Does Artificial Intelligence have a Last Mile Problem?

Artificial Intelligence – The Last Mile

Extract of presentation given to AI Summit London, May 5th 2016

Good Morning, I wanted to briefly talk today about something called the Last Mile.

The last mile was (and is today) an issue for the telecommunications industry, namely the stretch of twisted copper cable that connects the consumer’s house to the local exchange.

The reason this is important is because no matter what smart new high capacity fiber-optic technology is being built between the exchanges or exciting new multimedia is available to sell to the consumer, what restricts the unlimited potential of the internet for the consumer was and is this small twisted copper wire with historically only enough bandwidth for a limited download speed…

Now if I draw a general and non-technical analogy here to the situation we find ourselves in today in relation to Artificial Intelligence (AI) software, I think we have a last mile problem too …

It is about our inability to connect all this data and content that is becoming more and more available, with the human expertise required to turn that content into meaningful products and services that are usable and actionable at internet scale.

In the field of Ai or cognitive technologies, this, I believe is our ‘last mile’ issue.

So for example, today, instead of being able to…

• -tell an asylum seeker whether they are eligible to stay in this country, process their case, apply all relevant facts and judgement and grant them an instant entry visa, all instantaneous and online, we ask them to fill out an online form, consult an advisor, read up on their rights, wait, decide, appeal etc.
• -Instead of being able to assess a complex financial transaction for all applicable regulations and make an instantaneous and online decision based upon the level of risk I or my company are prepared to accept, I have to gather and consider that information, seek expert advice, discuss and then decide and so on…
• -Instead of having a secured loan application assessed and approved and transfer the funds instantly, I have to fill in an online form, wait for approval, speak to an advisor…

And so on…

And the reason for this, is of course, because this is where historically, the human expert comes in.

 An expert who consumes all the information that is available and applies context, and precedent and judgement and know-how to a unique set of circumstances to provide context specific decisions.

 That we trust.

And right now, it is our biggest constraint in AI, as we have manifestly failed to effectively emulate that expertise in software. Hence why so many AI bots and related online products and services are still fairly basic, routine and static.

So then I ask myself, what would we need in order to close this gap, to be able to do the things I described, to go the last mile and include expertise in software?

• -Well I would need to be able to absorb facts about a unique circumstance, and respond intelligently with more requests for information until I had enough to start its expert analysis
• -And I would need access to knowledge contained in public and non-public places, structured and unstructured data
• -And I would need to be able to learn from past successes and failures
• -I might need the knowledge of hard rules and soft social & cultural norms
• -I would need to apply know-how or judgement
• -I would need to infer facts and conclusions
• -I would need to consult with others
• -I would need to be deterministic, bold enough to make a decision (not probabilistic)
• -I would need to explain to me how it had come to that decision

And probably a range of other things….

I would need to be able to reason, instantaneously and in real-time across all of this complex content

If I could put a system that combines all of this expert reasoning power on top of the incredible AI data technologies being developed today, then it would be truly unleashing the unlimited potential of the internet into the real world…

The equivalent to replacing the twisted copper wire with high-capacity 8 fiber optic, bringing it straight into my home, giving me answers to all my most complex questions, solving my issues, providing me the complex products and services I need, in real time.

Think real time medical diagnosis, think instant digital justice, think ambient compliance systems, think highly advanced online consumer services… and on and on

The expert is now in the system. On their best day, every day.

So that’s what we at Neota Logic are working on. Feel free to contact us to hear about how we’re already doing that in many industries today and if you are an expert, how you can model your expertise and make it available at internet scale.

Rick Seabrook
Managing Direector EMEA
Contact: seabrook@neotalogic.com

Rick is an expert in business strategy, organic and acquired growth, and transformational change across a range of industries and professional services. Prior to joining Neota Logic in 2015, he served as European Managing Director of a Nasdaq-listed legal technology business where his role focused on strategic growth and commercial delivery. Rick also held the title of Managing Director (formerly Partner) at Accenture for over 18 years, where he led the development and management of services businesses across a range of industries, including Oil & Gas, Utilities, Telecommunications, and Mining.