Episode transcript The transcript is generated automatically by Podscribe, Sonix, Otter and other electronic transcription services.
Hi everyone, here is Ricardo Vargas, and this is the 5 Minutes Podcast. Today, I would like to talk about the virtual cycle in AI development. And this is based on, and for me, one of the best introductory courses on artificial intelligence is the Andrew Ng AI for everyone that is available on Coursera. If you don't need a certificate, it's it's absolutely free for you to audit. If you want to pay, it's something like 29 or $30 for the course. And, of course, I'm not, okay, I'm not part of Coursera or a business part of Andrew Ng, but it's a great course. In this course, he talks about one thing that really is, for me, one of the key concepts when you are developing any AI project. So, on top of thinking about the project management aspects of this project, what is important is that you understand how you can serve or take this virtual cycle. And I will explain using two, I would say two diagrams, very simple. The first one is a Venn diagram. For those who do not know a Venn diagram, it's two circles that have an intersection between them in the first circle; the circle on the left is what I can do, okay? So everything that is inside this circle, the circle of artificial intelligence, can do and can deliver. The right circle is the value for your business, and you will only find relevant AI projects if there is an intersection between these two. If it's something that I can do, it delivers value for your business.
Many times, I can do a lot of things that are great, cool, and fun but that have nothing to do with your business. So, there is no intersection. So, what you need to understand is where this sweet spot is. And this drives me to the second diagram. The diagram is more data drives a better algorithm that drives a better service. This means the user will have better service when using this AI. That will lead to more people using it. That will lead back to more data. This is the virtual cycle, which means you have more data. Then, you deliver a better algorithm that delivers a better service to your customers, which drives you to more customers using it. And when you have more customers, you have more data. Look, this is your desire, and this must be the strategic aspect of your product. So, every time you are delivering and thinking about an AI product, you need to think through these lenses. How can I get insight into my niche inside this virtual cycle of AI? It's very hard. This is why, for example, AI is dominated by very few companies, like 4 or 5 six companies.
They dominate because data is very expensive, and they have reached what Malcolm Gladwell calls a tipping point. They reach a point that is pretty much impossible for normal people to compete. Because imagine for you to find a 1 trillion token database for you to create a large language model to compete with OpenAI or Gemini. This is this is unthinkable because it's so expensive. It's so hard for you to get that it will take forever for you to do that because these companies have already reached this virtual cycle. They have a gigantic amount of data that drives to hundreds of millions of people using it every single day. So imagine, for example, if I decide to develop, for example, a large language model that will do project negotiations, okay? And let's suppose I, I want to develop my own large language model. Have you thought about how challenging this would be? Imagine me collecting data about negotiations in projects of all sorts to build an algorithm that is useful and will drive people to use it. And imagine that for a general purpose. A large language model like ChatGPT. This is natural. This is natural. You can ask many questions about negotiations and project management, and you will find the answers on ChatGPT Gemini or Cloud. So it's very, very hard.
This is why, for example, most of the companies that are developing this kind of model are using these large language models as a base for their work, and they are doing what is called fine-tuning. So, dear, the project is to create a virtual cycle on top of something that is already inside the virtual cycle. This is exactly what we are doing with the PMOtto. So we migrated to inside, uh, to inside the ChatGPT store to make sure that all the platform benefits from the different versions of ChatGPT. Why? Because it's just impossible to compete and create a large language model that will be valid and will drive usage outside these players. And I'm not saying that I like this from a societal perspective or from a business perspective. I'm just telling you that this is the reality. This is the reality. So, the world of AI is being dominated by 4 or 5 players because they have all reached this virtual cycle. And if you want to reach a virtuous cycle on AI, you need to create something that is absolutely unique, and that can quickly create and generate data through usage and improve extremely fast. And when I say extremely fast, I'm talking about a cycle of two, three, four sprints. We are talking about 60 days. We are talking about 90 days.
If you do not get the traction to get into this virtuous cycle in 60 to 90 days, something like a geometrical or potential growth, it will be very, very hard for you to compete. When I think I'm just using GPT as an example, when I think about ChatGPT 5.0, it's very hard for me to even think of something that could be better than what we have. If they incorporate math, I'm telling this many times, then it will be pretty much able to do everything. So this is exactly why it's so incredible, but at the same time, so hard to develop an AI project from scratch because you will be competing with organizations that are very far ahead in this game. Unless you change the game, it will be very hard, and it will be much more easy for you to incorporate that. They're part of the AI in your project and do fine-tuning on your project to deliver additional benefits. Think always about that, and I strongly suggest if you have not done so yet, let me do a free advertisement for Andrew Ng's course. It's very nice, and you should do it. It's very good and very, uh, worth the course for you to understand AI a little bit better. I hope you enjoyed this podcast, and I will see you next week with another 5 Minutes Podcast.
