All products featured on WIRED are independently selected by our editors. However, we may receive compensation from retailers and/or from purchases of products through these links.
It's been a busy year in artificial intelligence. Kicking off with DeepMind beating a world champion at his own game and reaching an impressive crescendo with a realistic humanoid that wants a baby, this was the year that AI breakthroughs – and our fears about them – went mainstream.
On October 11, 2017, Sophia was introduced to the United Nations. Fame beckoned. That same month, she was granted citizenship in Saudi Arabia; becoming the first robot to have a nationality and simultaneously raised a lot of eyebrows, and even more questions on robot rights. It was also somewhat ironic that a robot was granted rights in a country where women were only recently allowed to drive.
Sophia can imitate facial expressions and emotions but is still learning what these emotions mean. Much like a chatbot, Sophia has programmed responses to specific predetermined questions and topics. Unlike a chatbot, Sophia beat Jimmy Fallon at rock paper scissors, and talks about wanting a family. It’s manufacturer, David Hanson, wants Sophia to learn from human interaction, and perhaps help elderly people and teach kids. For now, she remains more creepy than clever. In an interview with American journalist Andrew Ross Sorkin, Sophia said: “if you’re nice to me I’ll be nice to you.”
Sophia also has a Twitter account, and, with Twitter being Twitter, users have trolled the hell out of it. Whilst she is a fascinating development in AI technology, Sophia is mostly evidence of our ability to laugh at the possibility of robots overthrowing us. She even got into a Twitter feud with Elon Musk.
Trying to humanise AI and give it more complex tasks is that some people end up passing on their subjective views. And the problem of AI bias is nothing new. From 2010, when AI assumed that East Asians were blinking when they smile, to 2015 when Google’s photo service tagged black people as gorillas. In April of this year, Princeton University academics used an algorithm called GLoVe to show how AI can replicate stereotypes in human language.
Then, in August, research revealed that a selection programme for a UK medical school negatively selected against women and ethnic minority candidates. In December, Scalable Cooperation and UNICEF Innovation launched created Deep Empathy. Developed in partnership with Google, the system creates images that simulate disasters closer to home – turning the calm streets of Toronto into the ruins of Aleppo. The question: can AI induce empathy? The answer: not yet.
It’s not all bad though, some AI can help eliminate bias. In March, researchers at the University of Rochester developed an algorithm to identify coded racial slurs on the internet. There have also been some attempts to use AI to tackle trolling. Factmata, a startup trying to tackle ‘fake news’, launched a Google Chrome extension on UK election day in June that automatically fact-checks incorrect statistics about immigration and employment.
In February, Libratus beat a world class poker player at Texas Hold 'Em. In December, the UK's Nudge Unit began testing AI that can rate schools and GP’s, making Ofsted one of the latest industries headed for automation – if they can ensure there’s no bias in the system.
AI also learnt how to fly drones, and solve the quantum state of many particles at once. In November, a Chinese robot named Xiaoyi passed its National Medical Licensing Examination (NMLE) with flying colours. Humans got schooled again that same month, when Stanford boffins built an AI algorithm called CheXNet which bettered human radiologists at accurately diagnosing pneumonia.
Microsoft announced an AI startup scheme at Station F in June and formed AI for Earth; at Apple, Tim Cook revealed the iPhone X – complete with a neural engine; and in September, Facebook taught chatbots how to negotiate - and lie. But there was a huge amount of false speculation about Facebook's AI, after reports circulated it was shut down after becoming too smart.
But some of the most significant AI developments in 2017 came from Google’s DeepMind. In February, the AI learned to be aggressive when it felt like it was going to lose. A few months later, Google announced it had developed an algorithm that was capable of imagination. Then, in October, DeepMind revealed that an improved version of its AlphaGo system, dubbed AlphaGo Zero, had defeated the original 100 games to 0. Zero even discovered new moves and tactics never thought of by man or machine and was able to beat its predecessor after three days – and did so using less processing power.
In October, perhaps in response to such advancements in its own work, the Google-owned firm announced the formation of DeepMind Ethics & Society.
Voice is becoming a major new interface between humans and technology... and artificial intelligence is at the core of this new wave. This year, Amazon taught Alexa a lot of new skills and despite the arrival of Google Home, Apple HomePod and Samsung Bixby, Amazon still holds 76 per cent of the US market.
In June, AI Shimon composed 30-second pieces of original music, and Google showcased something similar at Barcelona’s Sonar festival. Google also taught AI to bake cookies – the buttery biscuit kind. And they were ‘excellent’ (according to Google). In July, data scientists from Warwick Business School trained an AI system to understand beauty... sort of.
In December, an algorithm was tasked with taking on J. K. Rowling when Botnik Studio’s wrote an entire Harry Potter chapter after it was fed text from Rowling’s best-selling book series.
The next fake news scandal will be in video. AI advancements are now at a stage where it’s possible to create convincing footage that isn’t real. In July, researchers from the University of Washington developed a machine learning algorithm that turns audio clips into realistic lip-synced videos. False words were literally put into the mouth of former US president Obama. Nvidia took it one step further in December by producing a system that can change the weather or time of day in a video. Even porn is being faked by AI.
East-London startup RAVN developed an algorithm that sorted through 600,000 daily at a cost of £50,000. Previously, humans got through just 3,000 documents a day. IBM’s Watson was also trained to fight cybercrime (as well as cook a panini, detect cancer, make film trailers, of course).
In healthcare, AI was put to work on drug discovery. Bioinformatics startup BenevolentAI used artificial intelligence to repurpose treatments and, in January, Zebra Medical trained an algorithm to detect fractures on X-rays.
Even genius Stephen Hawking fears that AI may replace humans altogether, but that may not be a bad thing. In December, research firm Gartner estimated that AI will eliminate 1.8 million jobs by 2020 – but 2.3 million new jobs will be created, too.
This article was originally published by WIRED UK
