The Secret Language Of Facebook Bots

The news is of those cut to measure to feed fantasies of catastrophes & plots, and it has not by chance become “highly viral” flying – as always in these cases without making distinctions between reality and fantasy. In short, here’s what we’re talking about:

#artificial intelligence: two systems of A.I. license plate Facebook have begun to speak to each other in a language incomprehensible to the a man; the programmers of Facebook have therefore \”switched them \” before they could

The news has been taken up by many and more or less in this form.

Don’t panic. The story is suggestive and certainly not lacking ideas for science fiction lovers. But, titles to effect apart, what really happened in Menlo Park’s labs? Did we really risk the technological apocalypse?

To remove the doubt, just take a look at the technical blog of Facebook.

For some time now the Zuckerberg company has been working on the creation of bots capable of dialogue with man within a chat and conducting simple negotiations. These software use artificial intelligence mechanisms to learn from the interlocutor and become more and more good at concealing artificial origin.

A excerpt from chat-bots conversation: it is not that the intelligence of autonomous conversation is quite evident, nor the originality of the language, but researchers say that it is so (we believe…).

Copying is learned. Using widespread practice in developing AI systems, Facebook researchers decided to have two of these programs dialogue with each other so that they could learn from each other.

Specifically, programmers have instructed “bob” and “alice” chatbots to make their own arrangements to divide a hat, two books and three balloons.

At some point, however, the conversation between the two machines took a strange turn:

Bob: ♪ I can I everything else ♪

Alice: ♪ Balls have zero to me to me to me to me to me to me to ♪

The two bots have started to make use of words and grammar rules other than ours, for us incomprehensible but for them (seems) very clear. What happened?

The researchers had not imposed on the software to use exclusively the English language to conduct the negotiation. And so the AI system has discovered and used a more efficient method than conventional one to use words. In fact, he invented a new language.

No use of intelligence. The programmers then decided to “turn off” the machines, and to re-activate them only after setting English as the only language of communication.

Not so much because the new dialect would allow chatbots to exclude man from their conversations (taking over?), but because it would have been completely useless for a software-function intended for interaction with human beings.

Deep Thought meditates with “the head in your hands” on life, the universe and all of it. As you know, your answer will be “42.”

No one has ever been “in danger” and no computer has ever “thought” to self-organise itself against human beings. In fact, the system has only used words (for him) more rationally, that is the only tools that had been made available to it to achieve the goal.

It’s important to understand each other. The interesting thing, Mark Liberman points out from the computational linguistic blog of the University of Pennsylvania, is that the new language has still allowed bots to reach an agreement on the sharing of objects. This means that both machines were able to understand it.

However, this is not the first time that an artificial intelligence system has developed a new language. In April, researchers from OpenAI, Elon Musk’s artificial intelligence laboratory, published an article explaining how their bots “invented” a linguistic adaptation of rationalization of the current English language, eliminating unnecessary and redundant elements.

Let them talk. Can allowing computers to develop their own language have some practical usefulness? Absolutely, Dhruv Batra, visiting research scientist from Georgia Tech University at Facebook AI Research laboratories, is convinced.

10 PHOTOS Photogallery The artificial intelligence you use every day GO TO THE GALLERY

Photogallery Artificial intelligence that you use every day Virtual maintainers

They’re called Siri, Google Now, Cortana. Very kind and always available answer our voice commands to remind us of a commitment, to search the web, to send a text or to answer a phone call. ♪ Remind me to call Maria when I arrive at the office ♪ ♪ Add the peppers to the shopping list ♪ ♪ Tell me how to get to the station ♪ ♪ They are just some of the requests

These software use artificial intelligence algorithms for language recognition but also to learn our habits and preferences over time and meet our needs better and better. Video games.

Video games are one of the most widely used applications of artificial intelligence. The algorithms of AI allow characters, environments, stories to evolve according to the behavior of the player, creating situations that are always new and unpredictable.

