Facebook is amazed at the unexpected results offered by its artificial intelligence system

facebook artificial intelligence

There are many resources or time, with all that this entails, especially in terms of costs, that dozens of companies and research centers are dedicating month after month to such a simple and everyday topic as can be compression and development of Artificial Intelligence. I say simple and everyday because, despite the fact that humanity hardly knows anything about this subject, the truth is that little by little we have become accustomed to dealing with this type of machines, a proof of them are the virtual assistants that we have , for example, on our phones.

As practically all the researchers who dedicate their hours of work to the development and exploration in this field comment, the truth is that there is still a long way to go since we know practically nothing about the world of artificial intelligence. A very simple test of what I present is the latest project carried out by a group of researchers from Facebook where a series of results was expected and very different ones have been achieved.

Facebook Data Center

Researchers from Facebook, in one of their experiments, are totally puzzled by the results of their tests of machine learning.

Putting ourselves in context for a moment, as has been explained from Facebook, apparently the basic idea of ​​carrying out this experiment was to test their techniques of machine learning, a learning technique where an artificial intelligence system learns to perform certain tasks based on repetitive acts, that is, getting a computer to learn to perform any action based on repeating it a large number of times. Basically what they did at this point is to use this type of test so that a computer learn by himself to speak autonomously.

The idea for this test is as simple as using a series of tools to achieve that, through the recurring iteration of two chatbots that simply spoke, the resulting system learns to communicate. The truth is that this new resulting system was not intended to be capable of create a new language or anything like that, the researchers hoped that in this way they would obtain a tool capable of learning and chatting with a human interlocutor in a much faster way.

These were only the expected theoretical results since, as the title of this entry says, what the group of researchers in charge of the project really obtained was something totally unexpected How can it be that, after hours and hours of training, these conversations between the new artificial intelligence system and chatbots led to nothing less than creation of a new language.

digital network

A routine test of machine learning results in the creation of a new communication language.

Precisely because all the systems involved in this conversation began to speak in a language completely different from all the languages ​​we have seen to date, the researchers had to stop project execution and change model since they could not follow the conversation that was taking place between the different machines. Attending at this point to the statements of the person in charge of this entire project where he tells us about the conclusions reached by the whole team after carrying out this project:

The truth is that there is still a lot of potential for future work, particularly exploring other reasoning strategies, and in improving the diversity of sentences without moving away from human language

Although many may classify this test as a failureThe truth is that it is more than impressive to the fact that it, without human intervention, has resulted in the system having been able to create its own language to communicate. Undoubtedly, a new example of how we still have a long way to go before we can understand what is behind the world of artificial intelligence and how it behaves when faced with certain inputs.


Be the first to comment

Leave a Comment

Your email address will not be published. Required fields are marked with *

*

*

  1. Responsible for the data: Miguel Ángel Gatón
  2. Purpose of the data: Control SPAM, comment management.
  3. Legitimation: Your consent
  4. Communication of the data: The data will not be communicated to third parties except by legal obligation.
  5. Data storage: Database hosted by Occentus Networks (EU)
  6. Rights: At any time you can limit, recover and delete your information.