Join the conversation:
The first of three episodes recorded at the AI and Big Data Expo, that we produced in partnership with Exfluency, the AI driven translation and localisation system. The event took place at the RAI in Amsterdam on 26th-27th September . We recorded a series of interviews from the Exfluency booth with a number of the speakers and attendees at the event. Our guests for this episode were:
Gayle discussed the key components that Netflix pull from the online memo as freedom and responsibility, which is a unique way of leadership and management of individuals across the company. He said anybody in the company has the freedom to do anything they wish as long as it’s in the best interest of the company and then they have the responsibility to own those decisions. Gayle said the other piece of that is Farming for descent, which is unique to him due to his background in the Marine Corps and he loves getting feedback and giving feedback. He said at Netflix farming for Descent is asking for feedback after doing a presentation. He then went on to suggest that feedback should be direct but purposeful and meaningful to ensure the person is constantly getting better. Gayle also talked about guardrails which means that all colleagues have access to every employee at Netflix which takes away that rigid structure company and makes sharing information much easier.
Gayle said a traditional business continuity program would usually have one business continuity person and for every 2,000 employees at Netflix, there are two people. One being himself and the other is TJ Mead. Gayle does the business side and TJ does the technology side of things in relation to continuity. Gayle says that the culture at Netflix is too fast or dynamic to use a traditional model so they took core requirements out of each step in the traditional program and built a quantitative process of basically knowing what the current health of the organization is.
Gayle talked about how enterprise continuity existed basically as a result of Covid when Netflix was trying to figure out how to work from home like other companies that were going through the same challenging time. Continuity allowed Netflix to focus on just dealing with Covid and making the workforce more resilient, he went on to say that since continuity has evolved into connecting the business to the technology at Netflix to make sure everyone has a clear understanding.
Gayle said that due to being a relatively new team they are only now beginning to do reassessments to assess everything they have been doing for the past two years and as a result, they are able to identify changes in behaviours. He discussed that some scores have gone up and some have gone down which is not necessarily a bad thing. His role is to capture the context so that everyone has a clear understanding.
Gayle said 100% AI could be a big part of business continuity in the documentation of plans and then testing and evaluating the plans. Gayle spends a lot of time working with different scenes creating different scenarios and he believes AI could make this a smoother process allowing him to evaluate and test in short time periods.
Netflix are acquiring more content and licensing and things of that nature. He said that as a technology company that does entertainment, they want to make sure that the customer is always able to watch whatever content they want and discover new content. Therefore the goal and the continuity program are currently built to make sure that leaders and the subject matter experts for those different services and applications, can do that regardless of the types of disruptions that may impact us as a company.
Gayle said he is very confident. He loves it and believes in it. He says Netflix sees value in his job and therefore he feels his time is well spent.
Mahmoud said that data lineage basically means, the art of traceability of data. You can identify a specific piece of data that you have in your organization by spotting the origin of it and where did it go and what happened in between?
Mahmoud talked about the potential, and that it can be very useful for you to understand not just the current way of doing data. Mahmoud said to imagine that there is a process that does a certain calculation in order to calculate your net profit, and this equation has a problem on it. Because the wrong attribute has been used to calculate it. How would you know that? If you have a possibility to track and trace how the main elements that this formula came from, that can be very useful. It can also be very useful in doing migrations or modernizing the data. The trend nowadays is moving to the cloud and most organizations have a big, bulky, old fashioned data and data lineage can help you to slice and dice the data that you currently have in order to have a proper strategy to do the migration to the public cloud.
Mahmoud said Booking.com’s footprint from a data point of view is huge and on daily basis, they have to deal with terabytes of data coming from the logs, from our website and app. You have to adapt and you have to have the flexibility to integrate multiple technologies in order to reach the goal of having an enterprise data lineage which can be costly.
