Marqo co-founder & CTO Jesse Clark, on the winners and losers of AI, searching the way we think, lessons from Amazon + Stitch Fix, defining 'developer first' and so much more
Aug 8, 2023
auto_awesome
Jesse Clark, co-founder of Marqo, discusses the revolutionary framework they are building that provides search functionality to developers. Highlights include: the importance of searching unstructured data, the concept of search the way you think, the decision to make Marqo open source, and the winners and losers of the AI revolution.
Marqo aims to revolutionize search functionality by allowing applications to search anything with human-like understanding.
Marqo differentiates itself from other vector databases by offering an end-to-end system that includes both inference and transformation, eliminating common frustrations with AI technology.
Marqo's upcoming launch will introduce a managed service, providing convenience and ease of use for developers while open-source options remain available.
Deep dives
The inception of Marco
Marco was created to address the issues and limitations of current search functionality, driven by the founder's frustration with search experiences in 2020. The open-source nature of Marco allows for fast iteration and feedback, while the development focus ensures that what is promised can be delivered to developers, eliminating common frustrations that come with AI technology. By offering an end-to-end system that includes both inference and transformation, Marco differentiates itself from other vector databases that rely on developers to handle these aspects separately. The user experience is a priority, making it easy for developers to pick up and implement, reducing costs, maintenance, and time-to-market for new use cases.
The Key to Marco's Success: Solving Real Problems
The primary goal of Marco is to solve real problems in search functionality. By offering an end-to-end system that encompasses both machine learning models and the ability to curate search results, Marco tackles the issues of irrelevant search results and poor user experiences. Marco's commitment to building a solution that solves specific problems sets it apart from other AI technologies that often focus on showcasing shiny features rather than providing practical solutions. Marco's focus on value creation ensures that it can evolve and adapt over time to meet changing needs and demands.
The Launch of Marco and the Business Model
The upcoming launch of Marco will introduce a managed service that takes care of all the infrastructure management, deployment, and performance optimization for developers. By doing so, Marco offers convenience and ease of use, eliminating the need for users to deal with complex installation processes and time-consuming configuration. The business model centers around providing a managed service for those who value effortless and streamlined search functionality, while open-source and self-deployment options remain available for those who prefer more control. The excitement lies in seeing what users will build with Marco and gathering feedback to further enhance its capabilities.
Marco's Multi-Modal Search Engine
Marco is a multi-modal search engine that allows customers to search across different modalities, such as images and text. It is particularly useful for product search in e-commerce, where users can search for relevant products across various platforms like Amazon or Stitch Fix. Marco's focus is on real-time search, which presents engineering challenges in terms of providing instant results. The system uses AI and transforms information into compressed representations called embeddings, which are then used to compare and retrieve the most relevant results.
The Role and Future of AI in Search
AI plays a crucial role in powering Marco's search capabilities. By using AI models, Marco can augment its knowledge base and improve search results. The system converts text and images into embeddings, which are used to compare query inputs with existing representations and retrieve relevant results. Marco's end-to-end system handles the transformations and enables fast comparison for real-time search. Looking ahead, the future of AI in search involves continual learning and adaptability to user behavior. This allows for personalized search experiences and the integration of new features and innovations.
Searching the way we think with Jesse Clark, co-founder of Marqo
✅Lessons from working at Stitch Fix and Amazon
✅What it means to be truly “developer first”
✅The decision to make Marqo open source
✅Who will be the winners and losers of the AI revolution
Marqo is building a revolutionary framework that provides search functionality to developers, allowing their applications to search anything - text, images, video, audio - with human-like understanding. Marqo makes it possible to do things that were hard or impossible with keyword search, and is poised to completely reshape how we search.
Episode Highlights from Jesse:
“The amount of data is increasing exponentially. A lot of it is unstructured, it’s messy. We’re going to need to be able to search this data, machines will need to be able to search it.”
“We’ve got this tagline: search the way you think. You’re able to communicate very fluently, have it understand, and retrieve really relevant results.”
“Nothing exists today without open source. There’s certainly that somewhat altruistic motivation to give something back after being such a beneficiary. But it’s also just a very good way to get feedback and iterate very fast.”
“We’ve already seen the commoditisation of a lot of these technologies around LLMs. We’ve been very cautious about where we invest on that, because a lot of it is very hard to defend, it becomes commoditised. It’s a race to the bottom and the companies just become marketing companies basically. That’s fine, but that’s not necessarily what we want to do.”