Thanks to foundation models, i.e. basic models that are pretrained using large datasets, it is increasingly possible to customize financial proposals according to the characteristics of individual investors. But that's not enough. To be credible, proposals must be explainable. And therefore transparent

To work, AI uses humangenerated data (this is why it is recycled), managing to learn from them and create new (augmented) knowledge as per Network Science, thus supporting decisionmaking processes. But to arrive at human intelligence it is probably necessary to arrive at the synthesis of the two

Bots and generative AI can undermine democracies by manipulating voters into believing an idea is popular and shared. Just like in product marketing campaigns

The future lies in prompt engineering, which is the ability to know how to ask the right requests to an AI algorithm whose task is to write code in collaboration with a more or less expert human agent

More and more companies, especially on the web, are adopting algorithms to fix the pricing of their products. Some empirical cases have shown that this benefits retailers (+38% for German petrol stations) and this has opened a legal and political debate on the collusive potential of this approach but also on the possible positive impact. However the real issue remains the data and their access to correctly train the algorithms themselves

Using algorithms to find patterns in historical data on startups and predict the likelihood of success for a new startup could hold entrepreneurs back, by encouraging the repetition of similarly innovative ventures at the expense of breakthrough innovation