>>16565482
It's hard to explain. I've been working in NLP tech startups for 10 years now though, so AMA. I just ask you be more specific.

>>16565482
The limitation of AI is hardware, not software.
Without Quantum computing with quantum locked distributed nodes you will always have nothing more than a glorified data structure folding Pseudo PdP math origami.

>>16566022
It's not hard to explain at all.
90% snake oil sales
2% Call center optimization
2% Robotics
1% research
1% DoD/NSA data snooping/Game theory (illegal but they don't care)
1% Corporate data snooping/marketing optimization (illegal but they don't care)
3% porn.

>>16566054
oh, I forgot the .0001% of the bots that run the stock market and .000001% spamming pol and leddit.

>>16566022
I'm looking to take the CompLing track in my cognitive science program, but I'd like to know more about the field. It seems to me that the predominant research is all about LLMs, AI, and AGI competition between startups. I'm not vouching for AGI, it seems intractable. So I'm trying to explore beyond the LLMs fad. I'm more interested in multilingual processing and translation. Are any other alternatives to LLMs development? Also, can you share some authors who focus on NLP? Thanks

>>16566295
>I'm looking to take the CompLing track in my cognitive science program, but I'd like to know more about the field. It seems to me that the predominant research is all about LLMs, AI, and AGI competition between startups. I'm not vouching for AGI, it seems intractable. So I'm trying to explore beyond the LLMs fad.
Computational Linguistics is a broad field, I'm more involved in text analysis, but there are other things to study related to speech production, historical evolution, and some other things. But back to text analysis (because that's all I feel qualified to talk about), you should get a really really solid foundation in both Formal Language Theory / Automata, and Abstract Theory of Syntax. To not be a moron, you should learn how statistical NLP was done before 2014, the NLTK book is a good applied book. There's also "Statistical Natural Language Processing" book, it's got a picture of two dice on the cover, published around 1999 give or take a year, and it's quite famous among people who know what they're talking about. At some point if you start doing research in 2025-2028 you will inevitably need to work with LLMs, but very little research is actually about building LLMs. There is a lot of research to do in finding their limitations and characterizing their outputs from a CL perspective.

>>16566295
Learn about compilers.