In my last couple of articles I have explored what AI is and why it has become increasingly relevant in the modern information age, the next question is how can I get started on the journey? This is the right attitude, after all every outing starts with a single step. The first thing to think about is what do you want to achieve? Just because you have a hammer with the words Artificial Intelligence on it, do not see everything as a nail. Remember that your toolkit has many mechanisms in it that can help addressing your business problems, AI is best used when the rules required to understand complex datasets are either extremely expensive to implement / maintain or complex and time consuming to define. Once you have this as a starting point, work on the use cases you want AI to address. We all live in the data driven economy, the best way to start is with your data and how you can gain insight from that data. Once you have articulated the business problem you are trying to address and collected the data you then need, start to think about how you can use the data to gain insights. This will influence the decision making process when it comes to selecting the right models & AI frameworks you want to use.
Once you have an understanding of the problem you are trying to address there are a couple of options available to you. Either you can start to build models yourself, you can buy a model that someone else has built (and implemented into a software product) or you can work with models that someone else has taken part of the way. There are pros and cons of each approach, so it does depend on the outcome you are trying to achieve and constraints of your available budget.
Going it alone can be a daunting prospect! There is a great phrase โif you want to go fast go alone, if you want to go farther go togetherโ, choose a partner someone you can work with and just as importantly someone who has an ecosystem. AI is a team sport and one that requires close collaboration across a broad spectrum of disciplines in order to be successful. If you are going to build it yourself or work with a partner to define the models, then the first thing you need is the data. Implementing AI will be extremely challenging if your data is locked away in disconnected source systems. Getting the data into an integrated format that can feed a business process is the only way for AI to hit the objectives in the business case. If your data is not integrated and aggregated you might as well just run simple analytics against each separate system. The value of AI comes in find patterns across complex and disparate data sources.
Even if you have all the data, quality is still king. Without it you can be training your model on the wrong thing. Biases also present a significant problem as they can be hidden within your training sets. There are various kinds of discrimination, racial or gender or age and these need to be guarded against in the input data set. You need to analyse the data in line with your organisations principles, goals, and values. Selecting the right analysis tools that can help you understand how the AI made the decision it did for internal / external auditors or review boards should also be thought about in the early days. One key thing is that โthe computer told me tooโ does not stand up to regulatory scrutiny, so this needs to be considered; especially if the AI is part of a decision support process.
Compliance is always something that needs to be taken into consideration when it comes to data sources. Just because you have access to information, does not mean that you can use it any way you want. There are lots of ways that you can audit your AI models, and look at the data that goes into those models. You can either do it using an independent auditing firms or build the capability using your own data scientists. There are some interesting fields of mathematics that are helping, in the form of Algorithmic Accountability. It is early days on the role that regulatory bodies will play in this field, but one thing for sure that is that they are all keeping a keen eye on recent developments. Regulators need AI as a regulatory tool in their toolkit as much as businesses need it in theirs!
As a partner, HPE can engage with you whether you already on the AI journey or just beginning; we have a broad and deep understanding of how to accelerate you on the road. If you are just beginning to think about how to get moving then the Artificial Intelligence Transformation Workshop is an excellent starting point. This workshop is aimed at helping to understand the core of what AI can do and also identify what data can be used to support an AI implementation.
Artificial Intelligence and more specifically Deep Learning promise to radically transform many industries, but they also pose significant risks; many of which have yet to be discovered, given that the technology is only now beginning to be rolled out at pace. I think in my next article I will explore the darker side of AI.
Featured articles:
- AI promises and perils [Enterprise.nxt]
- How to prepare your business for AI [Enterprise.nxt]
- Want to know the future of technology? Sign up for weekly insights and resources
Matt Armstrong-Barnes
Hewlett Packard Enterprise
twitter.com/HPE_UKI
linkedin.com/company/hewlett-packard-enterprise
hpe.com/uk
Mattab
Matt is Chief Technologist for Artificial Intelligence in the UK&I and has a passion for helping customers understand how AI can be part of a wider digital transformation initiative.
