“Think of it like this: France launches the Concorde; that same day, literally anyone – pilot or not – can get one for themselves and fly it. That’s what’s happening right now with machine learning on GitHub.”
Though machine learning remains an emerging technology, the highly-technical field is in a period of commodification. Without any need to understand the technology itself, anyone has the ability to go online, download a machine learning algorithm or toolkit and apply it as they wish. While there are many benefits to this, there’s also a downside.
“What we’re seeing is a lot of people saying they can do things with AI, but actually can’t,” says Michael Summerville, a data scientist with Bloomberg’s Global Data team in London, and the source of the Concorde analogy. “A recent survey conducted by London venture capital firm MMC found that 40 percent of European startups classified as ‘AI companies’ don’t even use any AI in their products. The space is reaching peak hype – there’s a lot of money, more players, but those players aren’t necessarily from the highest-level background. There are diamonds in the rough, but there’s a lot of rough.”
If expertise isn’t necessarily a requirement in the machine learning space, it automatically becomes harder to find. For Summerville and his team, which uses natural language processing (NLP) to analyse data about European companies, this issue can present a real roadblock to understanding the true challenges faced by others in the wider NLP space, and how they’re being overcome.
Enter Data Pitch. This is an EU-funded open innovation programme and virtual accelerator delivered by the Open Data Institute (ODI) and the University of Southampton. It pairs corporate and public sector organisations that have data, with startups and SMEs that specialise in working with data. Data Pitch asks those larger organisations to define a challenge that the startups and SMEs will respond to with a proposal for creating a high-impact, innovative product or service.
The challenge set by Bloomberg was to develop NLP technologies that can be trained to autonomously answer subjective questions that an analyst or investment manager may have about a company’s Environmental, Social and Governance (ESG) practices. These ESG measures could include anything from a company’s impact on climate change to the diversity and inclusiveness of its workforce and leadership team. However, ‘good’ ESG practice is an often subjective notion – the definition is relatively flexible.
So, at a time when organisations are facing greater socioeconomic pressures than ever before to act in a more sustainable and ethical way, subjectively determining the performance of their ESG practices is of huge importance to socially conscientious investors – both individuals and organisations who wish to see their money invested in companies who behave in a way they approve of. In fact, the number of Bloomberg customers making investment decisions using ESG data has more than tripled in the past seven years. It’s a huge growth area for the finance industry, so anything that can speed up the process of evaluating ESG practices has value.
For its startup partner in this challenge, Bloomberg selected SummarizeBot, a Latvian data science and AI-driven company that specialises in information extraction, structuring and analysis in order to automate tasks that previously required human-level intelligence. SummarizeBot’s team officially started work on the challenge on April 1, 2019, and will receive financial and advisory support to develop its idea over the next six months.
For Bloomberg, participating in Data Pitch isn’t about finding a solution that couldn’t be developed in-house. In addition to collaborating with and helping promising startups, it was about identifying those diamonds in the rough.
“To us, Data Pitch is a science project,” explains Summerville. “It gives us an opportunity to partner with smart people and have interesting conversations about the problems they encounter in the challenge, not to mention hearing from them what’s going on across the AI and NLP sectors.”
Later this year, we’ll report on the results of the Data Pitch collaboration between SummarizeBot and Bloomberg.