Startups like to use the term ‘Artificial Intelligence’ when selling their products to describe a range of different things, such as intelligent ways to automate a workflow. When it comes to the more ‘traditional’ definition of AI–Machine Learning–there are still tons of startups that use it, but interestingly not nearly as much independent makers building Machine Learning powered products.
Building and supporting indie-made Machine Learning products is a good chance for the maker community to democratize the access to AI. Right now, many big tech companies, such as Google, Amazon and Apple, are providing the majority of AI services.
It’s important to have an independent alternative to those services. In the same way that there are now indie and privacy-concerned alternatives to Google Analytics, making Indie AI is great to distribute the access to new data, open research and more diverse products.
In this article, we are going to shine a light on a handful of indie-made AI products, which are utilizing Machine Learning and Deep Learning for different use cases.
TypeFont by Vasile Peste
TypeFont isn’t your typical indie product. It is an open source library created by Vasile Peste, which was hunted on ProductHunt a little over a year ago. Typefont rose to the top with nearly 800 upvotes and over 1,400 stars on GitHub, securing a Top 3 place on launch day.
TypeFont analyzes the font used in, for example, a book cover or a cereal box and identifies it within a database of available fonts. It is a great product for designers, typographers and web developers alike.
The open-source project is also a great opportunity for young makers to turn this into a ‘real’ product. If you are interested in Machine Learning and AI, turning this project into a mobile app, for example, could be a great first step in placing a finished Machine Learning product on the market.
Gladys by Pierre-Gilles Leymarie
The Gladys project is another open-source project. Even though it had humble beginnings, the Gladys project has generated around $20’000 in revenue last year through donations and training material – not bad for a project that started on a Raspberry Pi in a student home!
The Gladys project is an open-source home assistant, running on Raspberry Pi. Think Alexa/Google Home, but entirely open and free! Gladys can integrate with your calendars and a range of Internet of Things devices, making it an easy to automate your whole life.
This project is especially interesting since every bigger tech company is trying to push their own home assistant. Having a open alternative to those assistants enables users to have a say in the direction of development, make sure the product is ethical and finally leverage the community to ship features faster.
According to the founder Pierre-Gilles, Gladys started as a side project a couple of years ago in his student home in France. He used his newly gifted Raspberry Pi to control switches and automate a few things. The project quickly became viral in the French media, driving over 1’000 downloads in its first version. In 2017, the Gladys Project got monetized through video courses, and then via Patreon in mid-2018. Today, the Gladys project already has over $35,000 worth of downloads. Impressive!
Lamina by Yan Chummar
Lamina was just launched recently on ProductHunt, garnering nearly 400 upvotes and shoutouts from members of the ProductHunt team! Lamina is a way to easily add AI and Deep Learning to your own products through its ready-to-use APIs.
The product was built by 15 year old Yan Chummar, who wanted to add sentiment analysis (e.g. if a review is negative or positive) to his app, creating an easy and inexpensive alternative to big tech APIs and a welcome solution for model deployment in the process.
At the moment, Lamina offers sentiment analysis and entity extraction (turning a block of text into structured data) as API endpoints, with more functions to come. Especially for social and content-heavy sites, the sentiment analysis should be something that makers should look into!
The future of Indie AI
With more people picking up Machine Learning skills and more and more easy-to-use solutions on the market, we should see a rise in high-quality AI products in 2019. There are a lot of low hanging fruits (think about open source projects that are difficult to use) for AI-powered products.
The development of independent Machine Learning products can be difficult, but with easy-to-use frameworks like Keras getting continuously better and offering easy-to-use pretrained models, building ML-powered applications is going to become a lot easier, even for Machine Learning novices.
Machine Learning is a super interesting topic with new innovations popping up daily. If you are following the ecosystem and new research, you should start seeing opportunities for indie makers to create focused, ML-powered products very soon!