GPT Party 3.0. Alexander Zhuravlev: AI-revolution GPT Party 3.0, the largest Russian-language networking event dedicated to artificial intelligence, took place in Silicon Valley on March 9-10. More than 450 people gathered at Plug and Play to meet leading experts, entrepreneurs, and investors, discuss the latest trends in artificial intelligence, and gain practical knowledge. At GPT Party, Alex Zhuravlev, founder of the MentoVC venture fund, spoke on the topic of “AI Revolution.” Alex talked about how artificial intelligence developed and what the current prospects for AI development are, and also shared how to make millions of dollars using AI. Alexander Zhuravlev, Founder of Mento VC Recently, I was particularly impressed by the presentation of the venture fund COATUE, which manages more than $70 billion. The presentation was extensive – 120 slides. I shortened it to 30 and am ready to tell you the main points. I understand that many of you want to make a lot of money, and this is one of the reasons why you go to AI events. There are several ways to do this: invest in funds or companies, launch your own startup in this area, or develop your existing business through revolutionary changes with AI tools. My name is Alex Zhuravlev. I worked as a portfolio director at Altair Capital, invested in many successful companies, made more than 200 deals, and currently study more than 10,000 companies per year. Now I am working on my fund Mento VC, which has already started to show excellent results. My topic is dedicated to the revolution in the field of artificial intelligence. There will be a lot of charts and diagrams, but if you can get into them, then in the end you will have important knowledge about how to act and make money on AI. Dynamics of Artificial Intelligence Development Chart 1 Look at this technology adoption graph, it shows how many years it took for the Internet and smartphones to appear. Artificial intelligence is just starting its path of adaptation, it is only at the level of 10%, but the trend is obvious. Chart 2 Now about what happened in the first half of 2023 with AI, when ChatGPT 3.5 was released in December 2022. We see a significant increase in weekly visits. The second is Wall Street estimates, they assumed that NVDA would earn 7 billion in the second quarter, but they earned 11 billion, in the third – 18, and in the fourth – 21 billion. And the last graph on this slide is about AI Basket, reflecting the number of new projects and applications in the field of artificial intelligence over the past two years. Chart 3 This is also a technology adoption chart, showing that computers took 20 years to implement, the internet took 12 years, and mobile phones took 6 years. As for generative AI, that’s still up in the air, but it seems like it’ll take about 3 years. Chart 4 I really like this chart. It shows how many employees it takes to reach $1 million in revenue in a year, and how we’ve gotten there. We’re talking about S&P 500 companies. It currently requires five employees for every million, but in the future, with AI, it may take less than three. Chart 1. Implementation of technologies Chart 2. Implementation of technologies Chart 3. Implementation of technologies Chart 4. Ratio of the number of employees to revenue $1 million AI Investments Let’s take a look at what’s going on with AI investments. If you’ve been listening to panels, you’ve probably heard people talk about the future and AI: “here in 10 years, there in 10 years.” I’m going to show you the numbers so you can see the stats as well as hear the speakers’ opinions. This is data on how much money private AI companies have raised since 2020. We’ll also look at what these companies were valued at before they launched their product. For example, Anthropic was valued at around $3 billion, Inflection and Adept were valued at $1 billion. This shows how overheated the AI market is and how much money is pouring into the space. Investments in private AI companies Timeframe for the development of artificial intelligence Chart 1 As for the question of whether artificial intelligence is hype or not, in my opinion, no. Now we invest in artificial intelligence mainly in models, which include OpenAI, ChatGPT, Mistral AI and others like that. As for the timeframe. If we talk about driverless cars and transport, autopilots have been developed for more than 15 years. We are seeing results in cars only now, after 15 years, while in the field of artificial intelligence, we have achieved significant results in just five years. And it seems that in the field of artificial intelligence there are already specific applicable results, and we are discussing ways to use them. Chart 2 Another example from history. In the 90s, many telecommunications companies were actively developing, then cloud infrastructure appeared, and now its market exceeds 150 billion dollars. AI is actively developing now, but at present, more than 60% of investments are directed to models. Which of these models will be the winner, we do not know yet. There is no concrete conclusion about whether OpenAI, Anthropic or Mistral AI will be the winner. It is important to understand that we are only at the beginning of the journey, and there is still a lot ahead. Chart 3 This panel makes an interesting comparison of the development of AI with unmanned transport. The first level is cruise control, something similar to the emergence of basic content generation in Midjourney. The second level is Tesla Autopilot, which is comparable to advanced models such as ChatGPT and others. The third level is self-driving with light intervention, in AI this is the GitHub Copilot level. The fourth level is high autonomy, and we have not yet reached this stage in the development of artificial intelligence. It is important to understand that we have reached high autonomy in the field of motor vehicles in general in 15 years, while in the field of artificial intelligence it took us only five years to reach the third level. It seems that the fourth and fifth levels are just around the corner. Chart 1. AI