Microsoft Eases Integration with Semantic Kernel AI SDK by David Ramel #Azurehttps://visualstudiomagazine.com/Articles/2024/10/09/Microsoft-Eases-Integration-with-Semantic-Kernel-AI-SDK.aspx?utm_source=bluesky&utm_medium=social&utm_campaign=azureweekly
The basic idea is to provide unified API abstractions, especially for idiomatic C# code, to help platform developers and others work with any provider with standard implementations for caching, telemetry,...
FAKE NEWS DEBUNKER Extension for information verification: - video frames/thumbnails extraction - webpage/local file AI analysis - metadata extraction - Fact Check semantic search and more. chromewebstore.google.com/detail/fake-...twitter.com/veraai_eutwitter.com/WeVerify
I’d say that LLMs can perform semantic tasks, but only within the restrictive (but unreasonably effective!) bounds of Firthian-style distributional semantics. Of course, actual thought, knowledge or cognition is as far beyond them as the moon is to a goldfish, and that’s where people get fooled.
Early pipeline entity extraction + LLM allows me to do semantic searches, in ways no other tech currently does. Even with a high error rate, it's useful. For example, documents where a speaker is introduced and then later is quoted saying something using a pronoun, like: 'He complained that ... "
Agreed on the second assertion, but to me that calls into question the first assertion, especially since any LLM’s ability to perform any “semantic” anything is an illusion.
I do think that his use case — basically semantic grep and summarization — makes sense and is probably quite helpful, but it’s also true (going back to your point) that it doesn’t provide broad efficiency gains, especially because you have to look for what was missed, not just verifying what was /
tech people!) can’t tell the difference, or have incentives to obscure the difference, between bullshit generation versus semantic search/transduction, either because the interaction mode by necessity blurs the two, or because you can’t sustain a trillion-dollar valuation on the latter alone.
You can publish lots of papers but nothing is quite as cool as seeing your name in physical print 😍
I’m going to be a bit wishy-washy here and say that LLMs are both high-powered bullshit-generating industrial turbines /and/ unreasonably effective semantic Swiss Army knives, depending on whether or not you’re operating on text in its context window. The problem is that normal people (and many /
Quite. Speaking as an employer, I am also a working person. This semantic distinction is irritating. (and yes, it does mean I have to put prices up/cut staff hours back. Things. Are. Tight)