Artificial intelligence (AI) is steadily solidifying its position as a key player for digital transformation. As AI forges ahead into every realm of industry and disrupts business processes and changes customer experiences, it’s vital for companies to tap into this revolutionary technology to gain a competitive edge.
Recognizing the value of AI and machine learning in automating business workflows, enterprise AI solution provider Indico has developed an enterprise-grade AI and machine learning solution that solves the challenges surrounding unstructured content – all the documents, images, and text that comprises 80% of the data in most enterprises.
SourceForge recently caught up with Madison May, Machine Learning Architect and co-founder at Indico, to discuss the future of artificial intelligence and machine learning and explain the concept of transfer learning. May also shares how their newest open source project, Finetune, helps developers solve natural language processing tasks efficiently and seamlessly.
Q: Please tell us more about Indico. What is the company’s mission and vision, and who are some of your current clients?
Madison May, Machine Learning Architect and co-founder at Indico
A: Indico provides enterprise AI solutions for intelligent process automation. We’ve built a solution which leverages AI and machine learning to automate tedious back-office tasks and improve the efficiency of labor-intensive, document-based workflows. These business processes depend on a lot of unstructured content, like text and images. In most organizations, this type of content makes up over 80% of their data.
To date, the ability to use machine learning with unstructured content has been limited to organizations with a lot of data science expertise and the ability to provide massive data sets on which to build their data models. This is a huge obstacle.
Indico’s mission is to remove these barriers and to put the power of AI and machine learning in the hands of a much broader segment of enterprises. We are doing this successfully with an approach known as transfer learning, which allows us to train machine learning models with orders of magnitude less data than required by traditional content analysis techniques. As a result, enterprises are now able to benefit from the dramatic advantages of machine learning in a fraction of the time.
The majority of our enterprise client base today is in banking, insurance, and investments. We also have customers in media & publishing, information services, and industrial services. In addition, we have over 10,000 developers using our cloud-based service via APIs.
Q: For anyone unfamiliar with the concept of transfer learning, can you please provide us with a brief overview?
A: Transfer learning is a machine learning method where a model developed for one task is reused as the starting point for a model on a second task. Transfer learning applies knowledge gained to aid the resolution of subsequent tasks. Tasks that would typically require hundreds of thousands of examples can be tackled with just dozens of training examples per class thanks to the use of these pre-trained models.
Common use cases include document process automation, automated contract analysis, content and image classification, and customer sentiment analysis.
Q: Let’s talk about your new open source project, Finetune. What does it offer and who is it best for?
A: Finetune is focused on enhancing the performance of machine learning for natural language processing. It offers users a single, general-purpose language model which can be easily tuned to solve a variety of different tasks involved in text and document-based workflows. This saves substantial time and effort, while also improving accuracy. And it only requires a base level of IT experience.
Finetune is targeted at developers with the minimal machine learning experience. The goal is to get developers the best possible machine learning model with the smallest amount of code and no hyperparameter fiddling. In other words, you don’t need a machine learning Ph.D. to build an effective machine learning model. The interface is designed to be familiar to anyone who has ever used scikit-learn — a popular python open source repository for training machine learning models.
Finetune focuses on solving natural language processing tasks — everything from text classification (e.g. sentiment analysis) to sequence labeling (e.g. named entity recognition)
Q: Why launch Finetune as an open source project? How would you characterize Indico’s commitment and contributions to the open source community?
A: The Open Source community has been the driving force for innovation in machine learning. We’ve benefited from it greatly and embrace it fully. Finetune is one way for us to give back and continue to promote the benefits of transfer learning to accelerate its adoption and reduce the barriers to machine learning.
The Finetune project extends original research and development work completed by OpenAI to address a wider range of problems. OpenAI’s base project provides an illustrative model for increasing the accuracy and performance of machine learning models with natural language content and includes general capabilities for document classification, comparison, and multiple-choice question answering. The Finetune library packages that capability up for easier use and supports additional tasks such as document annotation, regression, and multi-label classification.
Q: Looking ahead, what trends or new technologies do you believe will change and impact the artificial intelligence and machine learning industries? How is Indico addressing these head on?
A: I think there are a few areas that will have an impact in the near term.
- The continued application of AI and machine learning to unstructured data. To date, most of the successful applications of AI have been with structured data; e.g., information typically found in tables and spreadsheets. Recent advancements in areas like Transfer Learning are changing this rapidly enabling enterprises to apply the benefits of these technologies to the 80% or more of data in their organization that is unstructured.
- The growth and advancement of transfer learning. Related to the above, transfer learning is beginning to gain real traction in the marketplace. Its effectiveness has already been proven in the computer vision space and now we are seeing important applications for natural language processing.
- AI as a feature vs. a stand-alone solution. We see AI increasingly being incorporated into other broader solutions. Enterprises are less interested in buying stand-alone AI technologies and more interested in using AI to solve business problems; e.g., to increase the capacity and efficiency of existing business resources, to automate manual processes; to speed response time.
- Augmentation vs. replacement. There has been a lot discussion going on about the fear of AI displacing employees. While this will happen in select cases, in most cases we see the technology augmenting the capabilities of people and making them more productive; allowing them to focus on higher value activities and services in their roles. The analogy I like to use is that Instead of replacing people with robots, we’re equipping them with bionic arms. Not every solution has to be 100% automatable — there’s a huge middle ground where machine learning can simply be used to make people drastically more productive.
Q: What’s on the roadmap for Indico?
A: I prefer not to share specifics about our product roadmap, but at a high level, we’re focused on three things:
- Continuing to make our solution accessible to both technologists as well as the line of business users. The ability for data scientists to collaborate with the subject matter experts in the business is key to advancing enterprise adoption and accelerating time to value.
- Identifying new use cases that are ripe for intelligent process automation. The best applications of AI start with clear business outcomes in mind. We can help our clients be more successful by helping them target the right types of projects.
- Enabling easier integration with existing technologies. As we look to impact improvements in existing business processes inside the enterprise, it’s becoming important that we are able to integrate with the existing systems that are driving them.
About Indico
Indico is a provider of enterprise AI solutions for intelligent process automation headquartered in Boston, Massachusetts. Established in 2013, Indico focuses on automating tedious back-office tasks, promoting efficiency for labor-intensive, document-based workflows, and synthesizing valuable insights from unstructured content such images and text. Indico specializes in data science, API development, and machine learning specifically for unstructured content.