Mahmoud goes on to talk about a suggestion he made in a presentation that he did at the event which was to support this by a backbone kind of architecture called the Data Lake or the metadata Lake, where you can store different types of metadata for the boosting of the data lineage initiative at scale at your company.
Mahmoud suggested to backbone it with an architecture, that can store different types of metadata because metadata is the key to unlock your lineage initiatives. He said try to think of how you get the business metadata, operational metadata, social and technical metadata, put them together in the data lake and then try to fit them together and try to find a way to traverse between them, that can be enabler for you to do your data lineage.
Mahmoud said it’s the heart of what we do at Booking.com, not just nowadays, but since the beginning because booking.com tailor our experiences anyway to the traveller or the person who use our service already using a lot of machine learning. Every single product team has a machine learning or an AI engineer next to it.
Mahmoud talked about the AI trip planner which is the usage of OpenAI and the language models in order to offer a different way to interact with our solutions. You can open a chat and you just naturally chat with and then it’s like a conversation that you’ll have with the computer to suggest you stuff. Booking.com have built a deep integration between these models and their models to ensure a seamless experience at the end with the customer.
Jaromir said it was a natural decision. Since Exfluency are into AI models and large language models, due to being in the translation domain. It was an obvious choice for Exfluency to attend this AI & Big data event because everyone is talking about AI, but Exfluency really want to listen, talk and partner up with companies.
Jaromir said it is a bit of provocative to inspire discussions. HI stands for hybrid intelligence, and that’s a great definition of our solution that Exfluency are not yet another artificial solution. The kind of solution Exfluency actually has, goes through technology all the way up to the human level where Exfluency has built communities around the concept. This hybrid intelligence stands for community members that do the data validation in order to create immutable data, in order somehow to provide a way that the data speaks for themselves and reflects the reputation and reputation written in the blockchain. Jaromir said it is the perfect combination of union between the machine and the human.
What are your other key objectives? Jaromir said that Exfluency are here to understand more about the problems that companies are facing right now. Everybody is so excited about AI and new opportunities and challenges that were discovered by some of the people in late November, maybe December last year. While of course, the technology is much older than that. So Exfluency want to engage people in understanding more also to make them realize they are not only opportunities but also challenges and the risks related to implementation. And Exfluency want to help companies to implement AI in their organization in the right way.
Jaromir talked about LLMs which can bring a lot of opportunities, they can easily generate content for you, they can summarize, they can basically save a lot of time, but at the same time, this opportunity creates a lot of risk. For example, it can result in totally loosing control over data and as a result can end up exposing sensitive private corporate data to potentially competition. Exfleuncy can help with this and create a tailored specific model that is effortless. It would allow a business to do things in the same way for example chatting to ChatGPT but the engine would be data specific.
Jaromir said for example, personalized marketing, in case of LLMs which is built on your data, that’s actually your model, that’s your specific model so that you can use these data to create localized marketing campaigns that will improve customer engagement. Another example would be where there is a model that never sleeps. It works 24/7, for automation and training and reacting to events that could actually create content, create notifications while you’re asleep.
Sabiha described the panel was talking about how do we safely scale AI across organisations, mainly focusing on ethics, governance, responsibility, accountability and how can we govern AI without stifling innovation too much?
Sabiha said the main points that were discussed – can we balance accuracy and explain ability or are there trade-offs offs and what are the trade-offs between bias, fairness and also accuracy? How do we make models more accountable, more fair?
Sabiha said that in most cases biases come from AI is trained on historical data. History has its own prejudices, own biases. But our morality and moral compass of society are changing. So we are mimicking past behaviour which is not acceptable anymore which needs to be changed. She said we need to be carefully inventing the future, not just mimicking the past behavior and another source which bias comes from is data scientists who are building they are taking decisions on data processing, data sampling or methodology choice, evaluating models. So consciously or subconsciously, the biases they have are creeping into the algorithms.
