Listen to this story
R&D at major innovation labs worldwide are key sources of building better AI systems. This is research that happens at scale and requires good financial support, compute power and other resources. Names that come to mind – DeepMind (founded in 2010 by Demis Hassabis, Mustafa Suleyman and Shane Legg and acquired by Google in 2014 for $650 million) and OpenAI (founded in 2015 by Sam Altman, Tesla CEO Elon Musk, Greg Brockman, Ilya Sutskever, Wojciech Zaremba and John Schulman with $1 billion in funding over two rounds). Of course, the traditional big names in tech such as Meta, Google AI, and Microsoft are actively contributing to AI research too.
But along with these for-profit companies (OpenAI was a non-profit company till 2019, now it’s not), many big and small non-profit AI and ML research labs have also come up that do not want to keep innovations restricted just to themselves or a select few but make its benefits available to everyone.
Sign up for your weekly dose of what's up in emerging technology.
Recently, Cohere, a Canadian startup that builds LLMs, has launched Cohere For AI – a non-profit research lab. Cohere says the lab seeks to solve complex machine learning problems and will focus on creating more points of entry into AI and ML research. The lab will be headed by ex-Googler Sara Hooker.
“My hope is to create the lab I wish had existed when I entered the field, and I’m very excited to have the opportunity to lead this effort,” she added.
Cohere’s co-founder introduced us to Transformers
Cohere has quite an interesting background. It was founded in 2019 by Aidan Gomez, Ivan Zhang, and Nick Frosst. Gomez was one of the co-authors of the academic paper “Attention Is All You Need” while in Google. This is quite a pathbreaking paper as it introduced the AI model architecture called the Transformer.
Cohere’s roots lay in 2017. The firm’s website informs us that four years back, a team of friends, classmates, and engineers started a distributed research collaboration that wanted to create a medium for early-career AI enthusiasts to engage with experienced researchers. They named it “for.ai.” Two of the co-founding members, Aidan Gomez and Ivan Zhang, went on to build Cohere while others pursued a diverse range of other activities in the same space. It claims that during that time frame, “for.ai.” was one of the first community-driven research groups to support independent researchers globally.
Whopping $170 million funding in a short span
In a very short span of time, Cohere has been able to raise quite a bit of money for its aspirations. In 2021, Cohere announced that the firm had raised USD 40 million in Series A funding led by Index Ventures, Radical Ventures, Section 32. Some of the biggest names in AI- Geoffrey Hinton, Fei-Fei Li, Pieter Abbeel, and Raquel Urtasun also participated.
In February 2022, Cohere raised USD 125M USD in a Series B round led by Tiger Global, Radical Ventures and returning investors Index Ventures and Section 32. Bill MacCartney (ex-director, Proactive Intelligence at Apple) joined as Vice President of Engineering and Machine Learning to lead the newly opened Palo Alto office.
In a recent interview, Gomez mentioned that one of the issues with Transformer models is that they need huge quantities of training data and compute power to run. Hardly anyone gets access to these high-quality NLP models other than those with access to a quarter-billion-dollar supercomputer (essentially the tech mammoths).
“What we want to do is foot the cost of that supercomputer and give access to all these organisations that otherwise couldn’t build products or features on this technology,” added Gomez.
Big and small non-profits in AI and ML have existed for quite some time now
There are many other small and big research firms that are not for profit. Some of the biggest names include The Alan Turing Institute (founded in 2015). Founding institutes include five universities, namely Cambridge, Edinburgh, Oxford, UCL and Warwick, and the UK Engineering and Physical Sciences Research Council.
As a mission, the Institute wants to lead the public conversation through agenda-setting research, public engagement, and expert technical advice and drive new and innovative ideas which have a significant influence on the industry, government, regulation, or societal views. In 2021, it received a funding of £10 million from EPSRC in June 2021. In 2018, the UK Research and Innovation awarded a total of £48 million in funding.
Allen Institute for AI was established by tech pioneer late Paul Allen in 2014. It has undertaken several ambitious projects to drive advances in science, medicine, and conservation through AI. Crunchbase data shows that the Allen Institute for Artificial Intelligence has raised a total of USD 17.9 million across two funds. The latest round of funding was the AI2 Incubator Fund I in 2020, which was of a total valuation of USD 10 million.
Wadhwani Institute for Artificial Intelligence, founded in 2018, is a not-for-profit research centre founded by Romesh and Sunil Wadhwani. The firm, at the moment,3 says it is building AI-based solutions in the agriculture and health domains, like management for cotton farms, maternal, newborn and child health and tuberculosis. It teams up with government bodies and global non-profits to make sure that the innovations its team of renowned researchers are working on reach those who need them the most. In 2019, Wadhwani AI received a USD 2 million grant from Google to create technologies that will help reduce crop losses in cotton farming with the help of integrated pest management.
The BigScience project is an open collaboration boot-strapped by HuggingFace, GENCI and the Institute for Development and Resources in Intensive Scientific Computing (IDRIS). This project wants to change the fact that most of the control of the transformative changes that LLMs can bring is in the hands of tech mammoths and wants to open up more participation in LLM research.
“To me, it’s critical that researchers continue to be able to share their models and datasets publicly so that the entire field can benefit from them. An initiative called BigScience has been launched, bringing together nearly a thousand researchers from around the world,” said Clement Delangue, co-founder of Hugging Face, while interacting with AIM.