Google's NEW GEMINI Takes the Industry By STORM! (NOW ANNOUNCED!)

I'm going to be discussing a defining moment in AI which is Google's Gemini program and many people are really underrating exactly what Google just announced so please watch a video Until the End to see exactly step by step why Google is probably going to be beating gpt4 gpt5 in the coming years based on what this team has accomplished and based on what Google is about to do now it sounds like a crazy claim but just watch the entire video before giving your opinion and you're about to see why Google is about to shake up the entire industry now here is what actually shocked the industry this is a Gemini now Google essentially is always trying to push the limits of what they do with AI and essentially what Gemini is going to be their large multi-modu large language model now this is of course I think the chat gbt killer that they're working on because they said that Gemini is involving planning and memory now this is something that chat gbt currently does struggle with so maybe Google are working on something to wear the AI can actually remember exactly what's going on and plan like Auto GPT their tasks more effectively now I think you guys need to pay attention to this paper right here because this was something that I looked at around two weeks ago and essentially this is scaling to 1 million tokens and Beyond with rmt now essentially this paper goes over how they can essentially allow large language models to have one million tokens in a context size window what that basically means is you know how when you're on chat gbt and essentially you try to tell it to write on a thousand word paragraph and it only can write 500 Words and essentially after around a couple of paragraphs between you and chat GPT it essentially forgets what happens this will allow for 1 million tokens now the reason I bring this in to a Gemini is because I know that Google are always innovating on what they're doing now this paper was released recently which means that Google could potentially integrate this in to a Gemini now the reason I think 100 they're going to do this is because the token size window length is something that has been a long barrier for large language models but if you're wondering what is even the capability of this and how is this going to I guess you could say not outsmart gbt4 but essentially perform much more varied and longer tasks is because let's say for example you have an AI that can literally remember a million tokens it's going to be able to process stuff that as long as the Harry Potter trilogy on book and essentially this is going to have a lot more applications than something that can only remember 4000 or 32 000 tokens in a context window which is really good and really game changing so we know that Google is trying to push the limit with Gemini because of course they haven't released this just yet now remember this is also what Google have said is going to be the direct competitor to a chat gbt because of course as you know they did release Palm 2 which does cover many different aspects such as medpom and SEC Palm which are two separate divisions which SEC Palm covers cyber security and medpalm which covers Medical Health in the health industry now the reason this is going to be so interesting is because they aren't trying to just create a large language model that is really good they are trying to create a large language model that outsmarts slash beat gbt4 now if you're wondering whether or not this is just speculation remember when we made this video the other day where we talked about Google's official statement and just how much they are investing in order to beat gbt4 you see Google knows that if they don't catch up now they may never catch up ever because with each Progressive cycle the AI is going to get smarter and the main point that I'm pretty sure that Google is afraid of is the point to where the AI can then train the next level of AI so right now of course you know that there are millions of different people who are potentially working on training these large language models based on the parameters and fine-tuning it with human feedback known as RL HF but of course this is going to be something that in the future these AI large language models are going to be able to do themselves once you reach that point there is going to be that exponential growth in which these AR large language models can self-improve and of course you're not going to be able to catch something that's that far ahead and of course self-improving because these automated systems can be left running for quite some time and by the time they've upgraded it's honestly not going to be a competition at that stage so with this in mind Google is investing billions into Gemini and I think this might honestly be a very insane model because Google have actually talked about millions of different things that they're working on figuratively obviously not Millions but they've spoken about many different things that they are working on such as their recent wave LM music thing which is just out of this world and nobody seems to really pick it up which we will cover in another video but with Gemini we know that you know text is only going to be the first level of I guess you could say our interactions with these large language models because we know that Google is more focused on the multi-model aspects which they talked about now if you're wondering what Gemini is currently being used for it's not currently being used it's still currently in training and we are waiting for it to be finalized and it doesn't seem to be much but one thing that you should know about Gemini is that I do

