🔴 New Paper — Silent Mandates: The Rise of Implicit Directives in AI-Generated Bureaucratic Language
How AI drafting creates implicit mandates without explicit agents, reshaping law, healthcare, and finance.
🔗 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5398989
#LLM #MedicalNLP #LegalTech #MedTech #AIethics #AIgovernance #cryptoreg #healthcare #ArtificialIntelligence #NLP #aifutures #LawFedi #lawstodon #tech #finance #business #agustinvstartari #medical #linguistics #ai #LRM #ClinicalAAI #SyntacticSovereignty
We're doing some natural language processing at Safetica, so we had to integrate a tokenizer.
First, we started with HF tokenizers that use Rust as the underlying implementation. Since we consume it as a C# package and native runtimes are not shipped for Mac (thanks Apple for preventing organizations from virtualizing Macs), I made my own C++ implementation.
Naive implementation in 60 lines of code, using std::map. Was slower than Rust's impl. By about 25%.
Swapped std::map for std::flat_map. It instantly outperformed the Rust implementation. By about 20%. Sure, my impl doesn't have the full feature, but we don't need that.
Flat map is frigging cool. It is a C++23 feature, so it is not available in all compilers, but you can find open-source implementations on GitHub.
Athens. Summer school dedicated for natural language processing.
Organizers are really great.