For AI to become a reality, it needs data. Lots of data, ideally well structured, from many different data sources. Without it, the models on which AI solutions are based will not work. The biggest problem in developing AI solutions is the lack of access to data. This is common knowledge for Fredrik Schlyter, co-founder of AI agency Insigo. In his view, open data is an important part of the solution to the problem. Access to data in customer relationships is crucial.
Terms like AI and machine learning have become natural parts of the Swedish language. ChatGPT is used for everything from writing essays in primary school to writing programme code in large companies. It is taken for granted that AI will soon, very soon, be able to solve a wide range of problems, creating both anticipation and fear. For the pioneers building AI solutions, concrete, practical problems occupy much of their time:
– It is often difficult to know in advance what results can be achieved with AI. Data is therefore needed to test the waters. A big problem is that data is often not available. Increased publication of open data would help,” says Fredrik Schlyter, co-founder and developer at AI agency Insigo.
Greenhouse – an example of the use of AI
Fredrik Schlyter and his colleagues at Insigo have so far taken on a wide range of tasks: natural language processing (NLP), computer vision, document anonymisation (annotation), automatic generation of newsletters and websites, and optimisation of greenhouses are just a few examples.
– For greenhouses, it’s about figuring out how water use, temperatures, humidity and a host of other factors affect the business, explains Schlyter.
Optimising greenhouses provides a good summary of the benefits of AI. When the number of factors and the relationships between them become too many for a human to monitor, an AI model can find patterns that can form the basis for improvements.
Want to learn more about open data? Read about how open data is used within journalism.
Among other examples of interesting AI applications, Fredrik Schlyter mentions the analysis of invoices to extract different types of data for further processing and analysis. This is an application of particular interest to municipalities, regions and public authorities, which have to comply with strict legal requirements when it comes to spending money. One of the most popular data standards and open data in Sweden is accounts payable, which reports incoming invoices from suppliers, where it is possible to identify fraudulent invoices with the right data.
Insigo has so far mainly been involved in consultancy work, but there are ambitions to offer finished products and services. Examples include automated solutions for newsletter and website generation. Insigo’s developers have therefore started to develop such solutions for their own business, with the ambition to eventually offer them to their customers.
Measuring open data
An interesting project that Insigo is involved in is an “educational” visualisation of how far Swedish municipalities have come with the use of open data, which can be seen here (source code here). It is being done for the National Data Workshop that the Västra Götaland region is hosting. In connection with this and other assignments, Insigo has worked with MetaSolutions, the main developer of the EntryScape platform for managing and publishing open and shared data.
– It is very rewarding to work with MetaSolutions, which plays an important role in increasing the accessibility of data, says Fredrik Schlyter.
Learn more about open data here.
He points out that the different people working together have different skills, which creates “an interesting chemistry”.
– Mixing different skills in collaborations allows you to come up with new solutions that you would not come up with on your own.
This is an important aspect of data use in general, and open data in particular. People with different roles and skills can come together around data to get a shared view of the problems that need to be solved, leading to new insights and better collaboration.