Blog: People Should be at the Heart of AI Design Process
In my previous posts, I’ve made the case for why it’s a strategic imperative for large companies to take a bold approach to ethical AI in order to match the pace of AI innovation. This is #2 in a series of posts on Enterprise Artificial Intelligence (AI) that takes a closer look at how global companies are currently integrating ethics into their AI programs, processes, products, infrastructure, and even culture.
[Here is a link to blog post #1, in case you missed it]
I took advantage of my invitation to the Atos technology days, co-located with VivaTech in Paris, France’s premier show for startups and innovation to catch up with Jérôme Stoller, Head of AI & Software Engineering at Atos and get his take on the barriers to adoption of AI and more importantly, how can ethics can keep up with accelerated adoption of AI?
As global companies race towards adoption of AI, it is imperative that they pause and consider the objective for AI. Stoller cautions against letting algorithmic biases creeping into their AI systems by ensuring diversity at the beginning will prevent ethical blind spots. He stresses that people are integral to the AI design process and in making sure the systems are not biased.
Here’s more from our Q&A below.
Mia Dand: What is the cognitive data center?
Jérôme Stoller: The Cognitive Data Center is a tool which aim is to raise the overall efficiency of a data center. It builds a model of the “normal” behaviour of the data center, and then, based on a real-time analysis of IT equipment and machinery data, is capable of anticipating potential problems. This means you can solve problems before they occur and therefore improve performance and reduce any downtime.
The Cognitive Data Center receives data from network, storage, servers, applications, cooling and energy consumption. It analyses in real-time all the events, displays significant graphs with predictions, associated with a confidence rating, about potential outages and the impacted elements on the data center.
MD: What are some of the barriers to adoption of AI in the enterprise?
JS: For AI to be successful we believe you need to have four vital ingredients in place. The first is the business case — what’s the issue you would like AI to solve? The second is the availability of data — AI doesn’t work without data and you need to ensure that accessible, actionable data (with enough depth and history) is available. Then, of course, you need to have the relevant infrastructure, with enough computing power. The fourth is security — you must be absolutely sure of the security of your data and your customers must trust that you are managing this appropriately. None of these is insurmountable but they are all very necessary to successful AI adoption.
MD: How will quantum computing change HPC (High Performance Computing) ?
JS: We believe that quantum computing won’t replace today’s computers. Instead, when the technology will be available, we are convinced that it will be used as an accelerator, for some specific workloads, eligible to the quantum acceleration. Like GPUs or FPGAs are already accelerators for some workloads, that general purpose CPUs are in comparison, slow to handle.
MD: How should companies integrate ethics into accelerated adoption of AI?
JS: The first thing they must do is understand exactly what they want AI to do and the problem they would like it to solve. This must be specific and identified. In this way they will also understand its broader impact. They must be careful to build algorithms without bias which means diversity of thought going into the design phase. They must also make sure that people play a central role in the running of AI and they keep people at the heart of the process with the ability to make alterations if necessary.
And that was it.
Stay tuned as I continue to cut through all the hype and jargon to bring you the inside scoop on how enterprise companies are ensuring AI is being deployed ethically in their organizations. Tweet me @MiaD or leave a comment below as I would love to hear about your AI trials and tribulations!
About the author:
Mia Dand is a strategic digital marketing leader and passionate diversity in tech advocate with extensive experience in building customer-centric programs at global companies like Google, HP, eBay, Symantec and others. Mia’s unique expertise is in leading complex cross-functional programs at the critical intersection of business, data, governance, and technology. As the CEO of Lighthouse3, an emerging tech research and advisory firm based in Oakland, California, Mia excels at identifying key industry trends and guiding F5000 companies on the responsible adoption of new & emerging technologies like AI for successful business outcomes. Mia is also the author of “100 Brilliant Women in AI Ethics”, a definitive guide to help global organizations recruit more talented women and empowering more diverse voices in this space. You can connect with her on Twitter or LinkedIn.