In an interview for listed media and technology group Naspers, the company’s global lead for data science and artificial intelligence (AI), Euro Beinat, has outlined his perspective on AI’s true potential.
What’s the difference between AI and Machine Learning?
Artificial Intelligence is a term associated with machines or algorithms that mimic cognitive human capabilities, such as learning, predicting or acting with common sense in response to their surrounding environment. The current wave of AI is much narrower and concentrates on Machine Learning, a set of tools that learn complex patterns from data.
These tools have been successfully used in natural language processing, translation, visual perception or anomaly detection. Applications are found in virtually every sector: retail and supply chain, e-commerce, news, media and entertainment, finance and credit, and many more. Because of this, machine Learning is widely recognised as one of the most important technologies being developed today.
Why are Machine Learning & AI so important, globally?
They are horizontal enabling technologies, in a way similar to electricity. The paradigm of “learning the code”, instead of “coding the code”, is very general and broadly applicable: if there is sufficient data we can develop systems that learn and adapt. These are fundamental changes that have the potential to transform, enhance and disrupt global business across multiple industries.
The opportunities we’re especially interested in are for turning products and services into curated, personalised, relevant, customised experiences. These technologies will also allow consumer services to be optimised and will help us better use data to empower and enable communities around the world.
What are your immediate priorities at Naspers?
My team will support the design and implementation of Machine Learning & AI strategies and solutions across the group and will develop and prototype new solutions. We will also review the potential to better utilise data across the group to benefit our customers.
Additionally, there is an ongoing debate about best practice around artificial intelligence and we’re keen to play a role in encouraging a global policy that can support the ethical use and development of the technology. Businesses, governments and communities have really only scratched the surface of what is possible in terms of Machine Learning & AI – we will actively engage with R&D institutions globally to explore and advance the area further.
What are the implications of artificial intelligence for data privacy?
Part of demonstrating expertise in this field is showing how the positive benefits of Machine Learning & AI can be achieved in a privacy-friendly, ethical and accountable manner. Robust practices have to be developed in line with individual business priorities.
For example, at Naspers, in addition to GDPR compliance, the nature of our business is such that we are developing an internal governance framework for data exchange and processing in applications. Our guidelines include, amongst others, transparency requirements, application principles, data security, data and model retention policies, anonymisation and related data privacy practices.
What do you believe to be the most exciting development in AI or Machine Learning over the past year?
It is hard to choose: the field is growing very rapidly with creative, unexpected uses appearing daily. Also, new algorithms and tools are published continuously and each one moves the science further. There are, however, a few developments that I would single out. Transfer learning, for instance, has become mainstream, especially in image recognition.
The trick is to use state-of-the-art algorithms, trained to recognize general objects like cats or cars, and upskill them to recognize very specific image objects, such as car parts or brand of watches. The net result is a much better computer vision tool trained with less data.
This is extensively used by the OLX Group, Naspers’ classified business, to describe posts automatically: it increases search relevance significantly and improves the user experience. Text and speech technology have also advanced very rapidly, crossing the point of being sufficiently precise to replace and augment many human interactions with machines.
What are you most excited about in the coming year?
The most exciting thing is that we are at the very beginning of a wave. In broad terms we can see this shaped by Machine Learning & AI, Internet of Things and Blockchain, with different levels of maturity. In addition to Machine Learning & AI developments that are specifically useful for Naspers, such as Federated Learning, Differential Privacy or Synthetic Data, we are closely following the development of hardware for machine learning, in particular the development of chips that will enable every device to learn at the edge of the network.
It is the bridge between AI and IoT and will change the way we design AI systems shifting from the cloud to the edge. We also see an increased democratisation of Machine Learning /AI tools, and this will enable more of our teams to innovate with machine learning, something we are very keen to facilitate and accelerate.
Read the interview at Naspers, here