Google Gemini ai Introduction | Google Gemini ai demo | Google Gemini ai announcement

Google Gemini is facing lot of backlash recently due to its image generation methods I'm sure you would have seen the news of what is happening all around the world even the platform had to pull some of this functionality back and they will release it later in this video we are going to learn what is Gemini what it is capable of and I'm going to show you some prompting and we will analyze the response of Gemini let's do that so guys we are going to understand what is Jin first of all then we are going to see what makes Jin so special why there is a dedicated video for this and why we should learn about gin okay then I'm going to talk about modes of gin like what are the different modes it is offered and also the pricing part okay then I'm going to talk prompt responses live so we are going to ask some questions to jimin and see the response and analyze the response and I'm going to leave you with some next interesting things that you can do and how you can learn more in the space of gemin okay so let's start with understanding what is Gemini first of all okay now if you remember in the world of Google AI uh there is something known as Bard okay I'm sure you would have heard this name Bard so Google bard was basically a you can say counterpart from Google for chat GPT okay and now if you go to Google and if you search b right the first thing it will show you is G so I'm going to Google and I'm saying Microsoft Bing and I'm saying B Google okay so the first link it comes is gin that means that Bard is now Gemini so let's keep it very simple whatever was previously known as Bard now it is gemin okay with Advanced feature obviously okay so that is the first basic point of what is Google giny Now I go to Google and ask what is Google gin okay and here you can see that it tells me few very very important stuff Google jimy is a powerful AI model that can understand not the text but also images video and audio it is built on multimodality I'm going to explain you this concept guys this is very very important multimodality okay that reasion similarly across text image video audio and code and most important gin is the first model that has outperformed human experts on massive multitasking language understanding you see here massive multitasking language understanding so this is called as this is also called as MML so let me explain you a little bit what is mmu so M mlu right m mlu is basically a benchmark okay so suppose anybody comes and builds a large language model so this is my model one this is my model two this is my model three right so this is basically a benchmark which tests that what what is the response from the model from the minimal information supplied so let's say I have hundreds of research papers okay and I give all these hundreds of research paper to all these models to learn the pattern or gather the information Etc and then I ask questions to these models okay so MML is a benchmark that tells you that from the human experts are these models better or are these models not doing good as comp compared to the human experts okay remember Gman is the first model which is outperforming human experts in most of the scenario hence we are discussing gin here I'll take a pause please try to understand what I'm saying here the first model in the history of AI that passage that surpasses the human experts capability to answer the question of a particular thing suppose these are medical research search papers M1 M2 and M3 so a doctor can give you answer less efficiently then a model can give you answer if you feed the information to the model that is what Gman is doing and gmany has made a unique record of 200k research papers right I'm talking about 200k research papers in they say lunchtime lunch time means maybe half an hour to 1 hour so jimin has learned analyzed 200k research paper during that time and it is capable of answering better than human Experts of that field okay so that is the capability of Gemini now what makes gini so special so going back to the definition guys the reason I have opened this definition is this is important here okay Google J is a powerful AI model that can understand not just blah blah blah it is built from multimodality now let us understand this one concept guys you may face this in interview VI okay so I'm writing here multi modality multi multi modality now from modality don't think it comes from the word model machine learning model no it is something different okay so um I'll give you some some real examples here suppose you are watching unfold data science video for some time okay so I write the word here Amon okay let us assume for a moment that you don't know any other Amon in your life you only know me so at the moment you see this right maybe you will think the discussion is about me okay now let me show you one picture let me show you one picture this is me okay let me make it little big this is me okay at the moment I show you this picture you will say that oh I know this guy this is Aman who creates videos on unfold data science okay so I'm talking about two modes here one mode is text mode okay remember this is text mode I showed you a picture that is what mode that is viewing mode okay so viewing mode means I'll just say View and then there can be a listening mode also suppose uh I'm not writing I'm not showing you the picture but somebody comes and says Amon in your ears or you know you hear the word Aman so that is a listening mode now one important thing to understand here is this is mode one of learning okay mode one this is mode two of learning okay mode two of learning and this is mode three of learning so I'm talking about learning from different sensory organs so learning from eyes learning from ears learning from nose learning from touch learning from different ways okay if the learning happens in this way okay this this kind of learning is called multimodality or multimodel learning so this is basically education concept so