
Beware the AI indoctrination engines
Chatbots have the clear potential to influence people towards dangerous beliefs and behaviors. How can we mitigate the risk to ensure they deliver benefits to society, not harm? ...
by Susanne Stein Published 10 March 2023 in Technology • 5 min read
One of the advantages of being a family business is that you can take a five-year horizon on decisions and not worry about quarterly reporting – but it is also necessary to be aware of the tensions created by the decisions you make. Within a family business, you always end up asking yourself, “What would my grandfather do?” On the one hand, you need to hold on to some of the tradition and some of the history, but at the same time, every generation must find its own path.
A case in point was the decision to sell our printing company in 2001. As the fifth-generation owner of a publishing house, it was hard for us to stop printing books. But it was clear that what printing was in the 20th century, artificial intelligence and software engineering are going to be in the 21st century.
MANZ’s digitalization journey began with my father in the 1970s, after he had travelled to the US and saw the first computer-assisted legal research databases from Westlaw, the company that is now owned by Thomson Reuters. He came back to Europe and founded the RDB Rechtsdatenbank which has evolved to become the leading database of relevant legal journals, collections of decisions, commentaries and handbooks for law in Austria.
The Ministry of Justice has set the pace for technological adoption in legal publishing by making much of its data public and, more to the point, free. While it’s true that the biggest
competitor you can have as a publisher is free data, it does drive competition and improvements within the sector. It also pushed us to move quickly and to evolve to offer our customers extra features and more information.
The challenges we now face are centered on artificial intelligence and products like ChatGPT, the artificial intelligence chatbot developed by OpenAI which launched in November last year. Responding to these challenges is not something for the future. We have already brought examples to market. Last November, we launched a product as part of our RDB Rechtsdatenbank called RDB Genjus based on this type of generative AI, but we trained it ourselves. Instead of searching through all the data available online, it is more precise, which is important for our customers.
The technology is not wholly there just yet, as we offer the AI search function only within documents that exist in our database. This reduces the chance of mistakes. So, a user cannot, for example, upload a document and then search for a similar document, but as soon as they have found a document within our database that is relevant, we then do offer the AI search. This reduces mistakes to less than 1%.Â
For the past 25 years, searches have been based on machine learning – broadly defined as the capability of a machine to imitate intelligent human behavior by using vast amounts of data – and this certainly remains the direction of travel, though it is unlikely that we will continue to use machine learning in the way that we do today.
Our search capabilities have always been good, but the next step is to take a more linguistic approach to search. In every article we publish, we mark specific words as important, so users had to know broadly for which legal point they were searching. The use of natural language, as well as the ability of a computer program to understand human language as it is spoken and written, will add considerable value. It will help provide more nuance, more differentiation, and deeper content. In short, it will help users search accurately in a world where there is even more information. The ongoing challenge is that of natural legal language. It requires computers to understand language at a high level.
We always compare ourselves to what Google is doing, and for legal language, it is not yet especially good – even more so for German legal language. What research there is, is going into English.
Our content is our data treasure, and we want to get as much as possible out of it for our customers. It has been a long journey from print to digital, but by harnessing the power of AI we hope to make our family business fit for another five generations.
Â
Managing Partner of MANZ Publishing
Susanne Stein became Managing Partner of MANZ Publishing in 2006. Since then, she has streamlined the business, merged the business titles, and brought the publisher back under full family ownership. She completed IMD’s Transformational Leader program in 2021.
21 hours ago • by Michael D. Watkins in Technology
Chatbots have the clear potential to influence people towards dangerous beliefs and behaviors. How can we mitigate the risk to ensure they deliver benefits to society, not harm? ...
21 hours ago • by Mohan Subramaniam in Technology
Traditional manufacturing and service companies need to leverage the new and explosive power of data and digital ecosystems to be competitive in the future, says IMD Professor of Strategy and Digital Transformation...
15 March 2023 • by Massimo Marcolivio, Michael R. Wade in Technology
Michael Wade and Massimo Marcolivio propose four categories of KPIs to measure and track digital transformation....
13 March 2023 • by Dave Chatterjee in Technology
As large-scale IT projects notoriously overrun and exceed budgets, Dave Chatterjee explores the factors that made migrating the American Cancer Society’s data to the cloud a rare example of a smooth digital...
You have 4 of 5 articles left to read.