Q&A With Inspired Cognition Co-Founders Graham Neubig, Pengfei Liu & Yusuke Oda
Next-generation Tools for Reliable AI System Development
AI technology is developing at a blistering pace, with mind-boggling tech demos of chatbots, writing assistants, and automatic programming tools that would have seemed like science fiction a few years ago. Moving beyond the tech demo to a system that actually works in practice however is still more of an art than a science. It takes a high level of expertise to choose the most appropriate AI technology for your target task, adapt it appropriately to a particular use-case, and make sure that it works on real data for the target application.
Inspired Cognition is building the next generation of tools for reliable artificial intelligence, which allow practitioners to easily build AI systems that work in practice. We sit down with Graham Neubig (CEO) to learn about what led them to tackling the creation of AI development tools for all.
Could you provide background on your experiences and what led you to this idea?
All three founders have experience building practical AI systems, and we know first-hand just how difficult and uncertain the process can be. We’ve built up some “tricks-of-the-trade” that are useful in figuring out the path forward, but we’ve seen the struggles of developers who are less familiar with developing reliable AI systems. In a way, the practice of AI development today is like the process of software development 25 years ago — buggy and fragile due to lack of appropriate tooling. That’s the problem that we’re trying to fix.
How did you explore creating your product?
Our initial explorations were in the form of research projects at Carnegie Mellon University, where we built open-source prototypes of many of our ideas related to efficient development of AI systems. However, as is common with research projects, these prototypes were not fully mature and lacked a lot of the important features needed to make them useful to developers. Inspired Cognition is taking many of these seedling ideas and building them out to be practical.
The first product that we have released is ExplainaBoard, a tool for exploring the outputs of AI systems. Using it, developers can explore the predictions made by a system, understand where the system is working and where it is failing, and take corrective action. We also have a number of other exciting additions to our suite in the pipeline!
Why are you the best team to champion this product?
I’m building Inspired Cognition, together with the two other co-founders, Pengfei Liu (CSO) and Yusuke Oda (CTO). In order to build this product, it is necessary to have a deep understanding of how AI works theoretically, experience with the practical aspects of how to deploy AI systems, and the ability to convey this experience to your audience. Our founding team includes both world-class AI researchers and educators (Graham and Pengfei), as well as skilled and experienced AI developers (Yusuke), to cover all of these dimensions.
What were some things when developing Inspired Cognition that surprised you?
There have been a bunch of eye-opening things, but perhaps one of the biggest ones is the fascinating variety of problems that developers are applying AI to nowadays.
In building the company, this was a good impetus for us to get out of our comfort zone and talk to many people in many different industries with different use-cases for AI. It’s been really insightful to see the similarities and differences between the different use cases, and this has helped us shape our product design as well.
How did you make sure your product was a reflection of your target audience’s needs?
It’s basically boiled down to talking to a lot of people about their pain-points in development of AI systems and synthesizing them into features or directions for our products. We’ve also deployed our product to users and have been collecting their feedback and issues, which has allowed us to prioritize parts of the user experience that are most important.
What has been some of the customer feedback?
It’s always great to hear positive user feedback, like one of our users looking at the UI of ExplainaBoard for the first time and saying “wow, this is so cool!” This has been really encouraging. But what’s been most useful has been the suggestions that users have given with regards to how we can improve the user experience, which has helped us ruthlessly simplify the user experience. We’re getting to the point where users can be self-sufficient and start using our products without too much help.
How does AlphaLab help you reach your goals?
We’ve only been part of AlphaLab for a short time so far, but it has already been invaluable in providing resources to help us accelerate the building of our product. The AlphaLab staff have provided very useful advice that has helped us prioritize and hone our marketing strategy. The financial support has also extended our runway and helped us focus on building out our product.
What is next?
As mentioned before, we have a couple of exciting products in the pipeline. They are aimed at having broad appeal to a wide variety of developers using AI in their products, so stay tuned.
What is the most important thing for readers to know about what Inspired Cognition does and how that’s beneficial to users?
The most important thing for readers to know is that if you’ve been following the process in AI and want to build a practical AI system that you can actually use in your user-facing products, get in touch with us! We provide software solutions and consulting services that will get it done.
How can consumers follow you?
You can follow us on Twitter or LinkedIn, where we will be putting out announcements shortly. You can also sign up for an account on ExplainaBoard and try it out; we will be sending out product updates via the user mailing list as well!
Outside of work, what do you like to do in your free time?
I love to do things outdoors (hiking, biking). Also, recently I’ve been trying to cook one dish from every country in the world — I’m about ⅓ of the way there right now.
