DeepSeek.com making waves!

Guys now it is telling me that “the info is beyond it’s scope”. Bachha hai thoda time lagega. However I expect it to slander India at any given opportunity in the future.

I believe only US and China are in a position to develop such AI models and not other countries like India. That is because they are using all the data stolen/stored/hacked/leaked from all over the world. Indian companies will gladly sell their data to the data “dalals” for few pennies. There were talks of Mr. Mahindra investing 10million $ and backing out as it was beyond his scope.

But I believe that India is in a dangerous position and we must develop a spine and create some AI based model catered towards Indians.

@vishalrao
I am using deepseek app downloaded from iOS app store
 
Last edited:
i feel it might be china's dumping technique to crush other Ai players and then dominate the market
 
WhatsApp joke.

First people claimed AI will take Human jobs, but now AI is taking away AI jobs :D
 
One day Indian govt. will be caught with pants down selling data to enemy states and we all will be "WAH sarkar WAH, Cinema kar diya"
 
Running the 8b params model locally on my mac. I still feel chatgpt provides better answers to my queries, maybe the highest param model of deepseek would give better answers. Leaving the hype aside, I believe China has done it's due diligence of making innovative technology cheaper and available for the masses meanwhile Indian industrialists are busy delivering milk packets and veggies under 5 mins 😂

I haven't tried programming on deepseek r1 as of now, if anyone has tried it let me know. I think it would perform much better than chatgpt which honestly produces the shittiest code known to mankind.
 
Last edited by a moderator:
Tried asking a few moderately difficult coding problems to both chatgpt(4o) and deepseek(r1), not to generate code but to explain a few stack traces and error messages. I still felt that chatgpt was able to come up with better responses overall, deepseek model would just get confuse and eventually returned a completely garbage solution which was fundamentally wrong.

I don't really know what is test data set they have used to "surpass" OpenAI's model


Source

We will know the truth once hugging face comes up with their own benchmarks with similar datasets on both models.

LLM is that uncle that knows everything but is completely wrong at times 😂.
 
Last edited by a moderator:
Now the API is also fully censored.

wirlKwF.png


muCNuhJ.png
 
You must log in or register to reply here.
Back