In the last few hours OpenAI has announced GPT-4, the fourth version of its famous chat bot based on artificial intelligence that has been surprising everyone for months. This is a giant leap in its deep learning capabilities when it comes to giving accurate and more accurate answers to the questions that are asked. Let’s see, therefore, what secrets this new version holds and why it will change the way we interact on the Internet.
Little by little, AI is changing the way we interact on the internet, but the most important thing is the way it gives us information. If the previous GPT-4 versions have stood out in something, it is that they have been able to give apparently coherent answers, but not entirely correct, and that is why the creation of the new version has become necessary. After all, for many users, AI has become a kind of oracle to ask and any misinformation can have dire consequences not only at the individual level, but also for society.
What’s new with GPT-4 and why is it revolutionary?
We have to start from the fact that we are faced with an application that is based on a learning model, which they have had to redo from scratch. Taking into account the huge amount of information that has to be processed to obtain an inference algorithm with the capacity to generate coherent, intelligent conversations and with reliable information so that a multitude of applications can use GPT-4 and therefore integrate its technology. It is clear that an ordinary computer cannot run it.
It is more because it requires such power capacity, to use it it is necessary to go through the checkout, since we are renting the power of a remote computer. Another different issue will be in large companies and public administrations, where they will have the necessary infrastructure to provide a series of services to the public and also for internal use. So there is still a lot of time for the material democratization of this type of applications and that they can be executed beyond data centers and servers.
What can you do?
- Is able to read entire texts and make summaries in one or a few lines with conditions. This is important when training the AI, as it helps us to know if GPT-4 “understands” the text that we have written to it. It can even help us to know if our writing is understood in the sense that we want to give it and not in a different one.
- You now have the ability to read from different articles and draw ideas from both to give a comparative answer to both.
- Among the capabilities that stand out in GPT-4 is the fact that it can respond in real time, this is ideal for creating customer assistance systems in various types of businesses in an automated manner.
- Another of the functions to highlight is in order to carry out more precise translations of texts, due to grammatical differences, conventional translators give horrendous results. GPT-4 will generate natural texts in the translated language and will even facilitate the internationalization of texts and works.
- Finally, GPT-4 has the ability to do market research through opinion mining to gauge public sentiment about a product or service.
As you can see, there are plenty of reasons to think that it is going to change the world and that it is a revolutionary technology. In any case, we must bear in mind that it is not an expert source and its operation depends on the conclusions of experts in different fields, so GPT-4 should never be seen as a replacement for a professional in any way. ambit
It is not available to users on foot
The processing capacity that GPT-4 requires to function is so high that it is only available to large companies and public administration. What’s more, due to the use it has of NVIDIA’s CUDA libraries, the green brand sees in GPT-4 the opportunity to place its graphics cards that have also become hardware for Deep Learning in recent years. However, they are not the only ones that are going to take advantage of it, since we cannot forget the improvements in this aspect that have been integrated into the company’s CPUs, especially the XMX units and the Sapphire Rapids cores with HBM memory.
