*This is a guest post by Martin McNulty, Chief Executive Officer of Forward3D.
You can follow him on Twitter at @martinmcnulty.*
Conversation between man and machine is nothing new within the world of film, as anyone who recalls Hal from Stanley Kubrick's
2001: A Space Odyssey will confirm. But Spike Jonze's new movie Her and the way the central character, played by Joaquin Phoenix, interacts with "OS1" -- a new operating system that promises artificial intelligence of the future -- shines fresh light on a topic that is becoming pertinent to our age.
Is Her the future of how we will engage with devices and discover information, or will key challenges restrict this direction of technological development?
In 1950, the posthumously pardoned Alan Turing, father of modern computing, devised a test for machine intelligence, of the kind you may be able to fall in love with. The Turing Test's tenet is simple: if a human is unable to reliably distinguish between a machine and a human's response to questioning, the machine can be deemed of equivalent intelligence. We're getting closer and closer to developing a system that could pass this test, although our best efforts remain a distant second to Spike Jonze's OS1, whose conversational ability replaces physical interaction entirely.
The debate around whether we'll eventually get there is becoming less relevant. Technology companies are investing billions in machine learning, natural language processing, and neural networks.
Meanwhile we find ourselves in a very interesting time; Moore's Law is holding true, as processor chips are becoming increasingly small and powerful. Kryder's Law is also holding, with storage density and affordability increasing rapidly. Add to all of this the availability of open-source software and education -- the conditions for imminent machine intelligence are perfect.
At Forward3D, a crucial aspect of our business is to adapt our online advertising for brands in line with changing consumer behaviour, helping ensure the brands we work with are both visible and relevant to their audience. An important and accurate indicator of consumer behaviour is web searches: the kind you enter every day on search engines like Google, or discovery engines like Siri.
In 15 years Google has gone from receiving 10 thousand to 15 billion searches a day -- that's a wealth of data, all of which points toward consumer demand and interest. Accurately parsing its semantic meaning requires the ability to programmatically understand natural language. When I search "Paris Hilton" am I looking for a luxury European hotel, or socialite gossip? It's a complex problem that requires complex technology, of the kind Google et al are investing heavily in.
We're already seeing these investments coming to fruition for consumers. Siri is a great example of semantic technology. Unlike traditional search engine results that are ranked in a list, Siri returns a single hyper-relevant result. This requires significantly more semantic understanding of the search to get right.
Modern consumers thrive on convenient information, and the extension of relationships beyond face-to-face. The advent of wearable technology such as Google Glass and smart watches will serve to exaggerate this trend, but only if their functionality and user interfaces improve. We'll need interactions, not commands, and they'll need to be both conversational and interactive.
IBM's Watson is the closest we've come to this. Back in 2011, when the supercomputer showed former Jeopardy champions a thing or two, it was the size of a master bedroom -- perhaps too large to power a smart watch. Today, IBM has shrunk Watson to the size of three stacked pizza boxes, improved its performance by 2400 percent, and announced a new $1 billion (£600 million) investment. More exciting is that Watson now has an API. Theoretically we could access its power through the cloud, using any internet-enabled device. The result wouldn't be a million miles away from OS1, although at this time you would be hard pressed to fall in love with it.
Computing power and storage continues to get smaller and cheaper and it is conceivable that the necessary processing power in the future could be squeezed into a smartphone sized device. That said, there is a fundamental limit on how small traditional processors can get -- a limit prescribed by particle physics.
To breach this limit we must look toward quantum computing, which is a long way away from being stable in the laboratory, let alone in a smartphone. Perhaps more likely is that the ever-increasing speed and reach of our wireless networking will facilitate OS1 style software to exist in the cloud. Without wishing to spoil the film, there are reasons to suspect that Spike Jonze's OS1 is a cloud-based technology.
Security and privacy concerns are a barrier that any would-be provider will need to overcome. A huge amount of training data is absolutely necessary for software of this kind to function; we're talking about everything and anything, from archived instant messages to bank statements. A provider would have to prove their value proposition, along with their infrastructure's security before an increasingly-savvy consumer hands over her data and faith.
It requires an intellectual leap to believe that we will one day confide our deepest feelings and desires to a computer, especially to the extent that we could inadvertently fall in love with our operating systems. But we know that in tech things move quickly and are subject to incredible change, as anyone who has observed the rapid adoption of tablets from all age ranges. What may seem like the sci-fi dreams of the distant future can become a humdrum reality within a generational blink of an eye -- who knows what tech might sweep us off our feet in years to come?
This article was originally published by WIRED UK
