Whether you wake screaming from nightmares of Skynet or have been cross- stitching a welcome banner for the robots for when they take over, the age of AI is upon us. Deep learning, the process which helps artificial intelligence grow, already powers popular technologies such as Google Photos' facial recognition feature, Skype's Translator application, (which instantly translates a spoken word conversation), and many more.
Systems for deep learning rely on what are known as neural networks—entities that function almost like a real brain. These complex networks are trained through a process of inputting large amounts of data, helping the “brains” learn how to perform tasks such as continuing a conversation, recognizing facial features, or winning a complex game.
There’s no doubt that AI is evolving and very real, as evidenced by the fact that over the past twelve months, Microsoft, Google, Facebook, Baidu and many others have joined the race to create the most sophisticated AI possible. In November 2015, Google open-sourced its deep learning project known as Tensor Flow, and just one month later Facebook followed suit with Big Sur, which is actually a server that is optimized for deep learning.
Google recently won one leg of the relay by creating an AI program called AlphaGo which was able to beat a human grandmaster at the Chinese game of Go, a game infinitely more complex than chess or other games previously played by computers.
The applications of AI are limitless. In reality we’re still far from unlocking even a fraction of its potential. But one very practical use case — improved translation— stands out in our globe-trotting, globe-trading world. Baidu, a Chinese search company, already has what is a very accurate voice translator for English, Mandarin, and Cantonese. According to MIT's Technology Review, it can recognize some short sentences better than a person.
Baidu is only one of many exciting emerging technologies in the translation field. Unbabel, a startup that bridges human knowledge and machine learning, allows machines to get smarter by learning from humans. Their platform Smartcheck’s brilliance lies in the fact that it learns from the human translators by “recognizing translation patterns, so that if it sees a particular way of translating a phrase, it will learn that is the correct way of saying it.” Over time the system gets more sophisticated and human-like, which is how machine learning works.
Many things we experience daily such as music, transportation, and finance are a result of artificial intelligence, yet most of us don't even know it. A recently-unveiled technology called Atomwise uses algorithms to enhance drug research, the results of which could be used to slow the transmission of Ebola. In the future, AI will likely be used to determine medical diagnoses, translate sign language, and enhance countless other applications. Even the simple ways it's currently being used in our everyday lives — predicting the things you want to buy, detecting fraud on your credit card etc. — will improve rapidly over the next few years.
We've come a long way, but in truth we haven’t even scratched the surface. Right now we think of AI as being split into three tiers: Artificial Narrow Intelligence (ANI), Artificial General Intelligence (AGI) and Artificial Super Intelligence (ASI).
ANI, sometimes called weak AI, is a computer that is good at one thing. AlphaGo, designed only to play the game of Go, is a perfect example. So far ANI is the only type of AI we’ve been able to create.
The second tier, AGI sometimes called human-level AI will have human smarts and will be able to perform as many intellectual tasks as a human being including: reasoning, planning, problem-solving, thinking abstractly, understanding complex ideas and quickly learning from experience.
Once we’re able to create this kind of AI is when things will get interesting because a computer that has these skills will evolve at an exponential rate, becoming tier three—Artificial Super Intelligence—before we know it.
Once we’ve got ASIs in our midst all sorts of walls will come tumbling down. Mysteries humanity has been struggling to solve since the dawn of time will be cracked wide open (for better or for worse). It’s quite hard to imagine what this brave new world will be like, but according to experts, this future is not that far off. A recent study conducted at the AGI Conference “asked when participants thought AGI would be achieved—by 2030, 2050, 2100, after 2100 or never. The results:
By 2030: 42% of respondents
By 2050: 25%
By 2100: 20%
After 2100: 10%
While this is not a guarantee of outcome, the fact that the majority of people believe AGI will be here by 2050 is significant.
Love it or fear it, Artificial Intelligence is very real and a lot closer than we think. While we’re still far from a point when it will be thinking, acting and growing on its own, we’re perhaps not that far.