Someone's calling about AI. Graph technology is ready to answer
This is a guest blogpost by Emil Eifrem, co-founder and CEO at Neo4j. He writes on why he thinks graph technology is emerging as a powerful way to make AI a reality for the enterprise.
According to Gartner, by 2025 graph technology will be used in 80% of data and analytics innovations, up from 10% in 2021. The world’s largest IT research and advisory firm is also reporting that an amazing 50% of all inquiries it receives around AI and machine learning are about graph database technology, up from 5% in 2019.
It’s a rise the firm attributes to the fact that ‘graph relates to everything’ when it included graphs in its top 10 data and analytics technology trends for 2021.
What’s clear from these figures is that graph databases are an essential tool — for developers, but also increasingly for data scientists. Google, shifted its machine learning over to graph several years ago, and now the enterprise is following.
From concept to concrete
I predict that within 5 years, machine learning applications that don’t incorporate graph techniques will be the vanishingly small exception. Graphs unlock otherwise unattainable predictions based on relationships, the underpinnings of AI and machine learning. And that’s why the enterprise is going ‘all in’ on graphs and why Gartner’s phone keeps ringing!
Graph data science is essentially data science supercharged by a graph framework, which connects multiple layers of network structures. The graph-extended models predict behaviour better.
Graph databases are also the perfect way to bridge the conceptual and the very concrete. When we create machine learning systems, we want to represent the real world, often in great detail and in statistical and mathematical forms. But the real world is also connected to concepts that can be complex. That’s why graphs and AI go together so well, because you’re analysing reams of data through deep, contextual queries.
Connections in data are exploding
Uptake on graphs is set to continue because data management is increasingly about connected use cases. After all, many of the best AI-graph commercial use cases didn’t exist 20 years ago. You couldn’t spotlight fraud rings using synthetic identities on mobile devices because none of those things existed. And yet they’re everywhere today.
Manufacturing companies would have a supply chain that was only two or three levels deep, which could be stored in a relational database. Fast forward to today, and any company that ships goods operates in a global, fine-grained supply chain mesh spanning continent to continent. In 2021, you’re no longer talking about two or three hops—you’re talking about a supply chain representation that is 20-30 levels deep. In response, many of the world’s biggest and best businesses have discovered graphs as a great way to get visibility ‘n levels deep’ into the supply chain to spot inefficiencies, single points of failure, and fragility. Only graph technology can digitise and operationalise it for that degree of connectedness at scale.
As global digitisation increasingly expands, the volume of connected data is expanding right along with it. We’re also facing more and more complex problems, from climate change to financial corruption, and it’s going to continue. The good news is we now have graph technology to access more help from machines to face the challenging situations ahead.
Welcome to the world of real, practical enterprise AI at last.