I don't claim to be an expert on AI but as someone who works in the 'digital industry' (whatever that means now) I think its important to keep track of notable shifts in an area that has become fast-moving and of huge potential significance. AI is surely one of the most important S-curves around, and despite AI being around for years we seem to now be on the much steeper upwards growth curve that charaterises classic technology adoption.
It’s been clear for a good while of-course just how important AI is to GAFA (Google, Apple, Facebook, Amazon) but a recent interview with Google’s Sundar Pichai showed just how big and fundamental Google think AI capability is. He declared AI to be:
“one of the most important things that humanity is working on. It’s more profound than, I don’t know, electricity or fire,”
Google's investment in particular, and their 'AI-first' strategy, seems to be generating rapid progress, but in some interesting ways. Last May they revealed their AutoML project which was a form of AI that was designed to help them to create other AIs. Late last year they announced that AutoML had beaten the Google AI engineers at their own game by building machine-software that is more powerful than their best human-designed systems.
In a couple of examples of both simpler and more complex tasks, an AutoML system beat the human-designed system in projects focused on categorising images for their content and a more complicated task of marking the location of multiple objects in an image (something that is important for autonomous robots and AR).
The implications of this are quite profound – the number of people on the planet that have sophisticated AI expertise is limited, meaning huge competition for talent but also representing a key challenge for any organisation looking to ramp AI capability (as surely many are). Google have of-course already created TensorFlow, the open-source software library for machine-intelligence (and Amazon recently announced the launch of Amazon ML Solutions, a new program that connects Amazon machine learning experts with AWS customers in order to help them identify new uses and capabilities for machine learning inside their businesses) but despite this, the development of your AI capability could until now pretty much only move as fast as the number and quality of expert brains that you could put on the problem. The potential to further democratise AI capability through using AI itself could take enterprise and consumer AI application to a whole new level.
Speaking about the automation of AI design and build, Sundar Pichai has said:
“Today these are handcrafted by machine learning scientists and literally only a few thousands of scientists around the world can do this…We want to enable hundreds of thousands of developers to be able to do it.”
Whilst it may currently be easier to adapt a current system to address a new task rather than build a new system from the ground up, the automation of this process could well make it far easier to design increasingly complex new systems. The role for humans in this process changes from designing and engineering to supervising and refining, for example to ensure that biases are not accidentally built into the system.
Whilst the times we live in are witness to some notable milestones in the development of a technology that will likely impact all of us, it is the potential to democratise the technology that perhaps will bring the most fundamental step change.
Leave a Reply