AnyIdea AI and AI Prio Score

AnyIdea's AI layer has both a synthesis AI and a generative AI. Further details can be found under FAQs - Which AI does AnyIdea use?

The Foresight module is primarily driven by Synthesis AI, which was developed in a 4-year research project with two academic institutions and FFG Austria. Based on an enormous amount of data (input), it provides important insights and a basis for decision-making (output).

AI scouts and data

The artificial intelligence calculations are based on so-called AI scouts, which can be created and managed independently in AnyIdea. AI scouts give the artificial intelligence a specific orientation and view of the context in which data is analyzed, linked, weighted and evaluated. Based on this, an AI prio score is calculated and a 100% data-driven assessment and prioritization of trends and signals is obtained.

Which data is taken into account by the AI, further details on this, categories and their properties can be found under Data categories and properties.

AI Prio Score

The AI Prio Score gives you a well-founded assessment of artificial intelligence with regard to the relevance of trends and individual signals. Both for each individual AI scout created and for an entire company when the sum and results of all AI scouts are combined and viewed together.

The development of the individual signals, the development of the sum of relevant signals, their distribution and occurrence in the respective data categories and their occurrence in relation to a time horizon of e.g. x months, provide information regarding the relevance, probability of occurrence and time to impact of technologies and topics.

Bottom up calculation

The AI prio scores are calculated bottom-up. This means that the totality of all signals is always considered and included in the calculations. The respective results are extrapolated and further aggregated "upwards" to macro trends. This results not only in AI prio scores for macro trends, but also for individual signals, signal groups or superordinate topic clusters.

The language models used by AnyIdea take into account not only synonyms but also semantic similarities. The models are continuously updated and brought up to date. In addition to these models, specially developed algorithms are included that can weight and sort the data, data categories and search results accordingly. 

The AI processes documents, articles and signals and extracts the most important information. This information is transferred to a multidimensional vector space using a deep learning language model and then transferred to a vector space of a lower dimension, clustered and calculated.

A thematic weighting is calculated on the basis of cosine similarity, the topicality and number of signals and other textual similarities.

A time weighting for signals at the lowest level is carried out based on their data categories, for example, scientific publications are classified as more theoretical and future-oriented than start-ups and, on the other hand, news tends to be weighted less depending on their publication date and frequency due to the increased number and volatility. In order to make a statement about the time behavior of signal clusters and topics, the number of all associated signals per topic is considered over time and a prediction (time series analysis) of the development over the next six months is calculated.