Keynote Panel: AI in Finance
summary
Panel reports concrete wins and limits: NLP models driven by recent deep learning advances process 5 million documents a day for central-bank and earnings analysis, earnings surprises still drive measurable alpha over weeks, and a de-identified consortium graph of up to petabytes reveals fraud islands that cut losses in virtual‑currency flows. Practical testing matters — vendors must label millions of synthetic records in a 24-hour vendor test or fail. Graphs plus embeddings enable privacy‑preserving link analysis for AML, supply‑chain what‑ifs, and new KPIs that outperform simple averages.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 1:43Jane Andry
- 2:39Steve Gustafson
- 3:32Michael Evans
- 4:40Evan Schnittmann
- 7:35Asking the right questions
- 9:11Using alternative data
- 9:55Optimization doesnt matter
- 11:49When will alpha decay
- 14:27Price inflection points
- 18:02Managing the signal
- 19:12Natural language processing
- 21:36Embedded vectors
- 23:21Natural image processing
- 25:39Deep learning in NLP
- 27:49Why is there so much money spent on advertising
- 29:46The dam is ready to break
- 31:09How is the graph to model knowledge
- 33:23Deidentified consortium graph
- 40:21Same sourcing
- 46:31Knowledge graphs and AI
related talks
18mKeynote: AI for Financial Services
26mKeynote: AI in the Real World: Today and Tomorrow
37mPanel: Data Analytics In Investment Banks
39mKeynote 3: How did you know? Explaining Black Box Model Predictions in Python
59mKeynote Panel AI in Healthcare
16mOperationalizing AI in Finance: Real-Time Analytics
59mAI in Finance: Scope and Examples
38mAI in action: How to use practically in iGaming beyond the hype?