The premiere of the Cologne Economic Talk was broadcast live from the KOMED hall in Mediapark Cologne (Fotos by Peter Johann Kierzkowski / CC BY). On the podium were:
- Maik Außendorf MdB, spokesperson for digital policy of the parliamentary group Alliance 90/The Greens,
- Dr. Robert Zores, Chief Digital Innovation Officer (CDIO) and Managing Director REWE digital
- Thomas Bartz-Beielstein, Managing Director of the IDE+A Institute for Data Science, Engineering and Analytics at TH Cologne and
- Lukas Stratmann, Chief Evangelist of the STARTPLATZ Al Hub.
The moderator was Antonia Koop.
My Top 3 Takeaways, if you cannot watch the whole talk:
- Bulletproof AI made in Germany/EU: Among other things, by establishing AI error reporting systems (similar to the US Aviation Safety Reporting System or the reporting system for safety-relevant incidents of the Federal Supervisory Authority for Air Traffic Control)
- LLM hallucinations are most likely features of current language models and not bugs that can be fixed with more computing power and data.
- AI regulation must be implemented efficiently so that it does not prevent innovation. This is not free of charge.
Recommended videos:
- Why AI is incredibly smart and shockingly stupid (Yejin Choi | TED2023) https://go.ted.com/6Wxz
- How to govern AI — even if it’s hard to predict (Helen Toner | TED2024) https://go.ted.com/rcQfm
Further links:
- NASA Aviation Safety Reporting System: https://asrs.arc.nasa.gov/index.html
- CIRSmedical, the reporting and learning system of the German medical profession for critical events in medicine: https://www.cirsmedical.de
- Error reporting and learning system for doctors: https://jeder-fehler-zaehlt.de
- A page that indicates whether your own data has been used by an AI: https://haveibeentrained.com/
- Gary Marcus: https://garymarcus.substack.com
- Melanie Mitchel: “AI now beats humans at basic tasks”: Really?https://open.substack.com/pub/aiguide/p/ai-now-beats-humans-at-basic-tasks?r=3u0kmu&utm_campaign=post&utm_medium=web