Juha Saarinen: Wise to move on artificial intelligence

AI will continue to creep into our lives, just like cloud computing has. Photo / Getty Images

Creation of a superior being that can out-perform humans is underway so developers better not get left behind

The end goal of computing and information technology has always been to create a superior being that can gather, store and process vast amounts of information faster than its human master's eyes can blink.

Disobedient and random humans are not as well-suited as machines to sift through large information troves. Knowledge is power, though, which can be bestowed on those who develop artificial intelligence systems that tirelessly process information.

Huge amounts of money and massive effort is being poured into developing artificial intelligence systems. Microsoft, IBM, Apple, Google and other large tech companies are racing to grab the lead when it comes to machine and deep learning, automated reasoning and natural language computing.

This is happening across the IT industry: NVIDIA, which is best known for its fast video cards for gaming, released the new Tesla P100 graphics processing unit recently. This has 15 billion transistors, and reportedly cost US$2 billion ($2.9 billion) to develop.

NVIDIA doesn't see the Tesla P100 used for video games, though, but for deep learning and artificial intelligence applications. Eight P100s in the small NVIDIA DGX-1 "supercomputer in a box" system can be had relatively cheaply - US$129,000 - and provide the sort of performance systems costing millions deliver.

AI will continue to creep into our lives, just like cloud computing has. Microsoft is betting on "conversations as a platform", for instance an immediate future in which users interact with bots developed as intermediaries for large computer systems (tinyurl.com/nzh-msconvos).

What you and I see of AI currently is only moderately usable - personal digital assistants such as Apple's Siri and Microsoft's Cortana try to figure out what exactly it was you said.

Sometimes they're spot on, other times hilariously wrong, like when you try to send a text with "Hey Siri" in a noisy environment (apologies, family and friends, I was researching this column, and hadn't suffered a brain malfunction or downed three pints of gin before sending those messages).

Then there's Microsoft's recent AI-pratfall with Tay the Twitter bot which hadn't been set up to learn only nice things from users, and which was subverted into a rather unpleasant artificial chat being quickly.

Bots are only the surface, though, of large-scale AI systems that are improving rapidly. For instance, the first iterations of Siri were neither here nor there.

Now, I use Siri on a daily basis, often via the Apple Watch, instead of poking at the screen of my iPhone. I'd like to have it on my OS X desktop as well, in fact.

AI will change everything faster than you think, with human-machine interactions becoming the norm and entire business and organisational processes being handled by automated reasoning.

Google, meanwhile, is using AI to improve search engine rankings, and even interact with user data.

At first I thought the Photos Assistant which created panoramas and albums automatically, using pictures and videos you've taken and stored in Photos, was intrusive and annoying.

Grudgingly, though, I have to admit the Assistant does a good job, probably better than what I could do, if I could be bothered, that is.

I'm still underwhelmed by the Google Inbox AI generated auto-replies to emails received which, if I were to send them off, would make people think I'm either weird or rude or both.

Which is not to say that I wouldn't like to have working AI-generated email replies. I would, I receive way too many messages and not having to process some of the more routine ones would be fantastic.

A programmable email inbox, powered by AI that understands the messages? I'd give that a go, even though I wonder what would happen when my machine-generated responses hit another AI-powered email recipient.

AI will change everything faster than you think, with human-machine interactions becoming the norm and entire business and organisational processes being handled by automated reasoning.

What this means for society as a whole I can't say, but if you're a developer or business looking ahead and you're not in deep learning mode, you'll lose out.