mathstodon.xyz is one of the many independent Mastodon servers you can use to participate in the fediverse.
A Mastodon instance for maths people. We have LaTeX rendering in the web interface!

Server stats:

2.8K
active users

#naturallanguage

0 posts0 participants0 posts today
Replied in thread

Is there a cognate lookup that doesn’t rely on a base language? There are plenty of multilingual options that force you to select an input or output option. Are there any that search close-enough words (whether cognates or simply spelled similarly)?

@mattia I want to know if these exist too, this is a really, really good idea, to have one that combines all languages.

LLMs are the closest thing we have so far but the accuracy is very low. And really, a tool such as this should be no more complex than a spell checker, and so a lookup should only take up a tiny fraction of the amount of memory and compute power an LLM requires for local inference.

Google's finally catching up to Perplexity, ChatGPT and Claude.

#AI #GoogleGemini #Google #Multimodal #NaturalLanguage #NLP #Search

======

Google's AI Search Evolution: Reimagining How We Find Information
androidfaithful.com/googles-ai

---
Google's Gemini 2.0 uses advanced reasoning capabilities to understand context and provide more meaningful responses, with key improvements including enhanced performance, multimodal queries, faster response generation, and nuanced understanding of user intent.

Firing up #Inform7 on my #Mac to write another #interactiveFiction #game. Getting stuck into the prologue. This will be a much bigger game than I've written before, including 7 main sections in the #code. It is going to take some time. So best get on with it sooner rather than later! Excited to be back #coding. There's something magical about using this #declarative / #naturalLanguage / #objectOriented #programming language and IDE. #GameDev #IndieGame #TextGames #Parser #TextAdventure #Inform

As one of those that is quite likely to use the word delve in both the original archaic sense of tilling the soil and the more current general sense of doing work and research in some detail, I'm more than a little indignant that its overuse by Openai generators is likely to cast doubt upon my own legitimate vocabulary.

And yes spelt is the past tense of the verb to spell. And yes this is serious.

Today on hashtag#MapScaping: our very own Jason Gilman dives into the world of #NaturalLanguage Geocoding and discusses the future of ML and AI in geospatial.

We're a little biased, but as far as we're concerned this is a must listen! Check it out, and get in touch with any questions you have for Jason and our team.

Full podcast: mapscaping.com/podcast/natural

You can read more about our work on natural language geocoding (and beyond) anytime on our blog: lnkd.in/drsyDSGV

mapscaping.comNatural Language Geocoding - August 9, 2024In this episode, I welcome Jason Gilman, a Principal Software Engineer at Element 84, to explore the exciting world of natural language geocoding.
Replied in thread

@adredish

#LLMs are product of having available online huge text corpus, with computational big data processing tools, and non-supercomputers with computing capabilities for looping few maths algos over & over; yet still wondering around the issues described as the
„natural language fallacy / unnatural language problem“ in 1976 by Drew McDermott:

» In this section I have been harsh toward #AI 's tendency to oversimplify or overglorify #naturalLanguage ... «

dl.acm.org/doi/10.1145/1045339

ACM SIGART BulletinArtificial intelligence meets natural stupidity | ACM SIGART Bulletin As a field, artificial intelligence has always been on the border of respectability, and therefore on the border of crackpottery. Many critics <Dreyfus, 1972>, <Lighthill, 1973> have urged that we are over the border. We have been very defensive toward ...