paint-brush
Moving forward with AIby@adhorn
490 reads
490 reads

Moving forward with AI

by Adrian HornsbyJanuary 28th, 2018
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

In the past few months, I have gotten the following question a lot:

Companies Mentioned

Mention Thumbnail
Mention Thumbnail
featured image - Moving forward with AI
Adrian Hornsby HackerNoon profile picture

In the past few months, I have gotten the following question a lot:

“What can I do today to prepare my organisation for the AI revolution?”

Indeed. Self-driving cars. Commercial drones. Smart cameras. Movie and music creation. Powerful & intelligent robots. Over the past few years, a new revolution has brought AI almost to the level of science-fiction. However, most companies are not worried about far-off futuristic applications of AI, they want to know what AI can do — today — for their organisations. Distinguishing the hype from reality can be a bit confusing, especially when you consider the attention that AI gets from the media and commentators.

So, how can your organisation get started and put AI to work for you?

Here are 5 simple things any company can start doing today in order to excel tomorrow.

Understand what AI is

Before figuring out any AI strategy, it is always a good practice to understand what are we talking about and especially what constitutes AI.

AI and the broader field of machine learning is a set of technologies that can be taught to perform tasks and that can improve over time. Similarly to how people learn how to avoid touching hot plates, the continuous learning in AI is the result of feeding back experiment results to algorithms.

One of the most important things to realise is that current AI is not general, but specialised: An AI systems used for classifying images won’t be able to learn anything else than classifying images. Moreover, in order to learn, the AI system must be fed a large amount of data, and not any data, data that is classified.

Indeed, if you want to teach an AI algorithm to recognise bananas, you will have to feed the algorithm with a large collection of images humanly annotated with a tag “banana” and also images that are not “banana”. In fact, most of the time, the more data the more accurate the AI algorithm will become.

AI = Digital Data + Algorithms + Humans in the loop

This is very important to understand because it dictates parts of your AI strategy.

Data collection

Training AI algorithms to solve problems may seem like a far-fetched idea today but the reality is that that day will come sooner or later. When that day comes, you want to be ready with a large amount of data in order to train the AI algorithms. This means one very important thing:

Start storing and preserving __all__ of your raw data because you will never be able to re-create raw data.

Why raw data? because that’s where hidden treasures hide: data correlation and patterns that you can’t yet begin to understand.

On AWS, the default place to store your raw data should be Amazon S3 in what is often called a data lake. Amazon S3 is highly durable, cost-effective object store but most importantly, it supports lifecycle policies. Lifecycle policies allow you to automatically transition objects from one storage class to another, e.g. from standard access to infrequent access and even to Glacier, thus reducing your storage cost. Remember to also enable MFA-delete protection and versioning on the bucket where your raw data is to prevent accidental object deletion.

Note: Make it ridiculously easy to collect and store any type of data. From API logs, system and application metrics to user behaviour, one line of code should be all it takes for anyone in the company to start collecting and storing new data type.

Understand processes that you want to augment

In other words, learn how AI can help your company.

As explained earlier, AI needs a lot of data but more importantly, it needs crystal clear instructions. As a business, you have to be very explicit about what AI can help you with.

For example, if your company is currently storing images and videos for end-users and using manually added tags to perform a search, AI could help you with adding automatically generated tags.

A good practice for each of the processes that you come up with is defining the data needs so you can make sure you start collecting the necessary data for that particular process.

In order to successfully apply AI to your business needs, you should first look at places in your organisation where data is being analysed to help making decisions. Wherever there is data being analysed, AI is most likely a good candidate to achieve this. Sales, marketing, social media, customer supports — many of these functions can be greatly enhanced by AI.

Here are couple of questions you can ask in order to help finding AI opportunities:

What are you currently doing that could be done faster or better?

What are you currently not doing that could help you make better decisions?

Start with the low-hanging fruits

When you start off your AI journey, the most logical first step should be picking up the low-hanging fruits.

Whether you have AI expertise in-house (data-scientist or ML researcher) or software developers, AWS offers AI services and frameworks tailored to meet your needs and level of expertise.

Developers without a data-science PhD can easily add intelligence into any existing application with a diverse selection of AI services that provide computer vision, speech, language analysis, and chatbot functionality. Those services are easy-to-use, API-driven and do not require any training whatsoever.

For example, you can add image and video analysis to your applications using Amazon Rekognition. You can build conversational interfaces into any application using voice and text with Amazon Lex. You can turn text in websites and RSS feeds into lifelike speech and create podcasts with Amazon Polly or localize websites and applications with Amazon Translate. There are a plethora of AI services available for you today to get started.

Learning — Enable skills diversification

Once you have picked up the low hanging fruits, you will need to get deeper in the AI rabbit-hole and especially tailor it to your company needs. High level AI services are great but they are pre-trained on general datasets so can’t be tuned to your particular use case.

In the long run, you will also have to create robust and sound processes for extracting data, training and testing your algorithms.

AI and ML specialist are currently a rare breed and it can take months or even years to find the right people for your team. But that should not stop you.

AI is not some sort of black magic! AI is mostly code, some mathematics and processes for managing data. Nothing that can’t be learned. There are tones of resources online to get started and learn. Most of the people I know, would they be given the time, would love to learn about AI. So, if you are serious about staying at the forefront of technological development and if you believe in continuous learning for your workforce (and I hope you do), invest in learning about AI and empower your current workforce to be the drivers of future developments.

There you have it. Understand what is AI. Take great care of your data. Find the processes that need improvements. Start with the low hanging fruits and slowly develop yourself into an AI-powered organisation, ready to tackle the future challenges ahead.

-Adrian