In the last 20 years, many search engines have come up like yahoo.in, dong ek, but google.it is a trillion dollar search engine. Microsoft’s massive distribution was also there, creating a buzz as well. In google2, a middle-class Indian boy launches his own startup Perplexity, which immediately becomes huge. Perplexity is an ancient startup looking to take over Google’s startup, the co-founder of a startup that’s just a name, lots of money around us. It’s another fast-growing player in the space that wants a much bigger piece of the pie. Confusion is another fast-growing player in the space that wants a much bigger piece of the pie. Confusion is another fast-growing player in the space that wants a much bigger piece of the pie. So what happens is that in December 2022, Google3 launches GPT, which acts largely as a search engine. You can ask it direct questions and it gives you direct answers. At the same time, Google2 was a company of people, but the game changed then. When the trillion dollar giant took 50% equity, Google2 wasted no time in announcing an emergency situation code rate and stopped Google2 projects and gave all attention to reviewing this situation. After that, in response to Google2, it started launching new AI models Bird and Jamna, while Google didn’t think of entering because training AI models cost a lot of money and the two players got billions of dollars in value and they became fast growing AI startups. So in this video we will talk about what Arvind Srinivas saw in the search engine space that made him decide to compete with the trillion dollar giant, what strategies he used that made Perplexity one of the fastest growing AI startups, can we ever learn important lessons from google3’s confusion? This story begins in 2017 when Arvind Srinivas completed his engineering from IIT Madras and soon after graduation, he joined Open AI as a research intern in 2018 and a year later in 2019, he experienced the architecture and development of google2 and GPT3. Then in 2022, he launched Open AI which went extremely viral. Now Arvind sees two major problems with Chat GPT. First, GPT cannot respond to recent news or real-time information because these AI models are only trained up to a specific date. If we ask them some questions related to current affairs, these models are unable to answer them, and the second problem is that whatever answer Chat GPT gives, there is no mention of the source anywhere, so the user cannot verify whether the answer given by Chat GPT is correct. And to solve these two problems, Aurobindo uses Perplexity A. Here Perplexity is an AI powered conversational answer engine it’s not a search engine, so you ask a question you ask a question it answers you directly The main difference from other chat bots like ChatGPT is you always get sources so it only says what’s already on the web and it clearly tells the user where the answer to the question is coming from on the web. That when both ChatGPT and Jamna couldn’t solve the problem of source verification and real time information in their AI model how did Perplexity do it and moreover Arvind didn’t have money to train his AI model well to solve these two problems when all the billion dollar companies were busy building their own AI model, it didn’t build its own AI model. These companies have used AI models in their products. So basically Perplexity integrates search engines with these AI models.
Now what’s the point, I’ll explain to you with a small example, if you ask Chair GPT to tell me today’s AI news, he won’t be able to answer because it’s only trained up to a certain date, but Perplexity has built a bridge between the AI model and the search engine, so that if we ask Perplexity to tell me, AI will tell this news directly, but AI will not tell this news directly. It will first go to the search engine, the search engine will go to different sites and it will feed its feedback to this AI model, so now the AI model will have the context of the latest news and it will find the important things from all these articles and write with the source. Now using this strategy, Perplexity gets three big benefits. Number one, every week a new AI model is launched, so all these AI companies fight among themselves and spend all their time to upgrade AI models, like recently Deep Suk came, Google launched its O3 Mini model, while Google 3 Minus is spent on improving these AI searches. Number two, each model has its own specialty, some are good at reasoning, some are good at writing. Confusion gives the user the option to choose the model, so that the user can choose the model according to his usage. Number three, now many models are censored on a certain content, such as Gemini does not give political opinions in deep research, while Deep Sac only answers in support of China, and because of this, there are many legal cases against these models, such as recently many media houses filed copyright infringement cases on Open AI, while Deep Sea is also banned in many countries, but it does not need to have its own model in Perplex City. Any legal issues if a model is sued. If there is a case, Perplexity can move that model to another AI model and because of this, Perplexity AI grows very fast and becomes a major player in the AI industry. In January 2024, with Nvidia’s investment, its valuation rose to $520 million. In April 2024, Perplexity launched Enterprise Pro, after which its valuation reached $ billion and it became a unicorn. In June 2024, Softbank invested in Perplexity, then 3x its valuation in just two months, i.e. $ billion. Then by December 2024 it will be worth $9 billion. That means it showed a 177x jump in just one year and thus Perplexity became the fastest growing AI startup. But now the question is whether the method used by Perplexity can be used by a giant like Google. It runs through Google and these models are their reputation. They sell to other businesses through Google’s API. If they themselves do not use these models in their applications such as the Google2 GPT interface, their reputation will be ruined. Why would any other business use their model if they themselves don’t call their model the best? And that’s why when Google2 launched, both these companies were limited in that they had to use their own search engines and AI models. And when Dipsic was launched, both of these companies had to invest time and resources to develop their AI models and then integrate them into search features, which is a very slow process. Perplexity integrated DC r1 op3 mini.in Because Perplexity does not put itself in the category of an AI chatbot but describes itself as an AI search engine which has nothing to do with these AI models.