๐——๐—ฟ๐—ฒ๐˜€๐—ฑ๐—ฒ๐—ป, ๐˜„๐—ต๐—ฎ๐˜ ๐—ฎ ๐—ฏ๐—ฒ๐—ฎ๐˜‚๐˜๐—ถ๐—ณ๐˜‚๐—น ๐—ฐ๐—ถ๐˜๐˜†. ๐Ÿ‡ฉ๐Ÿ‡ช I spent two days there last week for the 3rd Conference on AI in Finance at #TUDresden. I think it might be one of the most beautiful cities in Europe. I really enjoyed the first day with the PhD researchers, particularly hearing Tony Klein and Vincenzo Capizziโ€™s perspectives on publishing and where academia is heading in the age of #AI. It was interesting to hear the editorsโ€™ side of that conversation. One thing that struck me: more papers and submissions donโ€™t necessarily mean better research. The editors described clusters of similar submissions, concerns over AI-generated tables and results, and even papers ignoring the journalโ€™s format. All of this takes more time to check. Being able to pay conference or journal fees is no substitute for good research. The second day had a good mix of research, startups and practical work in finance. Alejandroโ€™s keynote and the talks by Eghbal, Kolja, Toghrul and Yifei were among those that particularly stayed with me. Our discussion around Alejandroโ€™s talk left me wondering: if AI automates more of active trading and prices absorb information faster, will there be less alpha left to find? We also talked about time series foundation models, the move towards multimodal forecasting, and AI tools and robo-advisors that could make financial advice more accessible to retail investors. Another finding discussed was more investor disagreement during AI outages. It made me wonder how much these tools push people towards similar views, and what that crowding could mean for markets. We also kept coming back to regulation and shared standards: how do we automate more without compromising the integrity or efficiency of the financial system? I also shared our work at #ZanistaAI on financial news, embeddings a