Paulo Carvalho has been working in the field of futures and foresight for more than 25 years. On one hand, he is a professor in foresight, strategy and innovation at the Faculty of Economics and Management at the University of Lisbon. On the other hand, he founded a foresight company five years ago, IF Insight Foresight, focussing on consulting, horizon scanning and strategic intelligence, as well as other strategy and innovation projects. He talked to Futures4Europe about Insight Foresight’s recently developed tool ORION and how it could revolutionise foresight practices.
Tell us about ORION.
Orion is a horizon scanning and intelligent system that incorporates different kinds of databases in powerful ways to capture, identify, categorise and make sense of different driving forces, such as megatrends, specific trends, wild cards and weak signals. Scanning stems from the principle of a holistic and extensive approach to explore the future. Our tool uses databases to enable individuals and organisations to extract, identify and make sense of the complex, holistic and systemic world in which we live. For about six years, we have used what we call a scanning and intelligence database, which through ORION can be organised both for consulting and also capacity building and training provisions.
Can you explain how the tool captures, categorises, and leverages drivers of change, insights and signals?
We synthesise data from both curated and non curated databases and integrate them with the knowledge of ORION AI, which is the combination of these databases with the power of generative AI. The database is organised around 20 megatrends that encompass around 1,400 different trends and signals. Each one of these elements and its metadata was curated by our team, or a third party. Lastly, the second layer of the database is designed for the identification and classification of non-curated information (such as articles) from various sources. These include more than 20,000 driving forces that keep growing and expanding. We can organise and extract specific trends, clusters, connections patterns from each layer or the entire scanning database using specific algorithms.
How does it help you find what you’re looking for?
When you ask a question to ORION AI, the system will go through a three-step process. Firstly, it will consult the first curated layer with megatrends and trends. Then, it explores the second layer, which consists of more fluid, non-curated articles. Based on this, it will go through the knowledge of the large language model, in this case GPT 4, to obtain the most relevant answers on the foresight horizon scanning query. In this sense, it is a co-pilot, an interactive conversation. Each time you interact and ask the system questions, you improve the way that you are talking and interacting with the system.
Who will be the primary users of your tool?
You do not have to be from a specific background or industry to explore and extract value and knowledge relevance from the database, hence why the many global forces and megatrends included. It's a global platform with global content in which you can make deep dives. Part of this will be available for public use, so you can explore multiple megatrends and the detailed description of each. After a certain level of depth, a subscription will be required to have access to the entire set of megatrends and trends and make use of the generative AI assistant. For example, you can look for trends related to the future of biotechnology, but if you want to go deeper and know more specific applications of this, you can make a deep dive and explore the topic more narrowly.
How does this tool address potential biases and ensure the ethical use of AI-generated insights?
In the construction of ORION, we still employ a human approach, in other words, the curation of the database and its sources. Of course, through keyword filtering and text, we cannot exclude our own biases. When we combine these with generative AI, we generate clusters or global forces not just from the top down, but also enabling the database to create its own clusters through a algorithms. What emerges from the algorithmic approach is different from our approach to organising our database, which is how we can identify blind spots and biases by comparing what our team has done and what the language model has done. When you use generative AI and large language models, even those trained by our own database, it is inevitable that biases translate to the generated answers. They will always be there, so, never take for granted the results obtained from the system. However, I believe that what you can get is so powerful and it can be so insightful.
What advice would you give to those who would like to use AI for strategic foresight?
My first advice is don't be afraid of exploring the new world of generative AI that is already disrupting strategic foresight, futures horizon scanning and strategic intelligence. We are using powerful algorithms to identify patterns and things that two or three years ago was completely impossible for many people who work in this field. However, you need to be aware of the biases that can ensue and iterate these with your team, workshops, participatory methods, hence why you should see this tool as a co-pilot. Use it with caution and be aware that we are just at the beginning of this journey in which generative AI has an impact in our personal and professional lives. Foresight, combined with innovation and strategy, will only continue to expand in the world of AI and generative AI.