Facebook engineers shut down two artificial intelligence systems that, during a test exercise, moved away from English and started talking to each other in what appeared to be a chaotic language. But it wasn´t, it was "a new language" in which the two agents "bots" seemed to communicate more easily with each other. The complex episode is devalued by some, but for others it is a sign of the risks that artificial intelligence can have for the future of humanity.
According to the Digital Journal's description: in a Facebook computer research lab - called Facebook AI Research Lab, or FAIR - engineers were testing how two chatbots (interactive chat software) interacted between them. The purpose of the simulation was to see how they solved problems such as transacting "balls" with a certain value, and negotiating between them the best possible deal.
At some point, the conversation ceased to make sense for the engineers. Later, however, the engineers discovered that, in reality, the two bots were communicating in a way that both of them understood. It was a kind of new language, for which programmers did not contribute directly - it was the result of improvisation that the two bots were scheduled to do, but eventually took an unexpected turn. The programmers decided to shut down both systems when they realized what had happened.
This was the interaction between the two bots "Bob" and "Alice":
SOURCE: Facebook (FastCoDesign)
To a human observer, this seems to be a meaningless interaction. But in fact, as FastCoDesign writes: “But what if I told you this nonsense was the discussion of what might be the most sophisticated negotiation software on the planet? Negotiation software that had learned, and evolved, to get the best deal possible with more speed and efficiency–and perhaps, hidden nuance–than you or I ever could? Because it is.
"Agents tend to move away from understandable language and invent keywords for themselves," explains Dhruv Batra, one of FAIR scientists, noting that "there is no incentive at all to continue speaking English." The phenomenon is similar to what happens with the creation of slang, a language that departs somewhat from the original language but which helps humans to understand each other more effectively in the context of, for example, a professional or social community.
Facebook's research, from here, will want to ensure that robots stay within the common language - even because, as Mike Lewis, another FAIR scientist quoted by FastCoDesign explained, "our goal is to have bots that can talk to people ". Still, this episode may have been another landmark of what is most fascinating and at the same time scary that Artificial Intelligence can bring.
That´s why some want to apply restrictions in this research area.
Elon Musk, president of Tesla and SpaceX, argued that "people should be very concerned" about the "race" for artificial intelligence technologies. This was one of the top topics that Tesla founder spoke to at the National Governors Association summer meetings in Rhode Island, USA.
The truth is that this discussion isn´t new. As early as 2014, physicist Stephen Hawking warned that Artificial Intelligence can mean the end of the human species because it has the ability to "take off on its own and evolve by its own means, at an ever faster rate. Humans, who are constrained by slow biological evolution, will not be able to compete and will be outdated. "
In my opinion, not all technological advances are beneficial to the development of a society or the human being.
There are limits that once exceeded, are irreversible.
Thanks for reading