Among the games that make wide use of artificial intelligence are the first-person shooters like Far Cry or Call of Duty, where enemies react to the player’s maneuvers and moves with the goal of eliminating it. Customer service

Many online services today use automatic systems to manage customer support chats. These applications, called chatbots, are based on advanced artificial intelligence systems that can understand what the customer is asking them, what is their problem and provide the most appropriate answer.

The real challenge for these systems, as well as for virtual assistants, is to understand the natural language of the human being that is in front of the screen. Adv Crime

Our cities, airports, stations are controlled by thousands of cameras. These images are examined in real time by human operators helped by powerful software that can recognize not so much suspicious faces as behavior patterns that can be an alarm signal.

With appropriate training, these programs, which manage to process thousands of images per second at the same time, may find out if someone leaves a suitcase at an airport, carries a weapon or is walking around suspiciously near a school. Foto: © Tom Blackwell / Flickr Prevision of purchases

The bigs of e-commerce could further increase their already large gains if they could predict the needs of consumers. Several companies including Amazon are working on AI algorithms that can understand what we will need in the immediate future before we realize it ourselves.

If we buy a science fiction best seller every month, we’ll probably be interested in what’s coming out next month. Then why not preview it, maybe at a discounted price?

But Amazon, selling practically everything from electronics to food, has a huge amount of data on the tastes and consumption patterns of its customers: an excellent basis for allowing electronic brains to process and guess our needs. Prevention of fraud

Even if we don’t notice, we’re very accustomed, even in shopping. We go to the same shops, we send the same figures more or less. If we do something very different from the usual, the banking system will notice. And it’s very important for our security, because this kind of surveillance is used to avoid fraud against us, for example improper uses of our credit card.

If we just made a purchase with the supermecato ATM under the house and after 10 minutes our card is crawled into a New York store, we probably have to worry. The AI systems that oversee the world of banking transactions are trained to notice all these inconsistencies and to report them to the banks that intervene avoiding us being cheated. Foto: © Sally Munkberg / Flickr Adv Digital journalists.

Trembled journalists: you have electronic colleagues who write news for you. Already today. And they have much less demands in terms of holidays, breaks and salary. Some online news sites including AP, Fox and Yahoo!, in fact, use automatic systems to write short financial or sports cut news.

At the moment these systems are not able to write articles for further information or news but they can without too much trouble to face the summary of a sporting meeting or the day financial on the stock exchange. Foto: © Arthur Caranta / Flickr Smart Home

One of the most interesting applications of artificial intelligence is home automation. These are all systems that manage the environments in terms of temperature, lighting, sound according to our habits and preferences.

Thermostats like Google’s Nest that are able to understand how many people there are a room, but also electric networks that optimize the operation of appliances so as to take advantage of the best energy rates. Or electric shutters connected to our smartphone that close themselves when we leave home and open up again when we return. The only limit is imagination. Smart car

The latest generation cars keep traffic under control around us, anticipate us on braking in case of sudden slowdown and, in the case of Tesla’s utopialot, they are also able to follow the lanes of the road alone. These cars are able to make decisions based on what happens in the surrounding environment.

To do this they use artificial intelligence algorithms similar to video games that learn and modify their behavior with experience. Like humans or almost. Adv Recommendation systems

The strength of artificial intelligence systems is to learn from experience. For example, they can learn what we like to watch on TV, what we usually listen to on the radio or order at the restaurant. Once we discover our tastes AI algorithms can give us suggestions for consumption and purchase in line with our tastes.

If we listen to songs with long guitar riffs, for example, we will probably be willing to listen to other songs, perhaps new or emerging authors that we are unknown to but with that characteristic.

Who uses them? Virtually everyone: Netflix to recommend movies; Spotify to signal music; even Facebook: the algorithm that chooses which posts to show us is based on experience and is based on an AI engine. In Depth Innovation Musk, Zuckerberg and Artificial Intelligence Amazon Innovation: in 10 years animal verse translators Digital Life Artificial Intelligence creates recipes from the photo Digital Life The future of fake news is

In an interview with Fast Co Desing Batra he explains how the use of a language developed by machines would allow us to better understand their operation and develop more efficient software, written in their idiom.

Leave a Reply

Your email address will not be published.

You May Also Like