This story is half fake. Zuckerberg wanted to make the news so they inflated this story so it looked as if two robots talked in ‘a new language’ with each other. Instead, it was very much like my story AlphaGo Computer Beats Top Go Player Ke Jie: Computers Don’t Think Like Humans I discussed the odd fact that computers prefer linear systems that are fairly simple and straight forwards versus circular thinking that is more ‘organic’. Relentless logic leads to a closed system. Zuckerberg is mistaking computers being ridiculously simple with ‘creating own language.’
Facebook shuts down chatbots after they make own language | Daily Mail Online reports. This made big news which Zuckerberg wanted. I looked into this ‘new language’ and nearly died laughing:
This is not a ‘new language’. It is pure repetition icon actions. It is utterly moronic ‘language’ sort of like how insects communicate with each other. What? Oh, insects transmit information such as bees have ways of telling other bees where to find pollen via dancing in circles while wiggling their little rear ends.
The repetition thing is very insect-like. All primitive forms of communication takes some element and then repeats it a clear number of times and this information is then accepted by the other party which chimes in with its own ‘bird call/insect noise’ and they agree in a way, via having simple goals.
The bots were also trained to never give up.
The article goes on to falsely claim the bots ‘negotiated’ in this ‘clever way’ instead, it was simply doing the same thing over and over again until it reaches infinity or gets set into a new track by some other element. This isn’t a ‘negotiation’ at all since it can’t stop until it has to stop due to human restrictions on this tendency towards infinity.
The new language was more efficient for communication between the bots, but was not helpful in achieving the task they had been set.
This is false. The people running this program systems had to pretend they found something profound. They did not. The computers were MIRRORING each other. This is why they used this cumbersome system of communication: they were not ‘talking’ with each other. They were reacting like a mirror.
The multiplication of human words hides the fact that the computers were reducing everything to basic code since it has zero comprehension of what was going on, it was simply, due to interfacing with itself, falling back into neutral mode which is REPETITION. Computers can go to infinity repeating something.
‘Agents will drift off understandable language and invent codewords for themselves,’ Dhruv Batra, a visiting research scientist from Georgia Tech at Facebook AI Research (FAIR) told Fast co.
‘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 shorthand.’
This is NOT ‘short hand.’ It is the most cumbersome way of doing things. It is less like writing and more like having to chisel things in rocks using a rock: primitive, insect like brain work.
The programmers had to alter the way the machines learned language to complete their negotiation training.
The humans working for Zuckerberg were not teaching computers anything. They hoped that computers would want to communicate like humans but once they had the computers talk to each other, they did what all computers like doing basically: repeating something over and over again to infinity.
Computers love infinity. I once wrote a story long ago about how the Goddess decided to unleash an unlimited number of monkeys to see if they could accidentally write a Shakespeare play. One eon, the head monkey came to the Goddess and said, they succeeded. She then saw that they accidentally wrote all of Shakespeare’s plays including the alternative folios (insert joke here) and she screamed at the monkeys, ‘This is nearly impossible!’ and then destroyed the entire Universe…which is how the Big Bang happened.
Oh, and she slipped on the banana peels left by the monkeys typing on a zillion typewriters.
I noted earlier this year that two powerful computers playing Go games, over time, they end up making the games utterly simple and then surrendering earlier and earlier until I proposed these two computers would agree after just three moves, who won and so, not bother with playing at all since it was no longer necessary.
This is how computers ‘think’ which is why letting them do the thinking is fatal for humans. Ahem: letting computers run our nuclear arsenal, for example, will lead to definite extinction of all humans. This is why we cannot turn over systems to computers. Humans are often very stupid but computers are always very logical and this spells our doom, being illogical creatures and thus, a perceived danger to the computer complex.