Our Learnings from the Early Days of Generative AI
It’s been an exciting few months at LinkedIn, as our engineering and product teams have been working hard to build some new and advanced AI-powered experiences for our members and customers. I have the opportunity to sit at such a unique vantage point where I get to see first hand the work that went into setting the technology foundations - from the technical resources, tools, engineering playgrounds and guidelines - to make it all possible.
We recently shared some big moments around new AI-powered experiences, specifically leveraging generative AI in our products. We introduced collaborative articles to unlock a new way of knowledge sharing with starter content generated by AI with prompts to drive contributions to the content from our LinkedIn community. We also integrated generative AI into the member profile and hiring experiences by testing a GAI-powered writing assistant in profiles and GAI-powered job descriptions for job posts.
At the core, our unique engineering approach is what enabled our teams to quickly move from ideation to exploration to deployment in less than three months so that we could get the latest AI technology into the hands of our members and customers to help them in their careers. Here’s a look into how we equipped our teams to embrace an entrepreneurial mindset as we build the next phase of our AI journey:
A Strong Engineering Philosophy for Building New Technology - Early on, we prioritized an engineering philosophy rooted in exploration over building a mature final product. The maturity for the right features and experiences would occur over time, but we wanted to make sure that exploration was encouraged by putting generative AI technology in the hands of every engineer and product manager that was interested. This included tools like our internal LinkedIn Generative AI Playground, which allows engineers to explore Linkedin data with the advanced generative AI models from OpenAI and other sources. Based on what we’ve seen with the playground, we brought together engineers for LinkedIn’s largest-ever internal Hackathon, featuring thousands of participants, who brought fresh perspectives and ideas. In the early phases, instead of pushing for something perfect, we wanted to embrace the journey of exploration.
Access and Tools are Fundamental - Interest in generative AI has been very high, but exploring the technology requires thoughtful access and the right tools. We committed to creating tools that inspire our engineers to explore and quickly identify ideas that have the strong potential to benefit our members and customers. We created the GAI LinkedIn Playground and the LinkedIn GAI-tway, which allows access to OpenAI models and cutting-edge, open-source models from Hugging Face. These are two really crucial tools that accelerated our generative AI exploration, and that we look forward to sharing more about in the future.
The Benefit of Knowledge Sharing and Learning Together - Throughout the engineering organization and across the different teams, you’ll often hear about our “OneLinkedIn” culture. Being collaborative is an authentic piece of who we are and a benefit for us when we first took on the challenge of integrating and supporting different uses of generative AI. Because of our culture, we encouraged different teams to share resources so that they could quickly develop in a time when the number of developers who could access certain generative AI models was limited due to capacity. We also passed on learnings from team to team about quotas, access, prompting patterns, and other best practices, so that they could better help one another.
Dedication to Responsible AI - AI has been such an important part of all the products, services, and experiences we provide our members and customers on LinkedIn. That’s a very big reason why we dedicate ourselves to the responsible use of AI, and exploring generative AI is no different. Part of being able to explore the ideas and uses of generative AI quickly has been because we can rely on aspects like our LinkedIn Responsible AI Principles to guide our work. It is a strength that helps us stay true to keeping members and customers first in everything we do.
Seeing the initial reactions from our members and customers to our new AI-powered features and experiences is very rewarding and our learnings continue. Overall, our team's work is contributing to an engineering culture at LinkedIn that values exploration, collaboration, and innovation.
Acknowledgements
Thank you to Praveen Bodigutla, Yanbin Jiang, Saurabh Gupta, Nicolas Nkiere, Xiaofeng Wang, Jie Bing, David Tag, Yi Liu, J.C. Van Ocker, Zhoutao Pei, Eugene Jin, Lei Le, Donghoon Jung, Michaeel Kazi, Jiarui Wang, Sai Krishna Bollam, Zhifei Song, Xiaowei Liu, Xavier Amatriain, Priyanka Gariba, and Sandeep Jha.