Cohere, a Toronto-based artificial intelligence (AI) firm that delivers natural language processing (NLP) models, has raised US$40 million (CA$ 50.4 million) in Series A funding. The funding will be used to hire additional staff and expand the existing voice AI platform to other sectors e.g., finance and healthcare.
The Series A Funding Round was led by Index Ventures, with participation from Section 32, return investor Radical Ventures, and numerous AI luminaries, including Turing Award winner Geoffrey Hinton, Fei-Fei Li, Pieter Abbeel, and Raquel Urtasun, who just started her own AI business. Cohere has already raised about US$ 50 million (CA$ 63 million) in total investment. Mike Volpi, Index Ventures partner, and co-founder, will join Cohere’s board of directors as part of the investment.
According to the company, its revolutionary natural language processing (NLP) software gives a more complete comprehension of human language, including semantics, emotions, and tone. As a result, it improves on existing software that supports machine-human interactions like online customer service chatbot chats.
The startup aims to develop NLP models that will be made available to a wider audience, rather than just a small number of internet titans. Cohere claims that their API allows businesses to deploy NLP capabilities throughout their organizations without the need for supercomputing equipment or AI knowledge, insisting that this “dramatically lowers the cost for enterprises of all sizes to access top AI models.”
In May, Cohere came under the public eye when it was revealed that the startup has been working with a few test clients, like Toronto chatbot service company Ada Support, to connect them into its software over the internet.
The startup leverages Transformers technology, which Aidan Gomez first conceptualized in the 2017 paper “Attention is All You Need.” Under the leadership of neural network pioneer, and “AI Pope” Geoffrey Hinton, Transformers was created mostly at Google Brain in Toronto (Canada). As members of the crew at the time, Aidan Gomez and Nick Frosst played a key role in the Transformers’ creation. Gomez worked at Brain as a student intern in Toronto and London, while Frosst was the first employee at Hinton’s Google Brain AI unit in Toronto.
Later, in 2019, they founded Cohere with third co-founder Ivan Zhang to take speech AI to the next level.
Transformers was a massive improvement over earlier sequence-to-sequence models as it employed the attention mechanism. Transformers allow an NLP model to examine the relationships between words regardless of how far apart they are, which is seen as an advance over the encoder-decoder-based neural machine translation method in natural language processing.
Read More: Skoltech Team creates Transformer Based Neural Network that names Organic Compounds
Some of the popular examples of Transformers models are GPT-2, GPT-3. However, according to a recent study, the problem with these models is that they are prone to learning things that they are not supposed to perform. As models like GPT-3 are trained on content found online, any bias or misinformation can cause a point of failure — this will result in the inheritance of bias by subsequent applications too.
To mitigate the concerns, Cohere’s engineers performed quality control tests to search for any flaws with the model before release, and the company continues to monitor its models after launch as well. Cohere will also post “data declarations,” which would include information about training data, its limits, and any dangers.
In an interview, Index Ventures co-founder Mike Volpi expressed his admiration for Cohere’s leaders, saying they “had a real focus on building a usable product that developers could just take and run with” for specific uses like assisting writers with article draughts, powering chatbots, and assisting with content moderation on websites.
