Agentic AI Is The Future- GEN AI Trends In 2025

Krishak's YouTube video discusses key AI trends for 2025, focusing on practical applications rather than theoretical comparisons. He highlights the rise of agentive AI, predicting increased job roles in building autonomous AI agents. Secondly, he emphasizes the growing importance of AI infrastructure providing GPU access and efficient inferencing engines for large language models (LLMs). Further trends include the development of even larger LLMs and vision models, along with the increasing use of AI edge devices for generative AI tasks and smaller, more efficient models. Finally, he notes the growing importance of AI security regulations and the expanding availability of open-source models, moving closer to the potential creation of Artificial General Intelligence (AGI).

hello all my name is krishak and welcome to my YouTube channel so guys this is a very interesting video for every one of you out there and I try to make this video every year specifically at the end of the year I say that hey what you really need to focus on the upcoming year right let's say now 2024 is going to get over in another 5 to six days and uh now we're going in 2025 right similarly I made this kind of video in the last year I told you hey you really need to go ahead and focus on gender application you need to focus on mlops tools you need to focus on creating amazing data pipelines at all and uh based on that I I made sure that I uploaded all those kind of videos in my YouTube channel along with that I also came up with yemi courses and trust me the kind of impact that has probably made in professional lives it's quite amazing right and many people were able to get jobs and all also now similarly I really want to talk about the AI Trends specifically in the field of generative Ai and what you really need to focus in the upcoming year that is 2025 okay so that is what I'm actually going to cover in this particular video and based on this saying whatever I'm probably talking about this particular Trends you'll be able to see similar kind of videos much in building applications will be uploaded in my YouTube channel based on the trends that we are probably going to follow in this video right so let me quickly go ahead and share my screen so here uh I've just written this particular title the influence of AI 2025 Trends you should not miss and definitely you should not miss guys for all those people who are probably following me every year you know you'll be able to see right initially we were we we probably discussed everything in my channel from scratch you know from mathematics to building applications to creating end to-end projects to probably deploy the solutions along with data pipelines and many more things i' I've probably uploaded more than 2,000 plus videos in my YouTube Channel right now I'm going to talk about the influence of AI or what is the AI Trend that is probably going to happen in 2025 so let's go ahead and discuss one by one right so the first thing first thing and I I can probably 100% bent on this particular topic it is nothing but agentic AI right so agentic AI I've already made a video I've also uploaded some of the projects uh you know in agentic AI it basically means right it is nothing but you are trying to create autonomous autonomous AI agents which will be able to solve very complex workflow so this is really going to probably be in 2025 you know you will be seeing amazing Frameworks you'll be seeing amazing Frameworks you'll be seeing amazing tools you'll be seeing amazing techniques to probably build agentic AI application right so I can 100% and say hey this is definitely going to happen because now many many many companies are specifically working on this right and because of this you can just see that complex workflows complex task can be easily solved you know without much human intervention and this way definitely I can guarantee on this your productivity your productivity will definitely be increasing right so this is one of the topic I am very very much you know I I'm I'm pretty much guaranted with this right yes in 2025 you'll be having lot of job roles where which will probably focus more on building agentic AI applications it will be building agentic AI Frameworks and manymore right some of the Frameworks even though I'm going to discuss it in my YouTube channel I just write it down over here uh some of the Frameworks that you have is like crew AI you have something like Lang flow Lang flow you have Lang graph right Lang graph you have uh fi data right so I will also be uploading a detailed playlist on each and everything as we go ahead so everything we will discuss in the form of practicals right but what I feel agent AI is definitely going to go uh ahead and even you know the CEO of Microsoft has said that agentic AI is probably going to replace even SL platform it is true because the kind of applications that you can probably see with this it is quite amazing okay now coming to the second thing if you probably see over here uh after agentic AI you know we are also going to see a lot of companies which will be coming up with amazing AI infrastructure okay and the main aim of this AI infrastructure will be specifically two things okay I will go ahead and write it over here one is providing GPU as pass pass basically means platform as a service right or uh pass basically means platform as a service like you have this AI infrastructure exess and you will be able to use their gpus to probably train the models and all right so this is one thing and the second thing that you may be able to get is nothing but inference inferencing inferencing engine inferencing engine or I can also say inferencing compute engine okay inferencing compute engine now see guys what does this basically mean right when I'm saying inferencing compute engine now you know that there are lot of different different LM models coming up right llm models coming up it may be open source it may be um you know paid right Opia is also coming up with meta is also coming up different different kind of llm models are coming up and these all huge companies are in Fierce competition for a small startup it is not possible that you probably try to create your own model completely from scratch unless and until it is not linked linguistic and all right if there are some companies who are specifically working in linguistic llm models definitely they can do this right and they can actually use this kind of AI infrastructures from various companies who have tie up with nvidias and all I've seen I've met multiple founders of different different AI infrastructure infrastructure companies who are investing a lot amount of money to provide GPU as a uh as platform as a service and along with that they also providing inferencing computer engine for the inferencing purpose right now when they are coming up with this kind of