dmi

Section 508

March 21st, 2017

3 Steps to Success in AI for Anyone

Expectations have been set high with Elon Musk talking about robots/automation taking over all our jobs, as well as Bill Gates and Stephen Hawking warning us about the threat of AI to mankind. Stephen Hawking later retracted his statement and said that pollution and “stupidity” are bigger threats.

On the flip side, too many companies claim to do AI when in fact they are just providing normal software services powered by algorithms or decision trees. An algorithm is not AI (or machine learning) unless it’s self-optimizing to achieve better results.

But AI is rapidly starting to impact every industry and part of society ranging from smart cities, self-driving cars and electric grids to robots, e-commerce sites, conversational interfaces and finance. Using Google as an example, they had leveraged AI in more than 2.700 projects by 2015 and this has accelerated since, according to Bloomberg’s Jack Clark. At the recent NEXT conference, Google emphasized that AI is critical to the future growth and success of Google Cloud (disclaimer: DMI is a Google Cloud Specialist partner).

AI also plays a key role in DMI’s end-to-end mobility offering with 10+ AI projects delivered in 2016 and accelerating in 2017. Every project provides its own insights and learnings.

As a starting point, we believe there are three important steps for any organization or project to leverage AI:

  1. Learn from humans, nature and existing problems
  2. Experiment and evaluate feedback
  3. Leverage the brains (and tools) of others

Here’s a somewhat more detailed explanation of what we mean:

1. Learn from Humans, Nature and Existing Problems

The biggest recipe for failure is usually jumping directly to trying to solve problems with AI. Before solving a problem, we need to really understand it in-depth. This starts by collecting data and insights.

Example:
To implement an AI-powered chatbot, start by implementing a human-powered chat service and see what people ask about, how they ask and to what extent you can solve their problems. Try influencing the questions posed by asking different questions back to narrow down the requests.

2. Experiment and Evaluate Feedback

Is it possible for an AI-powered computer program to solve the problem better or as good as humans with a reasonable amount of time and resources? If the answer is believed to be yes, then try it out in the same way as when Google Go went head-to-head with the world champion in Go (but do it in a less expensive way).

Prototype and see how the computer’s response compares to the human response. Evaluate and improve if it’s possible to get better or close enough. If it’s underwhelming, then maybe it’s better to focus on something else.

Example:
To automate customer care for broadband connections, prototype a service that predicts and provides advice on the solution based on the customers’ description of the problem including video and voice. Then compare the solution to the human response and resolution success rate. Once the success rate is better or close enough, then it can be implemented.

3. Leverage the Brains (and Tools) of Others

Once the problem (with data) has been identified and a solution found, select the right tools. One of the reasons that we are seeing an explosion in AI implementations now is that the cloud services provided by Google, Microsoft, IBM, Microsoft, Baidu and the Open AI initiative makes it easier for anyone to get started.

And there are lots of other great companies that solve specific problems in the fields of image recognition (Clarifai), cyber security (Sift Science), natural language processing (MindMeld), text analysis (Narrative Science), deep learning (H20.ai), commerce (BloomReach) and a myriad of other things.

Example:
Images need to be automatically categorised and tagged for an online second hand goods provider. Leverage AI-powered Clarifai and the service could be up and running within a few days and improved over time.

In conclusion, your job probably won’t be replaced by a machine tomorrow, but the opportunities for AI to improve customer experience, productivity, reliability and safety are here today. Get started immediately with the tips above or contact DMI for more insights. The journey has just begun.

Magnus Jern
Chief Innovation Officer

Tags: AI artificial intelligence chatbot machine learning

Connect with us

Job Openings

Want to be part of our growing team?

View More
Work with us

Learn how DMI can help you grow, or launch your business.

Get In Touch
Offices

See all of our locations around the world

View Locations