- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
- DeepSeek has rapidly accelerated AI adoption throughout Chinese society.
- Chinese tech giants have released a wave of free, open-source AI models.
- Startups are under pressure to abandon foundational research as investors prioritize practical applications.
Practical applications for LLMs such as? There are a lot of claims there but not much substance.
Sorry, but we’ve leaped through this cycle of the next ChatGPT revolution, followed by the next GPT killer, followed by the whole technology being overhyped anyway too many times. It‘s weird to see OpenAI being portrayed as this big villain but it‘s a triumph when others do it.
Truth is they spent much more than they claimed on this and want a return on investment by aggressively pushing it into applications that don‘t need it. It‘s the same old tale every time.
Practical applications for LLMs such as?
Translation and voice-to-text (and vice versa) applications to name a couple.
Personally, I have been using LLMs to help me code. They are not perfect, occasionally generating nonsense, and some are downright awful (looking at you, VSCode Copilot). But when they get it right, they saved me so much time.
I know LLMs are associated with tech oligarchs currently, but to dismiss all LLMs as
uselessimpractical is downright unfair.Yeah I despise LLMs ( really they are fine what I actually hate is how much it’s getting shoved into things) but I won’t act like they don’t have any use cases. It’s a bit obstinate to act like they have no use.
For coding, I treat it as an overly enthusiastic junior-mid.
It’s also great because I’m learning terraform and I just don’t want to look up the syntax. I just tell it to add something and it’ll do the lookup for me.
It gets it wrong sometimes which is where I need to do the lookup but it definitely has its uses.
DeepL works pretty good without word salad and it‘s getting better every day. I also don‘t think the Chinese market cares all that much about it.
LLMs will surely remove entry level coders… and some years from now there will be too little senior devs to debug it all. Some see a quick way around a problem, I see a problem in the making.
As for the last part: „Useless“ is what you said, not me. It surely isn‘t as useful as it‘s sold and that writing piece is a prime example for that. The point stands that it‘s sinfully expensive to develop despite what they claim so they need to sell it no matter what it takes. It‘s mainly their incentive and dishonesty I am criticizing.
I use it for translations regularly, and also tts. It works well.