Unpacking the “I” in “AI”

When is Artificial Intelligence going to become truly Intelligent? When will our tools be able to genuinely contribute towards human wellbeing? This and other questions were the subject of a conversation hosted by: SIGMA Focus on Asia: 2020.

Indeed machines are becoming less artificial and more intelligent. They will rely less on bottom-up big data and more on top-down reasoning that more closely resembles the way humans approach problems and tasks.

In turn this sees the usefulness of AI increase as it moves from narrow domain to wide decision-support. This will make AI suitable in scenarios which seemed unsuitable only some months ago.

What kind of Data Architecture do we need to drive real-time activities that are intelligent?

Data allows you to create a profile. In this context that means a detailed understanding of your customer. With this data you can serve customers better and you can predict needs based on known patterns. However what is even more critical than capturing data is making it actionable. Being able to collect the right data is one thing, but making it extremely useful requires a different skill and mindset. More so it necessitates an entirely new set of ethical guidelines around the utilisation of insights.

How do we move from reactive AI to predictive AI?

Traditionally ‘event driven’ computing actually meant ‘event triggered’. We are now moving into an environment which is increasingly predictive using large data sets to act on a data trend perhaps before an event actually occurs.

We can presently predict outcomes based on defined scope with a high degree of accuracy. In fact even with simple linear regression prediction we see accurate results in business environments which are non-complex.

What are the regulatory implications of intelligent AI?

It’s not how intelligent the machines are; it’s how much ‘control’ we give them through legal (or other) instruments.

The first element we need to consider is building trust. We are relying on AI more and more, but it hasn’t yet earned the broad confidence of society. Clearly we don’t build trustworthy AI by simply making AI better at detecting statistical patterns in data sets (such as through better deep learning activities) but through two key actions:

• Framework setting: coming to an agreement about an ethical framework that can build societal trust in AI, not by individual national activity but broad international consensus on a set of universal principles.
• Technological trajectory: moving AI to better understand causality. The common example used here is the cheese-grater analogy. A child can quickly understand a cheese-grater (how it works, how to grasp it, clean it etc.) but no existing AI can properly understand how the shape of an object is related to its function. Machines can identify what things are, but not how they correspond to its potential causal effects. If we develop AI technology to better understand causality we’ll trust it more and it will be more useful.

The European Commission’s framework on Ethics Guidelines for Trustworthy Artificial Intelligence is a great start and lists seven key requirements that AI systems should meet in order to be trustworthy:

• Human agency and oversight
• Technical robustness and safety
• Privacy and Data governance
• Transparency
• Diversity, non-discrimination and fairness
• Societal and environmental well-being
• Accountability

How do we improve customer onboarding with smart AI?

Most on-boarding processes lack intelligent, profile-based signposting which lead customers to the right option or product. Intelligent AI can allow us to guide customers through the right path with least friction based on the perceived needs of the customer. More so this could be done over an omni-channel approach.

Omnichannel strategy includes real-time customer data synchronization across disparate channels which the customer selects based on personal preferences. A banking customer, for example, could begin filling out an application online and complete it at a bank branch, over the phone, or another channel without starting from the beginning with an algorithm intelligently re-hashing the customer profile and re-assesing preferences at every step.

An omnichannel experience doesn’t mean trying to be everywhere with the same content or services all the time. It means being relevant where your customer believes relevance should exist.

Can AI assist in better customer engagement cycles?

It is a known fact that 94% of customers say that they are annoyed by repetitive communications and the present engagement practices practiced by most companies (Twilio, Feb 2020). Consumers want real conversations and businesses need to be able to manage personalised conversations at scale.

There is a major opportunity for businesses to deploy conversational AI to create proactive, ongoing dialogue with customers, rather one-off, reactive, or transactional interactions. This is typically what separates Virtual Agents from chatbots. AI also allows us to focus more deeply on customer experience (CX) through at leat three key methods:

• • Identify more accurate buying patterns
  • Deliver enhanced service over multiple touchpoints
  • Respond faster.

How can we widen the societal trust of AI?

In one word: literacy. We should take time to explain key AI concepts, like classification and confidence levels, ethics and fairness in machine learning, for non-technical audiences. There’s a lot of AI hype which subsequently ‘hides’ the tough questions.  It also shifts the discussion away from AI literacy to AI marketing-speak. 

Also; AI requires a deep partnership between the technologists and the humanists. It presents the need for Interdisciplinary Education and we’re not thinking enough about that.

Photo by Michael Dziedzic on Unsplash