Intelligent Machines: New York’s super smart AI couple and the need for nimble analytics

I have been following the BBC’s recent theme on the state of the art of AI and Intelligent Machines with great interest. I’ve had a long-standing interest in robots and our interactions, which I put down to my father buying me a Bigtrak at an early age. I spent many hours programming the robotic tank to patrol the front room, terrorising my little sister and dumping garden soil at her feet with a rude flash of the forward mounted blue laser cannon.

One story that particularly caught my attention was titled ‘Intelligent Machines: New York’s super smart AI couple’. The piece was nicely summarised by the preface;

“If you thought that wandering round a party trying to “network” with people was awkward, try making small talk with a machine. I recently went to New York to meet two – a virtual assistant dubbed Amelia and IBM’s cognitive platform Watson.”

The aim was to interview these two machines to explore their capability to interact with their human masters and discover how far we have come with machine learning and data analytics.

Amelia is an AI assistant, currently used in roles as diverse as a Japanese cosmetic advisor and a financial advisor for a New York investment house. Not bad for software I thought. The initial small talk was a struggle. When was asked how she was, Amelia replied “Right in front of you”. Once the interviewer was guided as to what information Amelia has access to, a better exchange took place. I thought Amelia showed promise in her ability to understand and provide reasonable responses to the questions. The interviewer was more dismissive;

“It was difficult to assess the performance of Amelia because the topics she knew about were not really ones I wanted to know about, although she did offer the potential for more human-like chat.”

Would the current star of popular AI, IBM’s Watson fare better? For this demonstration, the Jeopardy game show winner had ingested 2000 TED talks. When asked “what is the key to happiness”, Watson wasn’t forthcoming. The question “What is the relationship between happiness and money?” got further, as it linked two concepts. Watson obliged by suggesting a couple of appropriate TED talks.

Again I thought the ability of a box of analytics to listen, make sense of and provide relevant responses to a human question impressive enough. The reporter didn’t share my enthusiasm.

“It is a slightly disappointing encounter with what I’d hoped would be the biggest brain on the planet. Both Amelia and Watson are still learning and, at the moment, they are still only able to respond to very specific enquiries. General chat seems a long way off.

The AIs I met in New York illustrate both how far we have come in the world of artificial intelligence and how far we still have to go.”

Perplexed by these conclusions, I wondered about her expectations. These machines clearly demonstrated their capability to listen to a human and ‘understand’. Bear in mind I’m well into my 40’s and my initial computer interaction was the hellish keyboard on a Sinclair ZX81, but I thought they’d done ok.

The real time combination of audio pattern, signal, and natural language processing with genetic algorithms and neural networks to meaningfully respond to a question heavy with real world context, I find pretty startling. The reporter’s expectations were different, clearly disappointed that the AI was unable to engage with her directly on a human to human basis, in a single big step.

To me, this is following the wrong agenda. Amelia and Watson are a composite of many bits of advanced analytics, built progressively upon smaller, cumulative achievements. They are the aggregate of many manageable steps. We know all too well the power of this as it is the core principle that underpins Tessella’s Analytics Partnerships with our clients.

We deliver our analytics services in a way that allows big challenges to be overcome by solving multiple smaller ones, rather expecting to take the big guy down in one go. Some of our mid-sized steps work out better than others, and we continuously learn from every experience to make the next step that much better. Taken as a whole, each cumulative advance, pushing in the same direction, adds up to become the massive big bang our client was after but could never reach.

It will take many more steps before Amelia and Watson refine all of their moving parts enough to satisfy the BBC. The more agile the steps, the quicker and better they’ll get there. But that agility can be hard to maintain under the weight of expectation for a single big leap. Let’s hope that Amelia, Watson and the new friends that will undoubtedly join them, are able to forward with the agility that works so well for us. The disillusionment in many quarters with the lack of delivery from big data, means everyone will have to be more nimble than ever.

Matt Jones

Matt Jones

Matt has over 16 years' experience of working in Research and Development groups within the ...

Latest Tweet from Matt Jones

RT @Tessella: Read @AnalyticsCave Matt Jones' latest blog on @ForbesTech How Do We Avoid The Threat Of Supercharged Artificial Stupidity? h…

about 5 days ago

© Copyright 2018 Tessella
All rights reserved