People with fears of artificial intelligence becoming a little too real might feel a sense of alarm at the news that a pair of Facebook chatbots appear to have carried out successful negotiations in a language only they understand.
The social media giant wanted the two AI bots, named Bob and Alice, to carry out a negotiation on certain items, including balls and hats. The team behind the trial wanted to see if the robots were able to trade successfully, and perhaps even haggle on the value of items.
An extract from their exchange can be seem below, taken from The Independent’s website:
Reading this, one could be forgiven for thinking that negotiations broke down or there was a glitch in the robots’ software, but researchers believe there is a pattern and structure to this seemingly garbled conversation. What’s more, some of the negotiations that took place in this mangled English were in fact successful, suggesting that AI programmes are capable of finding coded ways to communicate with each other that humans may not be able to decipher.
In particular, it’s noticeable that the robots keep talking about themselves, with Bob repeatedly saying ‘I’ and Alice constantly using ‘me’. This is likely not a glitch, but a form of emphasis or a way of expressing quantity. One researcher, Dhruv Batra, suggests that saying the word ‘the’ five times, for example, could be a request for five of something, so the phrase ‘the the the the the ball’ might be a request for five balls. Generally though, it appears the actual meaning behind this unusual AI discussion is not entirely clear.
A number of media outlets have added to the hyperbole of the story by reporting that these peculiar negotiations led Facebook to “shut down” the two robots, giving the implication that they were so worried about the indecipherable scheming of Bob and Alice that they decided to kill them both off. However, Facebook researcher Mike Lewis has explained that the real reason for the shutdown was that they want to build robots capable of communicating with humans.
Many respected figures in science and technology, notably Stephen Hawking and Bill Gates, have expressed grave fears over the threat of artificial intelligence, and countless films have been centred on the preoccupation of AI becoming self-aware, ranging from 2001: A Space Odyssey to the Terminator franchise. Others suggest that there is only a risk from AI if the humans controlling it are irresponsible.
Overall, we should perhaps not be overly concerned by bots using their own form of language to communicate, as they are only taking the English we know and truncating it to a more direct, computerised form. Although the exchanges between the bots seem strange, the outcome of them was nothing more than that they traded virtual balls and hats, which is what they were programmed to do.
- How to find a circular reference on Excel - May 23, 2024
- Five life skills learned from internet marketing - January 3, 2024
- How artificial intelligence can (and can’t) help you write content - September 29, 2023
