In July, Uber released a new open source AI library called Plato Research Dialog System. A few months ago, Cisco launched its MindMeld Communications AI platform as an open-sourcing solution, after acquiring the company in 2017 for $ 125 million.
Why are so many new libraries being announced? There seems to be a trend for developers to use open source conversation AI and leave behind closed source alternatives. Most of the conversational AI tools that have been developed over the past three years are open source. And companies like Uber and Cisco want to define the default conversational KI stack.
In fact, in recent years the entire AI sector has shifted strongly towards open source infrastructure. The first spark was possibly Google's decision to offer TensorFlow in 201
This trend has now reached the area of conversation AI. Looking at developers' interest in two of the most popular cloud-based tools, it is clear that the number of questions about stack overflow has been falling since early 2018. (Note that botframework is the name of an open source SDK, but it is used to access Microsoft's conversational AI platform, which is not open source.)
While the interest As a result of closed-source tools such as Dialogflow and the Microsoft Bot Framework, monthly installations of the most popular open source natural language understanding systems have been discontinued (NLU) libraries have increased six-fold between mid-2018 and mid-2019. I expect open source to overtake the cloud APIs within 12 months and become the dominant force in the conversation AI.
Why now? In 2016, the chatbot hype reached its peak as companies examined chatbots and language assistants. For the creation of a proof-of-concept, the convenience of a fully hosted solution like Dialogflow is convincing, as it requires very little engineering effort and low up-front costs. Meanwhile, companies in various industries are using conversational AI to solve compelling business issues, and many prefer to control the tools and training themselves.
The market for conversational AI is beginning to mature, especially in banking and insurance. and healthcare. Erica of the Bank of America and Eno of Capital One are examples of leading banks that have assembled large teams to develop conversation AI. Challenger startups like Lemonade and N26 are on their way to building autonomous organizations. This becomes possible as the industry moves from level 3 to level 5.
As always comes with the market maturity differentiated products. There are already several competing open source solutions that address different types of developers. The primary goal of MindMeld is to enable developers to create use cases on Cisco-specific conferencing equipment and the wider Cisco ecosystem. Uber's Plato system works in a different vertical. It is optimized to meet the needs of researchers who want to develop and test new algorithms and evaluate them on standard datasets, much like other research-oriented libraries such as PyDial and the latest ConvLab. Then of course there are tools for developers who deliver products to production. Among the providers are Rasa (where I am CTO and co-founder), DeepPavlov and Snips.
When Open Source overhauls the Cloud APIs over the next 12 months, tools like Cisco and Uber are the dominant force in the conversation AI are just the beginning.
Alan Nichol is founder and CTO of Rasa.