is hype? Chart 2. AI models Chart 3. Comparison of the stages of development of AI and unmanned transport Chart 1 The graph on the left of this slide shows the plans of companies. About 60% of corporations confirm their readiness to implement artificial intelligence in their activities. This is a different reality at the moment. Only 6-9% have fully implemented artificial intelligence, the rest are either at the stage of creating an MVP or at the stage of pilot projects. However, about half of the companies have not yet even reached the POC stage or are at an earlier stage of development. Chart 2 This graph shows the speed of models to achieve human level in benchmarks. We have not yet achieved human level code generation, knowledge of school subjects and understanding of the general meaning in artificial intelligence, but we are very close. It is clear that earlier, starting from the 2000s, this required much more time, but now the curves are almost vertical. Chart 1. Implementation of AI in companies Chart 2. Speed of AI to reach human cognitive level Advantages of Artificial Intelligence As for the advantages, for example, GitHab Copilot already saves developers more than 55% of the time on software development. Runway saves up to 90% of the time of those involved in video editing. In the area of customer support, it was possible to reduce the response time by 45%. Surprisingly, artificial intelligence provides 79% better quality answers to user requests than real therapists. Also in this thread is an example of the time it takes to create marketing materials. In the past, for example in the 50s, it could be months; with the advent of computers – weeks; with the advent of the Internet, cloud computing, etc. – days, with artificial intelligence it is already minutes. This is an illustration of what was said in the panel discussion on marketing. An example of how quickly models can learn is Midjourney, which can now create images on a simple request of significantly better quality than before. It is important to note that this would not be possible without user feedback, which helps improve and speed up the models. The rapid growth of GitHab Copilot since January 2022 is another example. It has already accumulated 10 million downloads. The acceptance rate of corrections suggested by artificial intelligence is constantly growing. Using a specific fintech company as an example, we can see that user service costs were reduced by 95%, response time was reduced from 45 minutes to one minute, and customer satisfaction levels increased by 14%. Prospects for the development of artificial intelligence Currently, we interact with artificial intelligence mainly at the application level. We currently have assistants in different areas: for example, GitHab Copilot and ChatGPT, and in content generation – Midjourney and Runway. The most interesting thing is what we expect from the development of artificial intelligence in the future. We assume that artificial intelligence will specialize in specific skills, such as scientific research, medical consultations, etc. We are moving towards personalized artificial intelligence – AI scientist, AI doctor, and so on. In addition, in the content sphere, we can expect the emergence of AI influencers, AI artists and AI marketers. It seems that all this is already close to reality. When it comes to copy-palettes, the example of AI chess shows that progress in this area has been going on for a long time. From 1995 to 2005, many believed that it was only possible to beat a human using a combination of human and AI skills, but in 2015, a machine beat a human at chess, showing that AI can achieve a high level of mastery in the game. This example shows that AI is capable of many things, and the level we see now in the field of copy-palettes will inevitably increase. When mobile technology and phones first appeared, it led to new markets worth hundreds of billions of dollars. In essence, it was a new era. And the same is happening with AI now. There are still many questions about what the future of AI will hold. Will we have autonomous systems and infinite game worlds that can be constantly expanded? Will there be models that can think? There is huge potential in various industries: for example, software creation creates a market worth hundreds of billions of dollars, audio – tens of billions, video analysis – hundreds of billions, and many more, including the field of robots. These markets represent new opportunities, although there are challenges that must be overcome. Conclusions The previous panel discussed the importance of learning programming. Indeed, there is an opinion that in the future, English may become the main programming language, and it will be enough to write code. Regarding language models and networks, there are rumors that Apple’s combination of Siri, Apple Shortcut and Ajax GPT could give billions of users their own built-in artificial intelligence in their phones. Another significant area is the so-called private data, since large companies have a huge amount of data. If we introduce artificial intelligence into this data, it will lead to incredible opportunities. For example, in Johnson & Johnson biotechnology, we will be able to receive personalized medical services created specifically for us. In Netflix, we will be able to receive shows and movies created individually for our preferences. Autodesk can use its data to create any objects and details. In the field of games, we have already mentioned the creation of infinite worlds. In the field of shopping, artificial intelligence will suggest products that are relevant to us. If we compare the AI industry to self-driving cars, we can say that we are at a level where cars have autopilot, which prevents lane departures and warns the driver by vibrating the steering wheel. The next levels of AI development are already available and attracting large investments. You just need to remember the beginning of my speech, where we talked about millions and billions of dollars, and think about where you fit in this niche.
