UK +44 (0) 330 380 1130, Support +44 (0) 330 380 1131, USA Sales/Support +1 888 457 2611

AI – from Zero to hero


Maybe the fear that AI was set to take over the world started with Mary Shelley’s monster – an intelligent human creation which got out of control. Nearly 200 years after this story was published, the goal of making a truly intelligent machine is yielding increasingly impressive results.
Google’s recent success with AlphaGo Zero wasn’t just through better hardware; it was a better way of thinking about AI. With thousands of years of human knowledge sidestepped, and months of computer training compressed, AlphaGo Zero reached World Champion standard in just three days. Continuing onwards and upwards, a few weeks later it became the best player in the world.

Why games

Whenever a new milestone in game playing is passed, people ask what relevance this has to “real world” problems.
For an AI writing poetry, it would be hard to measure its brilliance.
But for a game, you can play AIs against people and each other to rank them. You can also measure ability and progress over time, using the handy ELO rating system.
And there is abundant test data – the raw material of learning – which Google’s latest program even managed to generate by playing itself.

Moving target

As someone who has been playing with AI for decades, the change in attitude to AI is striking.
For a long time, AI was defined as anything computers couldn’t do yet. As each new milestone was conquered, the goal posts were moved to the next unsolved problem.
Like the Wizard of Oz, things only looked impressive if you couldn’t see behind the curtain. Once you could see how things worked, you could see the trick behind the magic, and the problem was no longer deemed hard enough to need intelligence to solve.
An initial attraction of early simulated Neural Networks was their opaqueness. All sorts of nonsense was spoken at conferences as people confused incomprehensibility with ability. Not understanding how a system works risks ridiculous and unexpected errors – for example mistaking random objects for an ostrich.
How an AI works needs to be understood or, like the Wizard of Oz, the “brilliance” of the inscrutable AI can suddenly be shown to be a mirage. I will cover this subject more in a later blog post.
Game playing, though superficially a narrow field, shows relentless like-for-like progress. As computers overtake people in harder and harder areas, the Wizard of Oz’s tricks become ever more sophisticated. When even the tricks behind the curtain are not readily understood, AI’s position is ensured.


I will discuss AI applications of direct relevance to our cloud video platform in later posts. Suffice it to say for now that Forbidden already uses AI elements in its Blackbird codec and has AI concepts for a range of internal tools.
Third party AI video functions are well suited to the cloud, so are readily accessible to cloud platforms such as Forscene. Forbidden is working with major cloud providers to assist development and integrate with such tools.
After years of technical progress, the disruption caused by AI is just beginning. As my Italian friend exclaimed following a particularly difficult general election: the trouble is, the system of disorder is breaking down.
With practical applications poised to move from R&D to everyday life, hold onto your hats: widespread AI adoption is coming.
Are you ready for the extraordinarily exciting times ahead?
Stephen B Streater
Founder and Director of R&D

Pin It on Pinterest