Stories

By Hui Wang, Vice President Global Data Science & Machine Learning, PayPal

The recent developments around generative AI (defined as algorithms that can be leveraged to generate new content) have created an incredible opportunity for nearly everyone around the world to leverage the power of data, artificial intelligence (AI) and machine learning (ML) to drive the next wave of digital transformation. At PayPal, we are excited by the potential of generative AI and are actively exploring, assessing and executing on opportunities leveraging the technology that can improve customer experience, drive operational efficiency, and help us create a more customer-centric business.

Earlier this month, I had the privilege of hosting a conversation between Sri Shivananda, EVP and Chief Technology Officer at PayPal, and Ali Dalloul, Worldwide VP of Customer Experience Engineering for Microsoft Azure AI Platform. Sri and Ali shared their perspectives and insights around generative AI and how companies like PayPal and Microsoft are leveraging the technology. In the post below, I’ve outlined the discussion and some of the key takeaways.

AI isn’t new, it has been around since the 1940s. Over the past 10 to 15 years, however, there have been major breakthroughs in machine learning (a branch of AI and computer science that uses data and algorithms to imitate the way humans learn) and deep learning (part of the broader family of machine learning that is based on multiple layers of artificial neural networks with representation learning), which has brought us to where we are today with large foundation models.
 

So why is there so much focus and energy on AI right now? What’s changed?

Ali explained that, over the past decade, there have been three major breakthroughs around AI:

1. The economics of data has changed. The marginal cost to acquire data right now is very, very low, and it’s available everywhere – from user-generated content to entire-web-scale data.
2. The breakthroughs in compute silicon. Super computers today can actually mimic or even exceed the human brain's synapse processing power. We have the computational ability today to process these large models in a way that is unprecedented.
3. The breakthroughs on foundation models and transformer architectures. The work that Stanford did on foundation models and Google’s work on transformer architecture became the genesis for large language models that OpenAI advanced to another level and created the Generative AI revolution.
   ​

So where are we today with AI? And where are we heading?

The three factors above have brought us to a point that is truly unprecedented in the industry, and it has yielded three important outcomes:

1. The economics of the industry are shifting. We now have the entirety of human knowledge inside an API called GPT4, and it’s accessible to nearly everyone, enabling grassroots innovation in a way that was never possible before.
2. The barrier of access to technology has been lowered in every domain for nearly everyone. This wasn’t previously available in AI.
3. We are now seeing a massive boost to augmented human productivity and catalyzing creativity because, for the first time ever, we have technology that is mimicking what has been traditionally done by humans.

AI could enable us to boost the productivity of our economies. It could also help people become more skilled by flattening the learning curve between very skilled and unskilled employees. AI also has the potential to diminish a lot of our time spent on basic repetitive tasks, allowing us to be more productive and creative.
 

What role will generative AI play in technology companies?

When thinking about ways to leverage generative AI, it's important to think holistically, not just about a specific function, but end-to-end across all functions within a company. It's important to think about how we can improve our customer experience by better understanding them and their needs, and catering to those specific needs. By leveraging generative AI, we have the potential to improve everything from customer support and engineering to compliance and accounting.
 

What advice would you give to companies that are trying to transform their business with generative AI?

Ali summarized his thoughts in three buckets:

1. Understand what generative AI does and doesn’t do. Generative AI works really well with language, unstructured data, and at a generalized level of understanding. But it does not traditionally work well with structured data, numbers, and is dated on recency (due to model training dates), and referential integrity (i.e., citations reference); both of which are solved through Bing (for consumer search) or Azure Search (for enterprise search). So, it’s important to put into place solutions for some of these areas where generative AI doesn’t work well on its own.
2. Think carefully about where you want to derive value. Just adding an AI label on your brand won’t work long term if you don’t build a proper foundation.
3. Realize that a cultural transformation must happen to truly embrace AI. Ali said, “Even at Microsoft, a tech company DNA to the core, the amount of cultural transformation we had to undergo inside Microsoft was not trivial. The technology is easy. But the question is, ‘Can you change the company culture to reframe the mindset and embrace AI in a way that actually serves the organization?’”
   ​

How can leaders encourage a culture of innovation, as it relates to AI, within a large technology organization?

It’s important to take a 360-degree approach to innovation – not just for AI but for any technology trend. Sri shared his perspective on how to encourage a culture of AI within a large organization.

1. First, certain teams will inherently be at the forefront of enablement for the entire company, and they will need to create a foundation where people can unleash what they can do with the context they have.
2. Second, while leaders should encourage this culture, bottom-up inorganic innovation is extremely important. Each person within an organization has the context of their domain with a much better depth than anyone else. So, everyone needs to participate in the shift, not only in terms of learning, but in terms of applying that into the various contexts and coming up with the best ideas within a specific domain.
3. Third, it's important to explore across multiple time horizons. It’s also important to plan out the immediate future as well as plant some seeds that go beyond a typical three-year time frame.
   ​

How do you handle the “negative side” of AI?

“At PayPal, we have always been a customer champion,” Sri shared. If we always start with the customer and we think about the customer and what’s in their best interest and how we build a long-term relationship with them and protect them on this journey, that will guide us to always do what’s right. We’re not necessarily in the business of payments, we are in the business of trust.”

AI is a very powerful technology that has also created a lot of fear and uncertainty, but having a customer mindset as a company brings with it a lot of responsibility and accountability. No company is going on this journey alone, the whole world is going through it together. Companies like PayPal and Microsoft and their partners and communities need to come together to apply the guardrails of what it means to be a civilization unleashing the power of AI technology. Including the right ethics, the right constraints and controls, the right limits, the right understanding of what’s good and what’s not good guiding this technology. Technology at the end of the day is an enabler, it’s not the destination.
 

What skills do people need to acquire to be ready for this technology shift to AI?

Ali and Sri both agreed that the most important thing is to be willing to continuously learn. Technology will evolve. Methods and architecture will evolve. Learning to learn will set you up for success on the ongoing AI journey. It’s also important to become more comfortable with data. Data literacy is going to be very important and data science will be a foundational skill for all engineers. Leveraging AI and ML models will become a primary design consideration. It’s also important to tinker – just get in there and start to play with it. Everyone will need to be thinking about exploratory research that needs to be done. And, finally, fail! If you’re not failing, you’re not learning. Learn from those failures, iterate over and over again. That’s how you leverage the power of the generative AI trend.
 

The content of this article is provided for informational purposes only. You should always obtain independent business, tax, financial, and legal advice before making any business decision.