think that it does have the potential to be better than gbt5 and although that is a bold statement you must directly understand where this statement comes from so in the presentation Google actually discusses who is going to be working on Gemini now if you actually understand what that team has accomplished in the past and remember they accomplished this quite some time in the past so they've been ahead of open AI for quite some time if they actually dedicate their workload to this I mean two teams combined into one combining all their resources and forces onto one large language model I'm pretty sure confidently that they can really do some damage in the AI space so for those of you who don't know what they've done take a look at this right here you can see that there is a list of things and I'm going to explain to you every single one that they've done so essentially the first one is alphago now alphago was an amazing breakthrough for Google because essentially what they managed to do was create a computer program that combines advanced search tree with deep neural networks and essentially they take a description of the boy board and essentially process it through millions of different neuron-like connections and they were able to beat pretty much the best player in the world at this and there was a huge documentary it was a huge deal like this was a huge breakthrough and this is something that they did in literally 2017. so this is absolutely insane okay now this is not just what Google has done okay remember this stuff is from ages ago okay now Google is always ahead and with a little push I think they're going to be even more insane so let's take a look at the next one Google strikes again this is insane so Transformers was a paper that Google actually released in 2017 and essentially what they did was they introduced this new architecture now what you don't know about Transformers though is that open AI actually used the Transformer architecture as the foundation for building chat GPT so the Transformer model has proven to be highly effective in natural language processing tasks including language understanding and generation and open ai's GPT is based on the GPT generative pre-trained Transformer series which utilizes the Transformer architecture so to summarize openai used Google's framework to build chat GPT and remember Google introduces framework in 2017. and of course we had Google deepminds wavenet which was introduced in around 2016-2017 where they were able to generate lifelike voices from Simple Text prompts and this is one of the examples there must have been moments even that afternoon when Daisy tumbled short of his dreams not through her own fault but because of the Colossal Vitality of his illusion it had gone beyond her Beyond everything so remember this is something that Google had done ages ago and then in 2023 there are many different programs that are out there that are far superior to this but the point is that Google created this quite some time back so think about it like this with this now a multimodal capability what do you think they're really going to be working on and of course how can we forget Alpha full now Alpha fold was a major breakthrough because Alpha fold was essentially something that allowed AI to predicted the 3D structure of proteins solely based on their amino acid sequences which is a challenging problem in biology and computational biology so Alpha forward made significant headlines and pretty much was considered a big deal for several reasons which we're going to go into now so number one was the protein structure prediction essentially determining the 3D structure of proteins is crucial for understanding their functions and interactions within biological systems and experimental methods for determining protein structures such as X-ray crystallography and cryo-electron microscopy can be time consuming and costly so essentially Alpha fold represented the Breakthrough in computational methods for predicting protein structures which potentially revolutionized the entire field now Not only was this great it was also accurate and reliable so Alpha Ford essentially demonstrated impressive accuracy in predicting protein structures in a competition so in Casp the critical assessment of protein structure prediction competition a biennial event where researchers assess their methods of protein structure prediction Alpha fold outperformed other participants and achieved remarkable accuracy results so this was significant because accurately predicting protein structures from amino acids had long been considered a complex and unsolved problem in biology and Alpha full did this with some ease so of course you have to understand as well that this impacted biology and medicine because essentially drug Discovery bioengineering and understand disease mechanisms Alpha Falls ability to predict protein structures can significantly accelerate the discovery and development of new drugs by aiding the identification of drug targets and understanding the protein to protein interactions involved in diseases so this is something that is truly game changing and this was done quite some time ago so this is why I say Gemini the entire team behind this AI program is literally mind-blowing and remember Google has literally done some other stuff as well there's so much AI stuff that Google has done the only problem I think that Google has is they've literally got so many different things that they can do they haven't really picked one thing to focus on the reason

chat TPT has been such a internet sensation is because the one thing that they focused on is large language models and they focused on creating a chatbot that is really effective Google literally has text to video text to audio literally has many different AI softwares that can literally do crazy things but for some reason they haven't released it yet they haven't fine-tuned these models yet but I think now as we move into the future with Google combining their two teams Deep Mind this is going to be next level so whilst this is being trained we aren't sure when the release date is going to be I do predict that maybe it's going to be sometime next year because of course Google doesn't want to lose to gbt5 I do believe that this one being multimodal is going to be truly incredible and I do believe that it's going to pretty much shake up the entire industry when it's announced