gemini meetup

gemini primarily on US english accuracy of data and how grounding transformer models is what is the weight for each data item, giving it more and more data and asking it to learn from it llm are neural netowrk abstractions layer " something not allowed" entire english language is transformeer model supervised and supervised labeled examples, vs pile of mysetery data correlation and causation, correlation give any 2 parameters and ai will find how they are related, causation have clear logic on how the 2 are related gemini is natively multimodal lot of modals solve one problem really well in real life want a multimodal approach nano on edge devices like camera, model can look at the image and do face recognition ultra have real time conversation have voice text and messages, process of adding videos gemini natively multimodal sophisticated reasoning explain causation advanced coding google ai studio ; ui to play with ai choose model 1.5 is out, can log in with gcp account, up to date model stable model is more consistent gcp account free to create and get 300$ credit to use https://gemini.google.com/ to use it like chatgpt ask questons from images receipt from restaurant, asked question cost of single mango lassi and it scanned and did the calculation and explain in the prompt public version has limitations gcp has demos gemini pro vision accepts text and imagery as input and text as output gcp go to vertex ai examples and then you can get the code and modify it as well https://github.com/GoogleCloudPlatform go to the gcp, give it a site, or upload all your files and then ask questions from it 1 million searches 5$ install vertex ai sdk for python honda , you tell something happens, and then it asks you questions to find what the problem is multi modal and sizing of the model and developer support ecosystem google labs

experiment