why scrutinize if content was ai-assisted? in the era of llms like #ChatGPT , we should embrace their use just as we adopted spelling and grammar checkers. ai drafts can enhance writing while human oversight refines and finalizes the work.
- claims to outperform gpt-4 on indic languages across multiple tasks
- utilized 10^23 FLOPs for training.
- 20 times more indic tokens compared to llama
- next model is being trained on multimodal dataset - vision, audio and text
- team is working on custom chips in-house
India's first AI foundational model, Krutrim, was recently announced. Here's what we know so far:
- trained on 2 trillion tokens(similar to llama2) of desi data,
- understands 22 Indian languages
- can generate text in 10 languages
๐
#Krutrim#AI#MadeInIndia ๐ฎ๐ณ
6/6
In this context, think of a function as an "Intent," and the parameters as "slot values" similar to those used in training NLU models for chatbot development. It can also be used to extract structured content and obtain JSON responses.
Happy Exploring!
The latest ChatGPT release just unveiled an incredible function calling capability, and it's pure magic! Now, you can effortlessly transform any message into a function call.
Full Thread (1/6)
5/6
You can call the function based on the response and add the result to the function role message. It also keeps track of the state and updates parameters based on new user messages. so, if we send a new message like "no, find me a Japanese restaurant"
Back in 2015
Interviewer: Do you think computers are conscious?
Scientists: No way! Computers are far from it
In 2023
Interviewer: Do you think computers are conscious?
Scientists: The question doesnโt make much sense because we donโt even know what consciousness is.
#GPT4#AGI
Hey @makemytrip@makemytripcare , I'm trying to book a flight on your website but the final payment is showing differently than the payment being charged when I scan the UPI code. Can you please clarify why there is a difference? #makemytrip#flightbooking#paymentissue