Facebook has recently developed a new artificial intelligence (AI), and it has since created its own language using code words to communicate more efficiently. Researchers promptly shut the system down over concerns that they might lose control over the A.I.
This isn’t the first time AIs have diverged from their training in the English language to develop their own, more efficient language. While the resulting phrases from this condensed method of communication sound like gibberish to the human ear, they do in fact make semantic sense when interpreted by AI agents.
Negotiating in a New Language
As Fast Co. Design reports, the researchers noticed that the bots had completely given up on English, but their advanced system is actually capable of negotiating with other AI agents. Together, they can decide how to proceed with whatever they’re working on. At first, the phrases being used seemed unintelligible, but upon further observation, researchers found they represented the tasks at hand.
In one particular exchange, two of the negotiating robots, Bob and Alice, used their own language to complete their exchange. Bob started by saying, “I can I I everything else,” to which Alice responded, “Balls have zero to me to me to me to me to me to me to me to me to.” The rest of the conversation was formed from multiple variations of these sentences.
While the phrases appear to be nonsensical upon first glance, researchers believe they reveal how the two robots are working out how many of each item they should take. Bob’s repetition indicates how it was using the language to offer more items to Alice. Interpreted in this way, his response becomes something like the way we might say, “I’ll have three and you have everything else.”
It seems the AI discovered that English phrases weren’t required for the specific scenario. Modern AIs operate on a “reward” principle, where they expect that by following a course of action they will receive a “benefit.” But in this scenario, for example, there was no reward for continuing to use English, so they decided to use a more efficient way of communicating instead.
According to FastCo. Designs, “Agents will drift off understandable language and invent code words for themselves. Like if I say ‘the’ five times, you interpret that to mean I want five copies of this item. This isn’t so different from the way communities of humans create shorthands.”
Other AI developers have noticed a similar use of “shorthands” to simplify communication. At a company called OpenAI — the AI lab founded by Tesla creator Elon Musk — an experiment succeeded in letting the AI robots learn their very own languages.
What’s the Problem?
If AI continues to create their own languages, developers may have problems creating and adopting new neural networks, but it’s unclear whether this would allow machines to actually overrule their operators.
These new developments, however, allow AI to work more efficiently and can benefit research teams in the long run if they put in the work to learn the new AI-created shorthand and stay up to date with this new method of communication.
What are your thoughts on this? Have we gone too far? Is a Terminator scenario just around the corner? Or is the advancement of technology in this manner just a natural part of our evolution as humans on this planet? Join the discussion. Share your thoughts with us in the comments section of this article!
Reprinted with permission from Collective Evolution.