AI & Big Data Expo 2021: Ethics, myths and making sense of data

The global AI market is predicted to snowball over the next few years, reaching a $190.61 billion market value in 2025. By 2030, AI will lead to an estimated $15.7 trillion, or 26% increase in global GDP.

Industry analyst firm Gartner has also predicted that by 2022 companies will have an average of 35 AI projects in place. 

Of all the end-use industries, the market for manufacturing is expected to grow the fastest. Increasing data volume derived from the manufacturing value chain has led to the involvement of AI-enabled data analytics in the manufacturing sector. In addition, several industry initiatives, such as Industry 4.0, a connected manufacturing initiative by the Government of Germany, have proliferated the growth of AI-enabled devices in manufacturing.

Businesses have also noted that automating tasks, such as invoicing and contract validation, is a crucial use of AI. 

Meanwhile, 80% of retail executives expect their companies to adopt AI-powered intelligent automation by 2027. The most prominent use case for AI in the retail industry is customer engagement (chatbots, predictive behavior analysis, hyper-personalization).

On the downside, a lack of trained and experienced staff is an expected restriction in the AI market’s growth. 

At the AI & Big Data Expo in London this week, all the opportunities and challenges surrounding AI and big data were keenly debated.

The event, part of TechEx Global exhibition and conference, showcased some of next-generation technologies and strategies, and offered an opportunity to explore and discover the practical and successful implementation of AI & Big Data in driving forward business for a smarter future.

Of all the talking points over the two-day event, some of the more prominent discussions focused on the myths and misunderstanding of AI, as well as ethics, algorithms and how companies are coping with vast quantities of data.  

Dispelling the AI myths

Speaking at the event, Myrna Macgregor, head of machine learning – strategy (acting) & lead, responsible for AI+ML at the BBC, said: “It’s important to think about what the challenges are. There are some barriers, obviously. Firstly, the technical point. Very technical terminology is often used in conversation regarding AI and coding algorithms.And these are things that are unfamiliar to some of the stakeholders you’re going to be working with. 

“Secondly, there are a lot of theories about how AI is going to take jobs. That a large percentage of jobs will disappear in the next 10 to 20 years, and that’s really unhelpful. AI, as an assistive technology, is a tool to help people do that work. 

“Also, a lot of people think that AI is an easy thing to do. There is a perception, perhaps, that AI is something you can grab off the shelf and throw at problems, or you can just sprinkle a little AI on problem.”

Ethics and responsibility

Macgregor said: “Ethics and responsibility are not a category unto themselves. You have to think about what you’re trying to achieve from a societal level, but also your existing organisational values and mission, and integrating that into the technology you’re building. It’s not starting from scratch.

“I think the main ingredients of responsibility are to avoid negative inference, building something that works, and maybe most importantly, building something that works for home users. And that kind of corresponds to how we think about the BBC – the universality aspect, as well. And I think the two things you need ub order to achieve that are thoughtfulness – taking a pause and thinking about the impact of what you’re building – and also collaboration. 

“It’s important to bring in different stakeholder perspectives, so that you’re really reflecting that collaboration and different perspectives. From a BBC perspective, responsibility looks like upholding the values that we have as an organisation. So I independence and impartiality are very important to us in an immediate context.”

Robustness of algorithms

Ilya Feige, director of AI, Faculty, commented: “There’s general intelligence and super intelligence. And there’s a whole topic there that people worry about a lot, rightfully. But right now, what organisations really need to have answers to is fairness, explainability, privacy and robustness. 

“I feel like the first three are kind of well understood, I assume everyone’s talked about, or read about them a lot. But maybe robustness is the one least discussed openly, and it’s sort of the endeavour to know when you can trust your algorithm, or when you can have confidence in it. So where is it likely to go wrong? And there are a bunch of different ways that can take place, like, my data distribution has changed and so the model is no longer relevant. It used to be good, but now it’s not. 

“And there are other examples like parts of data space, where the model is just catastrophically bad, or even attacks. So there are examples of ways in which you can attack algorithms and provide data that fools algorithms. So robustness is a big topic that’s discussed less fairness, privacy and explainability.”

Making sense of your data

Mark Wilson, head of data governance, Handelsbanken UK, said: “Once you’ve got your data in silos or in a warehouse, how are you controlling that? How are you extracting the data? Do you have competency in that? So when it comes to governance I’m coming from the angle of have you got control of governance? When people look at the data, do they actually know what they’re looking at?

“It’s quite sad that a lot of organisations are still not on top of this to a large degree. I think they’re sold – you look around tech conferences, the vendors, everyone’s got a dashboard, everything looks great. But have they actually got the fundamentals in place? The data quality controls. There’s governance in terms of lineage, documentation of losses, and dictionaries. So you can sound safe, you can look at this PowerPoint, or this heat map because I can tell you, with 100% integrity, the data behind it is good. And that’s not happening and so many people are just sold on the dashboard.

“Everyone at a company has a responsibility when it comes to data. Who’s using data, which is invariably everybody. Everyone’s either taking data from something to put together a PowerPoint or to produce a management pack, or they’re taking data from something to be sorted and to be put in something else etc.

“However, you’ve got to have a centralised data governance function to set the standards. You should have data owners in place. So your organisational sets of data. This also comes from that doesn’t have customers, products and agreements. Services, fundamentally NHS a bank, National Trust services, customers products. 

So someone essentially has got to make sure there is a data strategy, framework, rules, and that everyone knows the standards the company has. And then somebody has to make that real. Because no one reads this dusty bit of paper. You need to have a chief data officer. This isnt something that should be left to IT. Data is not an IT problem. It is an enabler. 

“The systems will help you do things. But somebody who understands the business needs to be in charge of the data.”

TechEx Global, hosted on September 6-7 at the Business Design Centre, London, is an enterprise technology exhibition and conference consisting of four co-located events covering IoT, AI & Big Data, Cyber Security & Cloud and Blockchain.

(Source: AI News AI & Big Data Expo 2021: Ethics, myths and making sense of data (artificialintelligence-news.com)