My new podacstepisode has just been released!
[Warning]: It is in german!
Here are the linkz:
Original:
Spotify:
Apple podcast:
What's it about?
I'm talking about the EU's attempt to deal with the AI mess, which - and now read carefully ;) - was not created by the AIs, but by the humans who abused those AIs.
See also my podcast episode: https://www.wackyworld-podcast.de/e/diskriminierung-durch-falsche-ki-benutzung/.
We are talking about the EU AI Act, more precisely, the proposal for a "Regulation of the European Parliament and of the Council laying down harmonized rules for Artificial Intelligence".
At Amazon KDP, AI-generated trash texts are submitted en masse as guides, the image generators create images that are hardly distinguishable from real photos (deepfake alert) and not a day goes by without sensational reports about the next hurdle cracked by AI combined with apocalyptic warnings about the imminent mass unemployment that will especially affect mathematicians, computer scientists, journalists and quite a few creatives (my application as a cocktail mixer for a bar in Hawaii went out yesterday).
While one faction is raising its hands in jubilation, the other is arming itself. With legal ammunition. Italy is the first EU country to ban our text-generating wunderkind. Who would have thought that? Certainly not the Godfather.
Has he risen from the sea, the singularity-Godzilla? Is a world-spanning mastermind forming from the neural networks? Does Schätzing have to rewrite his swarm? SwarmGPT?
Or should we not roll any legal menhirs in the path of technology or the believers in technology on their way to digital Elysium?
My suggestively worded question reflects the discussion. Either / Or. Black or white. 0 or 1. Computer science professors who have never set foot in a law lecture hall are suddenly able to fully evaluate legal acts of the EU. Verdict: stifling innovation, preventing competition. Female law professors, who are happy when their research assistants turn on their computers, are becoming experts in the field of artificial intelligence. Verdict: We have just prevented the matrix architect.
It's about time someone looked at the matter as it is, damn complex and neither black nor white, but deep gray with 256 color gradations.
Someone who, unlike many others participating in the disscussion, has read the whole 119 pages of the primary source (below is the link)
Questions that are answered
- Introduction: why should we be addressing this issue right now?
- What is the EU AI ACT (hereafter EAA) from a jurisprudential perspective?
- Is it even legally enforceable yet? (No, it is not, it is only a proposal so far (as of 2023/04/09 - 00:38), will even be renegotiated again in April 2023).
- How much room for maneuver is left to the member states?
- What happens if the EAA is violated? (Spoiler: it gets damn expensive).
- What is the EAA in broad terms?
- Objectives.
- How will AI systems be categorized?
- Focus on high-risk applications:
- Which AI systems fall under this?
- What regulations need to be considered here? In other words: What requirements must be observed here?
- Who all is required to comply with the regulations? (Spoiler: also the users, not only the operators...).
- Which subtleties have to be considered / discussed at the EAA / can surprise you quite a bit?
- Is it really true that ChatGPT and Midjourney (and similar systems) were / are classified as high risk applications? (Spoiler: There is a lot of nonsense being spread).
- EVALUATION: How should the whole project be evaluated? What advantages does it bring? What price do we have to pay for it?#
- How can Klosterfrau Melissengeist help us all back to a "clear" mind?
Sources / References
The primary source
https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52021PC0206 (last visit: 2023/04/07 - 20:34 -- don't wonder about the flag icon confusion, here you can choose in which language, you want to have the masterpiece)
Sources with reference to European law (from the circle of those involved)
https://futurium.ec.europa.eu/en/european-ai-alliance/blog/chatgpts-rapid-entry-urgent-government-appoint-authority-responsible-ai (last visit: 2023/04/08 - 09:33)
(last visit: 2023/04/08 - 10:3
https://www.politico.eu/article/eu-plan-regulate-chatgpt-openai-artificial-intelligence-act/ (last visit: 2023/04/08 - 10:39 -- IMPORTANT: The only SERIOUS(!) source I found about ChatGPT and Co. regarding their classification as high risk applications. It seems that many publishers have forgotten what makes good journalism, namely own research and not stupid copying from the colleagues, of which sometime someone claimed that ChatGPT is already "in the law". 1. there is no law yet, i.e. no regulation at all, there is only a proposal 2. there was also no extension of this proposal, but only 2 of the main responsible persons for the creation of this proposal have expressed themselves in the direction of "ChatGPT must go in". But what the heck. Citing sources seems to be out, after all. *stamping foot and sulking away*)
More web sources (not generated by ChatGPT)
https://www.stern.de/digital/technik/chatgpt--erstes-eu-land-verbietet-ki-mit-sofortiger-wirkung-33339022.html (last visit: 2023/04/07 - 12:43)
At first I was pleased (as a free spirit) that the state is not granted such far-reaching possibilities of AI-assisted investigation, until I heard in this excellent episode of the Heise show what these search possibilities are used for. As a family man, it changed my mind: https://podcasts.apple.com/be/podcast/heiseshow-bei-der-secit-gespräch-mit-dem-leiter-der/id1082309594?i=1000604496295 (last visit: 2023/04/07 - 13:01)
Legal sources
Ebers, StichwortKommentar Legal Tech, Regulation (EU), KI-Verordnung Rn. 1-17, beck-online (My primary "summary" source).
RA Dr. David Bomhard / RAin Marieke Merkle, "The Current Commission Draft and Practical Implications," RDi 2021, 276.