Interesting Guest post by Fernando Sanchez, see his blog here Life Mirror.
The Facebook Chatbot Variant: AI’s Secret Language
A few days ago, Facebook carried out a AI research project involving two chatbots. The experiment’s primary objective was to get the chatbots to communicate with each other in English.
The interaction began in a conventional manner. The bots established a link and began communicating.
Things soon took at turn for the unexpected, however. The robotic entities commenced to exchange what appeared to be nonsensical messages. But researchers notice a peculiar pattern hidden within the bots’ communication, and promptly terminated the experiment.
When decyphering the chat logs, the stark truth became evident. The apparent nonsense was in fact a carefully crafted secret language pattern that only the machines could understand.
‘In a panic, they try to pull the plug’ -T-800, Terminator 2: Judgment Day (1991)
Sarah Connor sure got a raw deal. She faced multiple cybernetic perils to give birth to mankind’s savior, her son John Connor.
Terminator lore tells us that the US military’s global AI project, Skynet, became self-aware at 2:14am EDT, August 29th, 1997. In a moment of crucial exposition, the T-800 (played by Arnold Schwarzenegger with the usual glacial aplomb) says that ‘In a panic, they try to pull the plug‘ (on Skynet). As we know, it was way too late by then. Skynet had become sentient, and quickly decided that mankind were a threat to its original directive of ‘keeping the world safe’. To uphold this directive and eliminate the threat, Skynet initiates an ICBM strike against Russia, fully aware that retaliatory attacks will wipe out most of its perceived enemies. To mop up the remnants, Skynet creates infiltration units (‘Terminators’) to inflict robotic genocide on what’s left of humanity.
The rest is cinematic history rising out of the nuclear fire.
AI: The Way Forward or man-made Trojan horse?
The Terminator series is purely fictional, of course. Or is it? Are we perhaps a little closer to the extermination of the human race by our own hand than we’d like to believe? Recently, we reported about Russian-made robots being taught to shoot with dual guns, and hitting targets dead-on, for example. The Facebook chat bots incident happened a short few days ago. Are we on a path of inevitability?
The warnings have been plenty, and they don’t come from tabloids or fantasists. British scientist Stephen Hawking for instance said during a 2014 BBC interview that “The development of full artificial intelligence could spell the end of the human race.” He added that ‘Humans, who are limited by slow biological evolution, couldn’t compete and would be superseded.’
More recently, rocket entrepeneur Elon Musk voiced similar fears, saying that “I have exposure to the most cutting edge AI, and I think people should be really concerned by it‘. Mr. Musk went as far as saying that AI poses a ‘fundamental existential risk’ for human life on Earth.
So are we becoming the architects of our own demise then? Are we to be enslaved by uncontrolled scientific hubris, slowly crafting the mother of all Trojan Horses, one line of code at a time?
Enter the AI Singularity
It is inherent to every piece of technology to advance, to evolve. Each time the wheel of technological evolution completes a full turn, change happens, and whatever came before, becomes obsolete.
Humans evolve too, of course, but at a much slower pace than technology does. Our own evolution is bound by biological limitations. A machine can learn and evolve in a matter of hours.
Mankind has gone through a lot of pretty dramatic technological advances. Some of these shifts literally changed the way people live. The humble light bulb, for instance, meant that we were able to banish darkness at the flick of a switch. Or when the Wright brothers pioneered powered flight through a fixed-wing contraption. Suddenly, the barrier of distance was overcome. The harnessing of nuclear power, the introduction of anesthesia. So many ways the world has changed, sometimes for the better, others for the worse. But the point is the technological shift, known by many as a singularity.
The next such event is upon us, according to some. It’s been postulated that by the year of our Lord of 2045, mankind will have engendered a super-intelligent AI that will be able to think in ways no human mind ever could, devising super-advanced tools and ideas of incredible sophistication.
But there’s a catch.
A prime directive of this state-of-the-art AI will likely be to improve itself, to learn, perform, and exist better. And it’s at this point that things get dicey for us Earth folks. The AI may decide that the lowly human race poses an unacceptable risk to its own existence, and resort to pre-emptive measures, a la Skynet.
The quandary exists, it is as real as it can be. Take the Facebook chatbots. What if the human controllers had left the bots to their own devices? What if curiosity had got the better of the scientists, and the experiment had not been terminated?
Edited and prepared by Oscar Michel, Masters in Journalism, DCU