Image Credit: 2001 - A Space Odyssey (MGM)UPDATED:
UPDATED
In 2014 Elon Musk spoke out about his fears in an interview about Artificial Intelligence, saying that "I think we should be very careful about artificial intelligence. If I had to guess at what our biggest existential threat is, it’s probably that."
That same year, both Bill Gates and Stephen Hawking also added their collective weight to the argument, fearing that "the development of full artificial intelligence could spell the end of the human race."
They are only partially correct.
It is not the artificial intelligence we should worry about, rather who gets there first, and it could very well be a beast that 1 billion users are content in feeding every day.
Sound familiar ? Well it should as it was the premise behind the Bluebook A.I. in the movie, Ex Machina. To program the A.I. behaviour of Eva, the main character, Nathan, harvested personal information from billions of Bluebook users, using their search queries as indicators of human thought, and hacked billions of cell phones for recordings of human expressions and body language. Facebook has access to all of this already and more, given it owns Instagram and WhatsApp.
While M is not a fully automated intelligence (there is a human interaction and curation involved to aid it with more complex queries), Facebook does track what the human operators do to learn more about how to handle more intricate queries. “Everything the trainers do, we record every step,” says Alex Lebrun, an AI expert at Facebook involved in the project. This information includes what websites they visit, what they say over any calls made, and what they type in response to the Facebook user.
It goes deeper. Facebook revealed in 2014 that it had manipulated the feeds of over 500,000 randomly selected users to change the number of positive and negative posts they saw. It was part of a psychological study to examine how emotions can be spread on social media. However, on closer examination you could surmise that this experiment was also a study of human behaviour and thought, rather than just emotion. Any true test of artificial intelligence would need to understand how complex the human psyche is when responding to positive and negative stimuli.
But that's not all. Facebook also uses a code that keeps track of every time you delete a would-be post or message and sends metadata about that message back to its own databases. In other words, Facebook is looking to understand how humans explicitly think, behave, react in real-time, and also their tacit thoughts and reactions that never see the light of day online.
This is only part of the story. At Facebook, their Director of A.I., Yann LeCun, a Deep Learning pioneer, has access to more living data relating to human thought, language and behaviour than anyone else. "Much of our work at Facebook AI focuses on devising new theories, principles, methods, and systems to make machines understand images, video, speech, and language—and then to reason about them.” said LeCun in an interview with IEEE Spectrum this year.
You could think of Deep Learning as the building of learning machines, say pattern recognition systems or whatever, by assembling lots of modules or elements that all train the same way. So there is a single principle to train everything.
Facebook's approach is very different from some of the more 'traditional' methods of building an artificial intelligence.
The Human Brain Project (HBP) for example is a massive endeavour that aims to accelerate our understanding of the human brain, and develop new brain-like technologies.
The HBP is organised into different subprojects including:
- Neuroinformatics (searchable atlases and analysis of brain data)
- Brain Simulation (building and simulating multi-level models of brain circuits and functions)
- Medical Informatics (analyzing clinical data to better understand brain diseases)
- Neuromorphic Computing (brain-like functions implemented in hardware)
- Neurorobotics (testing brain models and simulations in virtual environments)
- High Performance Computing (providing the necessary computing power)
This all seems like a lot of work to, in essence, build a machine that humans hope will become artificially intelligent without understanding fully what it means to think like a human in the first place. Something which Facebook has access to.
Intelligent assistants like M are only the precursor to building a true A.I. that will one day function on its own, and mimic human thought and behaviour through its understanding of us, if not become an emergent intelligence in its own right. What Musk, Gates and Hawking fear is an artificial intelligence built by a handful of humans based on complex algorithms. What they should fear is an artificial intelligence built by Facebook that was schooled by over 1 billion humans.
UPDATE:
Facebook announced this week that its Parse project, its major push into the Internet of Things race, has been expanded with new SDKs for silicon vendors like Broadcom, Intel, Texas Instruments, and Atmel. With Facebook intending to funnel the interaction of data between devices, and devices and people, this expands the amount of data available for any A.I. project to use in understanding behavioural patterns. Not only will Facebook know you as stated above, it will also know what devices you use, why you use them, when you use them.
On August 27th 2015, Facebook hit 1 billion logons in a single day. It's reminiscent of the fabled date behind Terminator 2: Judgement Day (which was August 29th) when Skynet became self-aware. Within 10 years time we may well be reading the same thing about Facebook itself.
And we may never be able to pull the plug.
