Li Feifei, President of Database Products Business at Alibaba Cloud
Generative AI is rapidly shifting how companies process and store data. More companies are using vector databases, which store data as vector embeds in multi-dimensional numeric formats to handle generative AI workloads.
Research and advisory company Forrester estimates that the adoption rate of vector databases at 6% will rise to 18% over the next 12 months.
We caught up with the President of Database Products Business at Alibaba Cloud, Li Feifei, to understand how databases have evolved to fit the demands of enterprises in the generative AI era.
The following conversation has been edited for brevity and clarity
Q: How do database products help businesses leverage AI?
A: Human beings explore the world around us using the DIKW model (Data, Information, Knowledge, Wisdom). Starting with data at the foundation, then information is built on top of data. Then you explore and build knowledge out of the information. Then, at the top is wisdom.
AI is essentially bringing us from DIK to wisdom, from data to information to knowledge, all the way to wisdom. But don’t forget, data is at the bottom. It is the foundation. So, databases are a key element of that data layer, what we call a data platform.
As we accelerate towards the AGI (Artificial General Intelligence) phenomena, databases, more generally known as data platforms, will still be a critical element of that endeavor.
Q: Any examples of organizations successfully using database technology to accelerate their AI initiatives?
A: I will start with the entertainment sector, in particular online gaming. We have one overseas customer in Southeast Asia using our cloud-native databases, such as Polar DB and our data warehouse Analytic DB.
But when it comes to AI, what they’re really interested in is AIGC (AI-generated content). They started using the vector engine embedded and integrated with Analytic DB (ADB). In ADB, we have built and integrated a cutting-edge vector engine. And not only do we have a vector engine embedded and integrated into ADB, but we also integrate it with one of our large language models, for example, Tongyi Qianwen.
We host all of their transaction and business data together with a vector engine and large language model. Suddenly, you can convert the traditional Business Intelligence (BI) type of applications into more intelligent applications. You can query the database engine in natural language. And not only can you interact with databases using natural language, but also you can also use natural language to direct databases to generate content. For example, you can generate pictures and semantically meaningful conversations, you name it.
As you can imagine, this is extremely useful in the gaming context. For example, you can have intelligent NPC (Non-Player Characters) in your game design. That’s just one example.
Q: Any recommendations for businesses when it comes to choosing a database?
A: We utilize AI technology and cloud-native technologies. We orchestrate those different database engines and build what we call one-stop platform solutions so that our platform can recommend the best type of database based on your business scenarios and workloads.
And it can automatically set up and run that instance for you without you doing all the hard work. This is what we call the future, or next generation, of cloud-native database systems. We dub it a one-stop intelligent data platform.
In summary, these are really exciting times. There are a lot of versatile database systems out there. You want to choose the one that is the most suitable for your business workload, and you want to look for a vendor that has the ability to automatically recommend, and even automatically orchestrate, the best database instance for you, based on the diversity and the changing needs of your workloads.
Q: What’s the key challenge when it comes to implementing serverless database solutions?
A: Serverless is definitely an exciting technological advancement. In the past, cloud service products basically have to provision a set of servers, for example, four core eight gigabytes of memory.
But that comes with a cost. Basically, what happens if you provision a server that has more capacity than the actual workloads require? You end up wasting your server resources. Serverless is designed to precisely address that challenge so that the capacity of the server resources used by your cloud service matches the needs of your workloads and it adapts to the dynamic change of your workloads.
But there are challenges and pitfalls you need to watch out for. One particular challenge associated with serverless products is that you started with an initial quote for the provision of the servers. However, if your workloads are changing very dynamically over time and even overshoot, then going serverless might end up costing you more than having a fixed quote.
Q. How can this challenge be overcome?
A: As a cloud vendor, we are pushing out a set of new features that can automatically switch between the two pricing models so that we are watching out for you, in the sense that we are monitoring your workload and your billings very closely.
Combining serverless with AI-enabled technology, we are really going to be able to use and build applications onto the data platform just like playing with Lego because serverless allows you to deploy and run applications very quickly, very efficiently, and AI enables you to make, so-called intelligent decisions. Combining these two, you can imagine a world where building applications, deploying them on a data platform is as easy and as effective as playing with Lego.
Q: How do you see the trends of database development, especially in the age of AI?
A: We really need to integrate AI and data very tightly. We believe AI will enable databases to be more powerful. At the same time, databases will form a foundation to support AI to better advance the knowledge of human beings.
Another really critical piece in this puzzle is computing power. Cloud computing provides computing as a utility, making computing easily accessible to a lot of different applications so that you can use data, use AI anytime, anywhere you want.
To summarize, I truly believe what we need to do is to combine AI, data and computing. Together with those three elements, databases will truly become a one-stop intelligent data platform in the age of AI.
