Google’s recent Duplex presentation was a pivotal moment in the advancement of humanity. It marks a milestone in the development of intelligent machines, showing how machine learning can drive a dynamic conversation with a human, with the human unaware they’re talking to a machine.
In Iain M. Banks’ Culture novels, a futuristic human civilization lived alongside hyper intelligent AI minds. Minds are the ultimate data processing and decision making machines. Minds are the government. They keep the lights on in the Culture, they run foreign policy, build and maintain infrastructure (fleets of space ships of varying types) and handle galactic defences. Everything is connected in the Culture, allowing the minds to apply benevolent influence for the greater good. They have personal relationships and conversations with all Culture citizens. Minds are at the top of the Culture’s hierarchy.
Culture citizens ceded all their authority to the minds after realising that machines could outperform a human in every way. Humans then checked out to enjoy a life filled with adventure, whim and indulgence. As machines were taking care of every citizen’s basic needs, humans had it easy.
Is this the future Google is hinting at?
A Step Change
Duplex represents a step change in real world AI, and for the ambition, innovation and execution, Google should be respected. But, what next, what is the real motivation for producing Duplex? Producing thinking, sentient machines seems to be the natural conclusion. Where else can this effort be driving?
Knowing when we have produced machine intelligence is important, and for that we have Alan Turing to thank. The Turing Test defines a set of simple criteria for a machine to be thought of having human-level intelligence … when all physical factors are removed, can a machine convince a human through conversation, that it’s human?
AI and Case Management
So, what does AI mean for process automation specialists over say, the next 5 years? Beyond Duplex, we can imagine that anyone interacting with a process automation system will be dealing with an avatar, which will interpret the discussion and drive outcomes delivered by some lower level piece of automation, such as ‘apply for a bank loan’ and ‘book a flight’. The avatar would be free to execute that logic at will. It seems unlikely it will refer to a process flow chart to decide what to do, it’s more likely to be informed by a learning AI model. Just like Duplex. Time to finally throw out Visio, we will not think about business process logic in the same way again.
As BPM system vendors, we probably have a choice. Do we all rush out and build avatars or do we focus on enabling services, or both. This is not clear yet, but thinking about what avatars could become is a fun thought exercise.
Who Is Your Avatar?
Do avatars become individual, personal assistants, or do they represent some other entity, such as the government or commerce? Will your avatars communicate with each other, to take care of the detail of your life, such as transferring your energy supply? Flipper Community in the UK is already doing ad hoc, automated supply transfers in search of the cheapest deals, with delegated behaviour running as software. And how will avatars expect to access automation services? Micro-services may fit the bill, but they may themselves may need intelligence, and not just be dumbwaiters.
Today, if you want to integrate business software to an avatar you can try Amazon Alexa or Google’s Assistant. Neither would pass a Turning test, try asking! They do have some intelligent voice recognition but the actions they carry out are trivial. Is switching the lights on in your house from a phone a problem really worth solving? But we are at the start, and that’s ok for AI because by definition AI has to start small.
The Pieces Are In Motion
What a ride this will be! Gary Kasparov in his book Deep Thinking (thanks Anthony Clark for suggesting), describes the 40-year global effort to produce a machine capable of beating a chess world champion … he points out that AI solutions tend to start out useless, moving onto become curiosities, and then over time getting near to human like performance, and finally outperforming humans. If this is a pattern in all self-learning AI, we can expect some time to pass before avatars begin to take over our lives.
Caseblocks was developed to work with knowledge workers and computers, where machine and human decisions are merged to create an outcome for the organisation and often, a customer. We are now preparing for machines to become the dominant decision makers in end-to-end process automation. What a ride this will be!