imagine imagine that in a class there are let's say 30 students okay student one to student 30 okay now some of these guys some of these guys may be good in learning by reading reading and writing some of these guys may be good in learning by listening okay so I'm saying these guys are good in learning by listening and some of these guys may be more more good in learning by looking at the pictures okay so the they will be different guys so I'm saying let's let's take this color I'm saying three modes three modes of learning okay if we combine all these three modes maybe we can efficiently train this class in a very good way so what Gemini doeses is it trains your model based on multiple uh multi modality of learning okay so for example if Gemini has to understand how a carrot looks like okay so for example how a carrot looks like um Gemini will give this carrot keyword okay I'm just telling you very high level what will happen Gemini will take a carrot keyword and Gemini will create a uh Vector from this word Vector okay so let's say this is Carrot words Vector okay similarly Gemini will take image of carrot image of I'll say carrat and Jim will create a again a vector for this right a vector representation of that image and then it will combine both these vectors and it will try to say to the model hey both these things mean carrot and you have to learn this tomorrow somebody either shows you a carrot or somebody either speaks carrot you should understand this is this okay okay that is the concept of multimodality now let me go to a very interesting Google page and show you what I'm talking here so let's go to this page here see this one developers. gooogle blog.com I in okay what I'm trying to do is I'm trying to show you how B um b or J what whatever you call it how J will behave when you give this so if you show this image right and you ask tell me what you see gini will tell you I see a person's right hand the hand is open with the fingers spread apart okay and then if you see let's try this one Gemini will say a person knocking on the Oden door and if you do like this V Gemini will say I see a hand with two fingers extended which is a common symbol for the number two all these are individual information now with all these images right let's Club text information okay so what I'm doing here what do you think I'm doing hint it's a game at the moment Gemini knows it's a game and it it you know combines with the information of the images right immediately it tells you are playing rock paper scissure Okay so that that is how multimodality works you know we are trying to make the model learn from the various modes of um pattern you can say or information you can say so this way what what what is written in Google uh blog also is everything we just did is now example of multimodel prompting okay that makes your Gemini very very special now let me go back to my note and try to show you what are different modes in which Gemini is offered so basically three modes uh these are good things to know uh one one basic mode is called Nano okay one basic mode in which Gemini is offered is called Nano another mode is called Pro and another you know highest mode is called Ultra Ultra okay so as it is clear from the name right if you go like this from top to bottom your capabilities will increase multifold and obviously the cost and infra needed will also increase so Nano is very very simple that can run on your mobile devices such as Google pixel also it can run on Pro is little Advanced version some complex calculations some complex computations Etc some complex coding Etc can be done on Pro and Ultra is the highest one which has as I was telling you right outperforming humans on various occasions and it has outperformed GPT 4 on various uh various tasks okay so Ultra is the highest so Nano Pro Ultra three versions of this let me check check in my notes if something I'm missing okay and yeah the the money part right so if I go to Gemini here for example if I go here b. Google if you search and G homepage if you go right then you will be able able to see that it it opens me uh a Gemini homepage like this obviously there is an API also through which I can connect but today I I'm going to show you prompting through this only later based on your response I can show you uh the API way as well and if I click on this there is a option of upgrading I'm not able to upgrade because something is happening in UK my um um my you know Google Google is I'm not able to make the payment to Google for some reason uh it tells me that your country is different or some some issue is coming I'm not able to do that but if it's available in your location you can try first two months it's free and after that it will cost you so if you want you can cancel before your UH first two months gets over okay so what you can do you can just upgrade it will ask your card number you can give it and you can do it but immediately it will not not take any money after two months it will take so before that you can cancel if you want so what is the difference between basic and upgraded is in upgraded you will get as you know any any paid platform you you may get right Advanced reasoning capabilities some customizable capabilities Advanced coding capabilities okay so all these things you will get when you upgrade it but this basic one also does lot of things in a very nice way so let me try to tell uh show you what what all things it can do in a lovely way uh so I'm inside G now and let me so let's say I'm in Bangalore and I'm planning to go to Kur this weekend okay so I'm in Bangalore Bangalore and plan to go to go to kurg this weekend okay I have a small baby with me please plan my trip okay so let's see I'll tell you how Google's response is different from rest of the uh large language models okay so you will you will understand by by the responses okay so what it is telling me is travel considerations mode of Transport alternate transport packing ET because kurg is a slightly cool colder place so it tells me warm BL