llm models right they definitely require huge amount of data and these models usually will come up within millions and billions of parameters right billions of parameters now when your model is of so huge size the major aim is that you need to be able to access this models in uh in a way that you should be able to get a response very much quickly and that is where inferencing computer engine will be coming up in uh word like coming up coming up in that specific kind of use cases right you need to have a technique wherein you should be able to get the output or the inferencing of the output in a much more quicker way right so many many companies are going to specifically work on this uh I don't know whether you have heard about this or not Nvidia has come up with an amazing new GPU right it is something called as Blackwell Blackwell gpus I guess the name is Blackwell gpus and in recent event right and we they had also announced this uh Jensen who is the CEO of Nvidia he had specifically announced this and if you talk about this particular GPU it is definitely used for both training and for inferencing purpose right and what Nvidia is basically doing it is partnering with various AI infrastructure companies you know where they provide you this kind of access so even in Amazon you'll be able to get this particular access as we go ahead so this is the main aim you'll be able to see that more better AI infrastructures will be created for both of this particular task Okay the third thing that what I see over here is that as we go ahead you'll be able to see that we will be seeing more more large large models now when I say more large model I'm talking about both llm models llm models uh or you we are also talking about uh large Vision models right so both the specific models now right now we have billions of parameters as we go ahead you know you'll be also seeing trillions of parameters so that Fierce competition will definitely be there both in open source both in paid Source right like companies will keep on keep on uh working on this there'll be in a Competitive Edge to probably come up with this things okay then uh the fourth one that you will be seeing is something called as AI Edge devices for generative AI Edge devices for generative AI for Gen a okay now what are these Edge devices I hope everybody has heard about this Jon Nano and all now recently Nvidia has even come up with this new product that is called as Jetson Orin now what is basically Jetson oriented is it is again an AI Edge devices which supports uh which supports for tasks that are related to generative AI so this kind of work is also going to specifically happen because we are going to do some amazing work in this specific field right so these are some of the points that you also really need to focus uh and you will be seeing that this kind of Trends will start happening you know I will be soon getting this Jetson Nano so Jetson uh Nano I think it the product name is Orin and I will also show you a demo how you can probably run an object detection and that will be specifically for your generative AI use cases okay so I have probably included most of the topics that are there since we are talking about large models there is also one more amazing thing that will be coming up that is nothing but small models also right now if large models is there uh to run this kind of models in the a Ed AI devices in the AI devices we also require small LM models it can be llm models it can be lvm models right so here also we'll be having this one of the example is 53 right 53 is just like a small model from Microsoft which is open source you know and here if you have probably seen this this has very less number of parameters when compared to that very huge llm models but this will efficiently be able to provide you the right kind of response that you're specifically looking at right so this is where you can probably see the AI Trends is specifically going on uh one more thing that you will be seeing uh as we goad ahead the security that is related to AI will now keep on increasing now more regulations more regulations are going to come up since now the models that are probably created you know they're trained on some kind of data right they're trained on some kind of data so now as we go ahead you know this kind of models are going to become much more bigger when compared to parameters right we compared to different different parameters so a lot of regulations are definitely going to come up from different different countries and already many many count count have started writing some kind of regulatory Frameworks so that they can efficiently or ethically use this kind of models for uh ethical purposes okay so that is the main thing that you'll be able to see over here right uh so let me just think if any topic that I've actually missed over here yeah this looks good specifically now you'll be also seeing one amazing thing lot of Open Source models is also going to come up right open source models is also going to come up and why this is basically happening all thanks to Nvidia now envidia what it is doing is that it is partnering with multiple companies to create linguistic models from different different different countries you know it has tied uh it has having a partnership with now uh goo you know in India even in Vietnam recently uh it has done some kind of tie-ups over there in Indonesia uh Indonesia also they have have probably done some kind tie up and that main aim over here is to basically create linguistic models and this linguistic model should be open source and they should be able to use it much more efficiently so this kind of Open Source models is going to come up more and more as we go ahead and finally what I see one more Trend we are very near to create AGI right AGI yes many people will think that there's still more time but companies are doing their they they are trying to come up with this particular AGI you'll be seeing more good good models as we go ahead you know so these are some of the important AI trends that you can see and based on this you can also plan your learning process so that it becomes very much easy even to probably get jobs in this kind of sectors because more and more job job openings will specifically happen in these areas with respect to this particular trends that I'm actually talking about so I hope you like this particular video yes first of all my main aim is to Target each and everything that I've discussed over here even in my YouTube channel talking about the application part building part and all I'm not going to talk about comparison or which model is best and all no I'm going to focus more on application and Building Products this is what I'm actually going to do right so I hope you like this particular video this was it from my side I will see you all in the next video have a great day ahead thank you and all take care bye-bye