The current issue of The Economist features several articles on AI. These mention the „Common Crawl“ as a data source, rather than „mostly Wikipedia“. (I remember the Common Crawl from the LLaMA context and others, with diagrams available detailing the composition of the dataset besides the C.C.). Read more... 23 Apr 2023 - less than 1 minute read
Wonderful Research paper on “Mass Editing Memory in a Transformer”. Updating facts is one thing, but there is also the “right to be forgotten” which needs to be addressed by an actually lawful AI. This is based on GPT-2, and OpenAI say they’re not training GPT-5. One may wonder what the brainiacs are doing instead, and THIS sounds like something... Read more... 22 Apr 2023 - less than 1 minute read
Google Bard now supports coding as an explicit use-case. It improves on Github Copilot and also GPT-4/ChatGPT in terms of rights awareness: If Bard quotes at length from an existing open source project, it will cite the source. Read more... 22 Apr 2023 - 1 minute read
Following up on the publication of “AI security concerns in a nutshell - Practical AI-Security guide” by the German BSI (Federal Office of Information Security), I highlighted that another, more specific security concern to generative AI is prompt injection, a throwback to the SQL injection attacks of the… 2000s? 2010s? Anyway, OpenAI’s proposed... Read more... 21 Apr 2023 - less than 1 minute read
Use Cases for AI presented by Matthias Lichtenthaler of the Federal Computing Center 🇦🇹 (“Bundesrechenzentrum”), at Future-Law “Digitaler E-Signatur Tag 2023”: Virtual Assistant: Case Categorization(?) Consultation/Assessment Semi-automated intelligent processing/answering Read more... 20 Apr 2023 - less than 1 minute read