Sabiha said it depends on the sources of the bias, where the bias is coming from. So if the data is not representative enough of the whole population, then it’s important to have diverse set of data to include all different demographics of the population on which the model is going to be used for and have diversity in data science teams so that the biases are easy to spot and easy to mitigate.
Sabiha suggested an example, biases against women, bias for gender, age, different ethnicities. She said she works in a bank and we use models to determine to give loan to someone or not, because in history, women had lower income and higher chances of default. If we train on the same data, even today, our models could be biased against women.
Sabiha has seen a lot of exciting use cases in health care and climate change mitigation. Sabiha went on to say as Hollywood has taught us, with great power comes great responsibility. So we are in a time where we can shape the future of AI because it comes with huge responsibility as well.
Sabiha said there are regulations coming. One is AI Act and all Member States in EU will have to comply with that. Third-party auditing will be required. We have a lot of developers and data scientists now budding, but do we have enough skilled auditors?
Sabiha talked about how technology moves much faster than regulation and law so lawmakers have to be on their toes as much as the technology makers.
Michael said AI and Partners is a professional services firm for companies subject to the EU AI Act starting in 2021. It’s an archetypal professional services firm geared to help companies achieve regulatory excellence.
Michael talked about how innovation, generally speaking, outpaces regulation, but it’s necessary to catch the pace of innovation at the moment. It’s a risk-based horizontal approach to the regulation of AI, and it’s based on the risk of an AI system. Other jurisdictions have different approaches, but the EU has set the precedent with this.
Michael said it depends on the scale, nature, complexity of the business and the level of maturity. As with all technological changes, AI is going to be interesting. He thought world leaders such as Elon Musk and Bill Gates talking about it, has to be considered in a whole different spectrum and we’re entering a new paradigm.
Michael said it has been busy and the team have enabled this to happen. The team have been absolutely exceptional. The opening remarks were based on where we see the intelligence and ability to adapt. The sessions, had great panelists, had philosophers and risk management personnel from banks giving real insight on how to scale AI and of interest was the ethics piece and how frameworks can be implemented in banks. There’s so much to talk about, it’s difficult to cover everything.
Michael answered that we are due to recognition that needs to be it’s important and also no, there seems to be a separation of what it means to be ethical and if it’s part of responsible AI or completely separate. He said there is still a lot of ambiguity there that needs to be determined.
Micheal said he speaks with prejudice but he is excited about the opportunities in financial services and from compliance, whether it’s everything from fraud detection to trade surveillance to automation of KYC and onboarding processes
Michael quoted Einstein ‘Intelligence is the ability to change’. He thought investing in change, adapting and driving the culture to change, not being scared of change and actually taking maybe almost a from the entrepreneur’s playbook. Every day is a new day to start something new.
Michael said there was actually a report released yesterday by the Association of Financial Markets in Europe, and they think how it augments the compliance function to some may not be as exotic, but Michael thought the ability to equip new compliance officers in the space of the tools to do things that were unimaginable.
Joris said that banking traditionally has been a very data-heavy industry. So he thinks artificial intelligence holds a lot of promise for banking and both the front end and the back office part, AI can revolutionize the way banking is being done in terms of marketing, but also risk management, compliance, these sorts of aspects. Joris said artificial intelligence can have a tremendous impact and we’re already seeing that today.
Joris talked about responsible AI in terms of ethical perspective and what is desirable.
He said responsible AI is all about making sure that the systems you develop and deploy are aligned with human values and things we think are important and impact as a bank that we want to have on society in a positive sense. So responsible AI means aligning your AI systems, but also the structures and processes of developing and deploying them with these human values.
Joris said when it comes to discrimination or fairness issues, AI is basically automating the status quo and that doesn’t work out equally well for everyone. So you have the potential for bias there and negatively impacting people, for example, comes to credit scoring or mortgage approval. He went on to explain the problem that these systems are become increasingly more complex and this explain ability issue, can we still understand the reasoning of an artificial intelligence? He also mentioned AI in general is the accountability issue where because these systems are so autonomous, we cannot actually locate responsibility to the programmer, the team, the business owner.
