Facebook’s team of AI has tried a new approach to teach chatbots how to chat like humans. The FAIR (Facebook Artificial Intelligence Researchers) tested an exclusive 164,000-utterance dataset named Persona-Chat for the bots to identify patterns of the communicative aspect of us humans.
According to a report received today, January 29, this data set involves over 160,000 lines of dialog from the Mechanical Turk (MTurk) marketplace of Amazon. MTurk, being an online crowdsourcing marketplace, provides a platform for people and businesses to coordinate on human intelligence and perform tasks that present-day computers are yet to learn.
The concerned test of Facebook was applied to neural networks that are used for the chatbots. An entirely new group from the MTurk assessed the results who carried out conversations with a persona-driven bot. Then, the outcome was compared with other Facebook AI bots and humans.
The persona-driven bot wasn’t up to the mark in terms of fluency and consistency as compared to the humans. Although, it did outperform the chatbot that was trained on movie dialog. This further opened potential in infusing human-like nature to the bots, which can even give rise to controversies considering future possibilities.
Last year, Facebook forced itself in shutting down one of its AI systems after the bots started communicating between themselves in a new language that was deviated from the provided codes. The recent test thusly carries weight, and it must as well be noted that the persona-bot could pick up our natural ways of communication better than scripted dialog-throwing.
The test from the last year showed that the Facebook AI chatbot or ‘dialog agents’ started creating their own language, while the FAIR team was busy improving the chatbots. These dialog agents were using Machine Learning algorithms to create a new language without human input, and they even turned out to be ‘incredibly crafty negotiators,’ as observed by the researchers.
Is the day finally coming when robots will rule over us? Or shall we be able to control their outgrowth before the ultimate danger? What do you think, Dear Reader? Share your views in comments below. Stay tuned with us for our coverage of the next development on this matter.
