooms and blogs.

The goal of creating the first version of the neural network was clear and simple to predict the next word, taking into account everything that was written above in the text. The new improved version has been given greater flexibility. It is able to generate texts that are the closest to human texts in terms of quality and style, provide clear answers to questions and translate phrases into all languages. In the future, the model can be improved, and then there is a high probability that readers will accept the robot as a person with above average intelligence.

The neural network adapts easily to the style and content of the text and “writes” realistic passages of quite high quality. But so far, not everything is perfect. Neural network developers are faced with various disruptions.  The model sometimes repeats itself, writing about events that cannot happen in principle (fire under water, for example). And the transition from one topic to another is also limping, so it takes at least 4-6 attempts to write a really good article.

If a neural network writes on popular light topics such as news, music and entertainment, half the time a good result is achieved. But the technical and narrowly targeted articles are still not very good, roughly speaking.

The neural network is much better at translating and answering simple questions. And since these tasks relate to language modeling, the neural network is likely to become more humane and accurate in the future, when the amount of data increases.

The Life ASAPA neural network opens up ample opportunities not only for creating voice bots and text assistants, but also for generating a digital copy of a person. That is why we will continue to follow the development of these neural networks – they are likely to surprise us more than once.

Subscribe to our newsletter

By clicking the button, I accept the terms of the Offer for the use of the site and agree with privacy policy