An intresting thing I found was that this is slightly different from Gemini Pro. If you ask it's name, it will not give any name. If you ask "are you powered by Gemini", it won't say it's powered by Gemini. I did manage the LLM to tell it's name once, to which it replied "Serenity"
Thanks I started following your great tutorials… I would like to do many this… mainly chat with a book…. Convert pdf to text file with ocr (text is images in pdf) … detect anomalies in large text file eg. foreign words (to the language) … have double language chat e.g. German and Spanish in same chat …
8k tokens output and not ready yet for businesses. Google did though provides regional based access albeit no european based yet. Overall its currently faster than cgpt although the contents somewhat need to be adjusted as required
I made a similar video as this! After playing with this for about 24 hours, I'm finding Gemini is very buggy (chat triggering exceptions) and is suffering from a lot of hallucinations. It's been giving me links that lead to 404 pages. One example, I asked it to explain binary to me where it started off ok but then proceeded to calculate 1011 (11 in decimal) and added it backwards (left to right instead of right to left) and got 13. SMH. I'm sure they will work it out but this is not ready for general use yet. Anyone else having experiences like this?
The model we use in the API is not the same on Bard. Bard is more verbose while on the API it is 'direct'. Like when we ask "what is the 3rd planet of the sun" Bard gives a more verbose answer, while the API just gives "Earth". Maybe a system prompt can make it more chat friendly.