Speaker 1: In the next eight minutes, I'm going to show you how you can cancel your ChatGPT Plus subscription along with Perplexity Pro so that you don't have to pay $23 every month still yet use the best model in class with web search that can give you a research assistant without having to pay any single person or any single company. So if this is somebody from OpenAI or Perplexity watching, I'm really sorry. One of the best AI models DeepSeek R1 is now powered also by search data. All you have to do is you have to first go to chat.deepseek.com and then you can go ahead and then give any question like I want a research report on alternatives of hip joint replacement. Enable the think, enable the search, you have a research report at the end. The report is not just random report like what ChatGPT gives you. So I've done the same thing with ChatGPT and you can see that it thinks about it. You can see the thinking process but you cannot see all the good references that DeepSeek gives. It's more like a research paper but if you have to properly get research report with citations like from which website it is getting something then I would say DeepSeek R1 thinking plus the web search is like a complete package for you. So how do you do it? Just simply go here and new chat and after you go there all you have to do is enable these two. By default these would be disabled. DeepThink R1 search. First enable this and also enable this. So what this is going to do is this is going to first get the entire search data and then think something and give you the final response. Let me go ask a simple question. One thing is like you need to understand when do you want to use it. You can't use it for all the prompts. Maybe a bad idea if you want to use it like oh I want to do what else 2 plus 2. I mean if you want to do what is 2 plus 1 you do not need web search. Then you do not need a thinking model. So you need to be very clear about what you need to use it for. So you have to be mindful about it for these kind of simple prompts. You can just new chat disable these two what is 2 plus 2 and then you can just send it and then it will just do 2 plus 2 equals 4. So for a very stupid prompt like this you don't need a thinking model. You don't need web search. But let us say that you need web search for something like where you are working on a project. You want some guidance. That is exactly where this model comes in. And then it is quite handy and very successful like the question that I asked. You want to understand what are the alternatives for hip joint replacement. Then you can probably go ahead with this model and with research the search and then you are basically putting together a team that can do search and then think through it discuss through it and then get back to you. So what I'm going to do right now is I'm going to just go ahead and then ask about something called lightning attention. Lightning attention is a very new thing that was launched by minimax. It's a it's a different way of approaching transformer models. So I want to see if it can go to the web search about it. Get back to us with a report. OK. I want a small report about lightning attention from minimax. Make sure you also add a table of differences between lightning attention versus Mamba versus the traditional attention. OK. So we have enabled the search. We have enabled the deep think R1 model. And as you can see here first it found 45 result. So it's going to take the result. Now it's going to do the thinking part. So as you can see here it started with thinking but everything that you are seeing is happening in real time. I'm not like editing this video making a cut. So you can see literally it is happening at the time. One of the best things about deep seek R1 model is the speed at which it can do thinking. So I can go ahead and then give the same thing. So I'm going to go to chat GPT click new and I've got O1 and the bad part is search is not even available if you happen to use O1 there. So that's a shame. So I have to go here and give GPT 4.0 enable the search give the same thing and then send it. So it's not even a thinking model in this particular case what chat GPT offers you. This is mind you this is after I pay twenty three dollars including taxes every month. So without having to have chat GPT plus without perplexity pro you can do this with this model deep seek R1 here. Why is it not chat GPT. OK let me refresh it again. I've got the model here. Chat GPT is not liking me I think. Don't tell me the server is down please please please please. No. OK. I'm going to go back here. And as you can see here this is done. It says lightning attention in minimax O1. So it figured out that it is an O1 model. That means one you can have the initial confidence. Lightning attention is a core innovation in minimax O1 a family of models minimax text O1 minimax VL O1 the visual language model designed to produce process extreme long context up to 4 million which you have seen if you have seen my recent video you can see while maintaining the computational efficiency and performance comparable to GPT 4.0 and Claude 3.5 sonnet. Then there is a reference one reference two and you have all these things and you have got the table as well which we exactly ask lightning attention this their mamba is their traditional attention is there. Now traditional attention is quadratic computation required which I made a mistake when I made the video. I said exponential like I just threw it in there. So a lot of people pointed out so which is good thing. So traditional attention quadratic computing computation requirement lightning attention is their linear mamba is their linear. So this is the biggest advantage here. And we have got references. We have got one we have got nine. The good thing about references itself is that it is not giving me any like random crap as reference. So the references like I think in SEO the search engine optimization language we would say high domain authority. So it comes from high domain authority website and it yeah it has given me a report. Meanwhile I have to pay twenty three dollars with chat GPT plus and also perplexity pro whatever that you have to pay. I think if you do not care about like a private search a lot like if you want to consider yourself that you know you don't care about what the company is going to do with your prompt then honestly you can close your eyes blindly go to chat.deepseek.com and then start searching for anything and it is not just for like research reports like this. So I can give you a program programmatic question like a programming language related question as well. So for example if I want to do something so I can go say I want to design a Manny animation and I want to explain Pythagoras theorem with that help me. And now it's going to search the website and OK in this case I think it realized that it doesn't have to search so it probably didn't search here. So it is just going to do the Pythagoras theorem or Pythagoras theorem just by itself. So this is an excellent feature a free website. You can kill your perplexity pro and chat GPT plus using this free website. It's a thinking model. Let me know what you feel about this model. See you in another video. Happy prompting.