Imagine a future where robots not only understand your emotions but can also generate revolutionary ideas, perhaps even challenging established theories like Einstein’s theory of relativity. This scenario, which seems straight out of a science fiction novel, is becoming increasingly closer thanks to advances in artificial intelligence. Among these, Chat GPT stands out not only for its ability to process and understand human language but also for its potential to go a step further, projecting us into possible future transformations that we are just beginning to comprehend. But how close are we really to this future? And more importantly, are we prepared for it?
Will AI Dominate the World?
Think of a robot capable of identifying your emotional state, or perhaps a robot with the ability to generate new ideas or create scientific theories, or even propose improvements to existing theories, maybe a robot that changes Einstein’s theory of relativity. Currently, research is ongoing on these types of super artificial intelligences, which are supposed to be capable of performing multiple activities, like humans, much faster and more accurately. However, we already have some systems that, due to their capabilities, come close to or even surpass humans in some aspects.
We’ve all heard of Chat GPT, we’ve heard good things, but also bad ones, like Chat GPT taking over the world, or Chat GPT being a scheme created to control the entire planet. But Chat GPT really doesn’t have the computational capacity or sufficiently powerful algorithms to perform multiple tasks, since Chat GPT, like other current artificial intelligences, is focused on a single thing, although in the case of Chat GPT, its capabilities go much further.
Initially, Chat GPT was developed to fulfill the role of an assistant that processes text and understands natural language like English. In short, from the beginning, Chat GPT was conceived as an artificial intelligence model that understands human language and can communicate and respond accurately to generally basic or specific single-topic questions and things, like a customer service chat or something similar.
However, during its training, it became evident that Chat GPT was doing much more than expected, as if Chat GPT had been trained to play soccer but knew how to play all Olympic sports at once.
And this sparked a lot of curiosity within the scientific community, and analysis of these natural language processing-based artificial intelligences began. OpenAI was one of the first non-profit companies, founded specifically to conduct research in the development of artificial intelligence, and one of its largest stakeholders was Elon Musk, and on the other hand, Microsoft, which is currently the main stakeholder.
GPT in Spanish stands for Generative Pre-trained Transformer, which means that Chat GPT is a system of artificial intelligence that is pre-trained to generate new text and, in quotation marks, think and know how to respond to a conversation with a human being.
There is a very important term regarding Chat GPT, and that is the Large Language Model, which in Spanish means modelo de lenguaje grande. In simpler terms, a Large Language Model is an artificial intelligence model trained massively with a huge amount of data, this data in the case of Chat GPT is all the public information on the internet. But it was not only trained with public internet information but also with human-supervised training, specific training, along with other training techniques to minimize certain parameters. And that is how Chat GPT has a lot of knowledge, as it has complete encyclopedias in its brain, all of Wikipedia, articles, news, and much more. Thanks to quality training, artificial intelligence can learn a multitude of things and be very efficient.
But we know that there is a gigantic amount of data on the internet, which is billions of bytes, so to be able to train an artificial intelligence of this type, a very large computational capacity is required, and this means that the costs are extremely high, and it is very expensive to train an artificial intelligence like this.
And that is why Satya Nadella, the CEO of Microsoft, seeing the great advances that OpenAI had with Chat GPT, decided to take a risk in investing billions of dollars in servers and infrastructure, making them available to OpenAI to continue with the development. The version of Chat GPT 3, which was the one that really revolutionized and had the greatest impact in the media, generated a lot of fear among researchers or even Elon Musk himself, who before this, decided to withdraw his shares from OpenAI and propose for many months that a pause be taken in the development of artificial intelligence while it was understood how AI could have greater capabilities than those initially programmed.
And I believe it is important to make a parenthesis here because scientists and researchers in the field of AI did not consider that a model like Chat GPT could have such advanced capabilities. GPTs were only trained to reach a certain level of understanding in language and generating new language, but Chat GPT did much more than that; Chat GPT suggested, created new information, programmed with Python, or with practically all programming languages, and did many other things.
This generated a lot of doubt and controversy to this day because releasing Chat GPT can create risks that are not foreseen in the long term.
