What a year it has been! At Silo.AI, we’ve just hit our first full year of operations with 50 AI projects executed in Finland, Sweden, UK, Central Europe, and US. AI is impacting every sector with most demand emerging from finance, industrial and manufacturing clients. We are very proud of our team at all three offices: in Helsinki and Turku, and as the latest addition, in London.
Despite the hype for AI, 2018 was still a year of experimentation with lot of proof-of-concept and pilot projects. However, in the second half of the year, we started to see more AI deployments impacting real-life business operations. We expect this trend to continue, leading to more transformational AI use cases getting implemented in 2019, as AI is moving from “nice to have” to “must-have”.
Our purpose as a company is to build AI for people. In other words, we believe that people and machines collaborating is more powerful than automation alone. We are building human-in-the-loop solutions which learn and improve from human feedback through every transaction. Use of AI will generate more data for the machine to learn faster. It is clear that organisations that leverage their current employees to train AI faster will develop competitive advantage.
2018 has also been the year of growing attention towards AI ethics and privacy. In Europe the General Data Protection Regulation (GDPR) came into force in May. The GDPR has a lot of implications for the design and implementation of AI, which our Data Privacy Expert Erlin Gulbenkoglu has previously outlined. We also made a less technical interview with her on the topic. In the US, the need for a clear direction for ethical use of AI has been seen in the recent appointment of a Chief Ethical and Humane Use officer at Salesforce. Google also published its own ethical principles in June.
AI interpretability is essential for making ethical AI that takes privacy into account. Interpretability means that the results the AI gives can be understood with a variety of techniques that help pinpoint what data was used to come up with the result and how. At Silo.AI we got a deep dive into AI interpretability in our monthly Silo.AI Academy webinar.
We’ve been constantly evaluating and learning from our projects and clients, especially their data, technology and people. One of our key learnings this year has been to improve our offering in a way that it supports planning the final deployments early enough. We noticed that in many client projects the focus on the deployments needs to start already at a proof-of-concept phase. Design sprint, part of our proof-of-concept projects, helps tackle the issue and establish a clearer path towards a transformational AI solution.
At Silo.AI we want to support cutting edge AI research and share some of it with the rest of our research community. Silo.AI Academy webinars are technical webinars intended for researchers and other machine learning experts, that have included topics like State Space Gaussian Processes with Non-Gaussian Likelihood, Bayesian Networks and Non-Stationary Spectral Kernels. In addition, within our own group of 20 PhDs we have a weekly research club, where we look into the most recent techniques and methods of AI, such as applying Multi-agent Reinforcement Learning.
Over the year, the Silo.AI team had the pleasure to speak or present at nearly 100 different events and participate in at least another 100 more, ranging from top research conferences, such as NIPS and ICML, to business events, such as Hanaholmen and Slush, where we hosted an AI hour with PwC and Peltarion. Towards the end of the year we were listed one of the top AI companies in Finland. Next year we will continue to expand internationally – it will be great to see where it takes us.
On behalf of the entire Silo.AI team, we would like to thank our clients and wish everyone Happy Holidays and an excellent year 2019!
CEO & Co-Founder
P.S. Join our team next year? We’re hiring!