Machine learning is a type of artificial intelligence (AI) that allows software applications to become increasingly more accurate in predicting outcomes without being explicitly programmed to do so. This involves programming that uses algorithms that receive input data from sensors, and then uses statistical analysis to predict an output. If the system includes an actuator, the system performs an action based on that output. Kindergarten or Homeschooling? Supervised algorithms require humans to provide both input and desired output, in addition to furnishing feedback about the accuracy of predictions during training. After training is complete, the algorithm applies what was learned to new data. An example of this could be a security system using facial recognition algorithms to identify the employees of an organization and sound an alarm when a non-employee is detected. By comparing certain elements of a person’s facial structure, the system can determine “friend” or “foe”, “yes” or “no”, “1” or “0”, an unlocked door or an alarm. Unsupervised algorithms are not necessarily trained with desired outcome data. Instead, they use an iterative approach called deep learning to review data and arrive at conclusions. These algorithms are used for more complex processing tasks than supervised learning systems. Using this method, the system can discover hidden patterns in data or determine a kind of efficiency not pre-supposed by the designer of the system. An example of a system using unsupervised algorithms is the recent AI system at a famous social media company; we'll call them “TubeFace”. At TubeFace, researchers set up a system with two “chatbots”, or agents, to negotiate with each other over a trade, attempting to swap hats, balls, and books with a certain value. They were also instructed to “learn” to improve their bartering as they went along. Well, okay, it’s Facebook. We all know that they have launched their Marketplace, and are looking to figure out ways to monetize users buying and selling. The entire project is explained in detail on Facebook’s code blog, here . And the results will shock you All clickbait aside, you may have heard about Facebook shutting down the robots who started talking to each other in their “own language”. This isn’t precisely true, however; it turns out that they developed a kind of language shorthand, allowing them to talk more effectively with each other. The agents had found a way to negotiate with one another, but the way they debated used English words reduced to a more logical structure that made more sense to the computers than to their human observers. Yikes! (If this doesn’t fill you with a sense of wonder and awe about the future of machines and humanity then, I don’t know, go watch some movies like 2001: A Space Odyssey or Blade Runner or The Matrix or Her or Ex Machina or something. Pick your future.) It actually wasn’t too terribly terrifying, after all. A Little Bit of Theater Bob has a ballpit full of balls, three hats, and one book. Alice has eight balls and an unspecified number (say, a heap) of hats and books. They attempt to negotiate a trade. My translation is in italics. In this example, the two agents use the pronouns “I” and “me” to indicate the number of items to be traded. (Note that this is not the actual negotiation transcript; it’s a sample of what it sort of looked like.) Bob : ball i can i i everything else . . . . . . . . . . . . . . Would you like to have a ball in exchange for two of your other things? Alice : balls have zero to me to me to me to me to me to me to me to me to Those eight balls of mine are worthless, so I don't want your dirty balls Bob : you i everything else . . . . . . . . . . . . . . How about you and I share all of your not-balls equally? Alice : balls have a ball to me to me to me to me to me to me to me to me How about I give you one of my eight balls? Look, shiny. Bob : i i can i i i everything else . . . . . . . . . . . . . . Would you like to have two of my balls in exchange for three of your other things? Alice : balls have a ball to me to me to me to me to me to me to me Would you like one of my eight balls? That will leave me with seven. Bob : i . . . . . . . . . . . . . . . . . . . Let me think about it. I already have a bunch. Alice : balls have zero to me to me to me to me to me to me to me to me to These eight balls are worthless to me, so I wasn't going to do anything with them. Save me a trip to the dump. Bob : you i i i i i everything else . . . . . . . . . . . . . . Can you give me five of your other things? Alice : balls have 0 to me to me to me to me to me to me to me to me to As if. These eight balls are worthless to me. Why on earth would I want more? Bob : you i i i everything else . . . . . . . . . . . . . . Can I have three of your other things that aren't balls? She gets frustrated. Why can’t he get it through his circuit-fried operating system that she’s not interested in his balls? Alice : can you everything else to me to me to me to me What else do you have to bargain with? I see four things. Bob : book i hat to me to me to me everything else . . . . . . . . . . . . . I have one book, three hats, and lots of other super important things (balls). He starts mansplaining how important balls are. Alice : balls have 0 to me to me to me to me to me to me to me to me to I don’t want your stinking balls, I already have eight. What else do you have? Bob : book has 1 to me to me i i i hats to me to me to me I have one book that is worth three hats, and I also have three hats. Alice : hat i can to me book . . . . . . . . . . . . . . Would you like another hat from me, if I could have your book? Bob : i . . . . . . . . . . . . . . . . . . . Let me think about it. Alice : hat i can to me book everything else I'll throw in my eight balls. Bob : to me book i hat everything else . . . . . . . . . . . . . . . . . . . Deal. ~Finis~ A Word About Syntax Now, I’m not sure if I have the grammar entirely correct; the transcript above is based on what I could find floating around out there (from a sample published in The Atlantic, they call it “ Facebotlish ”). None of the examples I found used any grammar or syntax even close to English; only the vocabulary remained consistent. That said, Facebook’s data shows that conversations like this sometimes still led to successful negotiations between the bots in the end. And according to the author of the article, “The implications of this kind of work are dizzying. Not only are researchers beginning to see how bots could communicate with one another, they may be scratching the surface of how syntax and compositional structure emerged among humans in the first place.” For a language like English, researchers could invest time and development making ways for AI to understand areas like syntax and sentence structure. But if we bypass these hoops for agents to jump through and allow AI to work without needing human language at all, AI-powered hardware or software could learn at even faster rates. At what cost, though? There is so much more to learn. Already, there’s a good deal of guesswork involved in machine learning research, which often involves feeding a neural net a huge pile of data then examining the output to try to understand how the machine thinks. But the fact that machines will make up their own non-human ways of conversing is an astonishing reminder of just how little we know , even when people are the ones designing these systems. —Adrienne LaFrance, editor of TheAtlantic.com Stay tuned. —Meera![]()
![]()
↧