After seeing these advances in processing capacity, data analysis, and inference with Chat GPT, and seeing that it could easily understand other languages and create texts, it was thought to release this great discovery to the world, but before that happened, many more things happened.
Big companies like Google and Meta felt somewhat threatened by the emergence of an autonomous competitor superior to any existing system. And that’s why Google began to release much more information and opened forums where advances in AI were shown and the upcoming release of its own artificial intelligence was announced. Before this, one of Google’s most important engineers had resigned, stating that this AI had become self-aware and that Google was not limiting the risks and wanted to release this AI to the entire internet.
And with Chat GPT, we can see the enormous analytical capacity that comes very close to self-awareness and the cognitive level of human beings.
OpenAI finally presented Chat GPT 3 to the public; the reception of this artificial intelligence historically exceeded the number of visits to any emerging website and set a record for the number of downloads during its first month of release, so much so that it was more popular than any other application in the history of modern humanity.
Chat GPT 3 was released, however, OpenAI already had Chat GPT 4 in development; a few months later, Chat GPT 3.5 was released, and a little later Chat GPT 4, which is currently the most powerful artificial intelligence in the world, speaking of capacity and results.
However, in November 2023, something unexpected happened at OpenAI, and its CEO Sam Altman became involved in a series of internal problems, and Sam Altman was fired and lost his position as CEO. This generated a lot of controversy, mainly because everyone wanted to know the reasons why Sam Altman was fired from Open AI. He was being accused of lack of sincerity, which raised suspicions among the Open AI board of directors, assuming that Sam Altman had dangerous individual interests, such as wanting to develop a super AI that could get out of control and dominate the world.
While this was happening, many of OpenAI’s employees resigned, putting the company’s continuity at risk. Ilya Sutskever, one of the highest scientists at OpenAI and co-founder of the company, seeing that the company was falling apart, decided to accept the master move that Microsoft promoted.
And Microsoft offered Sam Altman a job as an independent director of a new project in an AI laboratory. And they also offered vacancies to all OpenAI employees, leaving OpenAI on the brink of collapse.
After this, it was determined that Sam Altman would resume his duties as CEO of OpenAI.
However, without speculating, all these events leave many doubts and many theories because we can imagine the crazy idea that in the OpenAI laboratories, they are really creating a super artificial intelligence or a general artificial intelligence that threatens the persistence of humanity, which is one of the greatest fears that have always existed, shown in movies and science fiction.
In the coming days, an episode of my podcast “Programa tu mente” will be released, where OpenAI and the history of the most influential companies in AI development will be discussed.
Using Chat GPT is very easy because it understands our way of communicating without difficulty; it’s like talking to any other person, only that Chat GPT doesn’t leave you on seen and responds immediately. It also has knowledge of science, history, culture, and almost everything you can find on the internet; Chat GPT knows it. It also works as an excellent personal assistant, so it is a very powerful tool for many applications. When you enter Chat GPT, it’s like using WhatsApp; you write a message, this message is called a PROMPT, and it is the text input that Chat GPT will analyze and determine how to respond.
Getting into technical and computing topics, Chat GPT is developed using TensorFlow, Numpy, and Pandas as the main libraries for Deep Learning, and Python as the main programming language. It also uses various specific algorithms for natural language processing. In its 4.0 version, it has over a trillion parameters, and as for infrastructure, it lives in the Azure cloud, since Microsoft is its main stakeholder.
Clearly, there are many more technical details regarding the distribution and infrastructure of the system, which are too complex to explain in this episode; I invite you not to miss the upcoming episodes and to comment if you want an episode where I talk technically about the operation of Chat GPT.
With Chat GPT, you can learn about a specific topic, summarize, organize your ideas much better, get historical information, perform data analysis, statistical analysis, mathematical calculations, and much more. In addition to this, in its 4.0 version, you can customize the GPTs to perform specific tasks and be experts in doing something specific. You can also send images, reports in multiple formats, and audio messages, and even hear Chat GPT’s responses through its new integrated audio interface, along with many other things.
The capabilities of Chat GPT go far beyond our imagination, and not only does GPT do incredible things, but there are also other artificial intelligences competing for that top spot, and others that are focused on performing slightly more specific but incredibly novel tasks, which I will talk about in the next episodes of the podcast “Programa tu mente.”