Joris said that one important step is to mainly look at this problem as a technical problem which he said was a misunderstanding of the problem of responsible AI to think you can solve it with more data, better technology, or to just train your data scientists and assume that things will be okay. And so what needs to happen is a proper form of governance. So you need to align the structures and processes with the values you consider important for artificial intelligence. He said that is what they have done at De Volksbank, they have changed the development process. De Volksbank have a new ethics team that is particularly looking into the ethical aspects of their AI systems throughout the life cycle of an AI system.
Joris said ethics is about justifying the principles that underlie your decisions. He said when it comes to business ethics, you need to be able to be held to account for the decision that you’ve made as a company, as an organization and these decisions need to be somewhat aligned with public values. De Volk bank cares not just about the shareholders, but about the customers, society and their own employees as well.
Fatima said it is important for us to focus on collaboration, whether as an organization or an individual or even at the country level, because that’s how the world is evolving, evolving, especially in the space of technology. She explained technology as borderless, and it connects individuals and nations and communities.
Fatima talked about the huge access to talent due to the nation being almost 240 million with 30% of that population being under 30. She said as the citizens of the country it is their responsibility to create jobs and create opportunities as well. However, she talked about how technology platforms enable you to be sitting in Pakistan and connecting with all over the world. Fatima talks about how the economy has taken a bad hit whether it is political instability or whether it is the macroeconomic indicators, and also suffering from floods as a consequence of climate change. The states in every crisis there is opportunity. Fatima talked about the governments own commitment towards engaging with the tech industry.
Fatima talked about the footprint of Abacus. She mentioned the intent of associations both on the private sector as well as the government owned entities, a testament of the fact that the government of Pakistan has prioritized the technology industry as one of its mainstays in terms of enabling the economic activity and also trying to get back on our feet after the last few years, which have been a bit, difficult. She said the government is pushing the private sector entities to represent even in the US, the tech crunch took place in California last week. Fatima said is it about the government saying, we’ve got this opportunity, we will sponsor you, we will subsidize, just go ahead and connect with potential stakeholders.
Fatima said the board nominated her as CEO two years ago in 2021 and she has been associated with the professional services industry for almost 25 years. Her journey has been through Coopers and Lybrand as well as Pricewaterhouse Coopers, and it’s sort of intertwined the journey of our organization and her own rise, she doesn’t come from a technology background, her own discipline has been in the human resource side which involved human capital management, change management, governance, more from an HR perspective. She said she just grew into the role of a business leader by virtue of being appointed, being put on the board. Her role as a member of the board gave her great experience in terms of overseeing a diverse portfolio. Abacus is now 35 years in operation and they were established in 1987, Abacus started from corporate finance and consulting, human resource consulting. In Pakistan, Abacus was one of the few companies in the world that had the ability to decide whether to go independent or to be acquired.
Fatima talked about how this was a very difficult decision because there was a pivot which meant there was a jump from international to local and lost 50% of our clients, 60% of our people, one office, 60 people. Fast forward to 2023, where Abacus are 4500 people, ten different businesses, predominantly technology and outsourcing. They have one of the largest BPOs contact centers, as well as customer experience setups, as well as obviously the technology portfolio and digital transformation and cloud emerging technologies, etcetera. She said human capital element is extremely important and why it continues to be important is when you have digital transformation, then you have to take the people along and you have to focus on as business leaders of reskilling and upskilling and preparing them for the change that will impact them at a personal as well as at a professional level. She goes on to discuss her current challenges. She then talks about their presence across 28 countries and they work with our international firms which are headquartered in the UK. So her job is to make sure their delivery centre of almost more than 4000 people is running, very efficiently and effectively to cater and deliver across four continents.
In partnership with
PRCA members receive 10 CPD points for listening to this podcast if they log it on the CPD programme.