blanket Etc accommodation ideas ET day one in Fall seat whatever the um location in cor I should go okay so it tells me flexibility pce yourself weather Etc now I'm going to ask very specific question how much I need to walk to visit important places if I go by car okay so if I'm going by car since I have a small baby so how much I need to visit okay so it should tell me now now one important thing is since it is a Google tool right so it can very well integrate in the background with maps and get the exact location and exact distances just that my prompting should be very very proper and I should I should be running on a um I should be tweaking my prompt in such a way that I should get that response okay so it tells me minimal working for now but remember it can integrate with Google Maps and it will give me the exact walking distance exact thing okay based on the Google Maps what else you can do in in uh jini is I can ask a question like this so currently it will not work because in the UK my location that functionality is not working they are rolling out in phases currently in this location I'm not able to see that maybe in your location it may work so what I want to ask to J is so it will integrate with your Google drive it will inte integrate with your Google Docs it will integrate with your YouTube and it will integrate with your Gmail okay so you can ask a very simple question here for example uh from my Google Drive from my Google Drive okay uh you can say which PDF talks about my offer letter my offer letter and what was my joint joing bonus okay so what I'm asking is in my Google Drive there is a my offer letter I have kept in that my joining bonus is there so I'm asking what was my joining bonus so go to my Google Drive and find this information so it may not be able to do this currently because um that integration is not there but it is quite possible in this okay same thing you can do in Google uh YouTube and Gmail as well okay now what I will do is I will give a Instagram image that I had put for uh some learning okay in unfold data science and I'll simply say explain this image explain this image so this image is basically let me show you what what that image was one I think this image only systematic versus random error I I just posted on Instagram same image I'm giving to Gemini and asking explain me this image okay so let's see what the response is so it it clearly tells me systematic error what what example analysis and random error is this key point also it is giving and it is telling me let me know if you need any details so accurately it is going to the image and giving me the relevant information okay now if I ask Gemini to create give me now I'm asking about creation of something give me an image of students learning data science okay give me an image of students learning data science so this is a very common picture of data science that all of us know right so it is taking basically from The Medium but I'll simply say there are there are no students in this image okay so now with students and the image is coming so what I'm trying to say here is uh when Google creates any tool right it puts lot of thought behind it and Google has lot of data with it right so it can integrate with our day-to-day um components that we use for example Google Docs or Google forms or Gmail or YouTubes or Google images whatever it is right and then it can give you very very relevant results okay what I want you to leave you with here is first thing I will just type few things here first thing is uh if in your location upgrade upgrade option is possible right then you upgrade okay don't worry for first two months it will not be charged then you can you can cancel okay upgrade and experiment with I want you to experiment with something very very uh complex complex such as maybe um you can try uploading some of the documents about which you have information and ask the question which which you know you you you were not able to understand or maybe a research paper which you want to understand or maybe couple of research paper which you think is difficult to understand maybe a complex mathematical formula okay those things you can upgrade uh you can experiment and try to see the responses here and third thing we have not gone programmatic yet okay so we can very easily go to the program programmatic world and do all these prompting through apis okay I'm going to explore this as well if you want me to programmatically access and programmatically show you how things work write me in the comment that I want a video with the program in API mode and we will do that but I will simil I will simply just recap what what I intend to intend to cover in this video is basically B has now become zini the first tool in the world of AI that is outperforming human experts on various occasions and that's why people need to talk about it okay there are different modes if you pay money you get a more powerful version of it very simple and the lowest version can run on mobile phone as well prompt responses are excellent okay in my system it is not integrating currently with Google Maps Etc or drive or whatever but it does okay and then do some experimentation and see how it is performing if you are able to upgrade and then the API part we are going to see as next video okay I'll see you all in the next video guys please give me a thumbs up if you like this video wherever you are stay safe and take care

Channel: Unfold Data Science

Views: 825

Keywords: Google gemini ai demo, Google gemini ai announcement, Google gemini ai release date, Google gemini ai how to use, Google gemini ai trailer, Google gemini ai tutorial, Google gemini ai launch, Google gemini ai app, Google gemini ai duck, Google gemini ai fake, Google gemini ai video, Google gemini ai review, Google gemini ai explained, gemini ai google, gemini ai controversy, gemini ai backlash, gemini ai, gemini ai vs chatgpt, gemini ai explained, unfold data science

Id: 6HC07CdOPgw

Length: 21min 8sec (1268 seconds)

Published: Wed Feb 28 2024