What are the novelties of this brutal version and what would be its dangers with ChatGPT-4

Are we facing the definitive birth of Skynet? That is what many of us wonder now that the boys of Open AI proudly announced the release of their new language module GPT-4. This is the evolution of the core of this Artificial Intelligence system that has revolutionized the community with the implementation of the previous version in ChatGPT.

The power of AI in this type of platform has been more than proven, concluding with certainty that its potential was considerable, although “fortunately” it was not a perfect AI in its execution and frankly it could be disturbing in any chat window.

The experience became more accessible to the public with the implementation of this technology in the Microsoft search engine, Bing and its chat assistant. This led to the proliferation of viral cases and reports where the broad areas of opportunity in the system were verified.

Now with the announcement of the release some obvious questions arise that we will try to answer below.

What is GPT-4, what’s new and why is ChatGPT-4 not yet available?

Under the description of Open AIGPT-4 is a much larger “multimodal (accepting image and text input, and emitting text output)” language model, which, while less capable than humans in many real-world scenarios, exhibits superior performance. on a human level in various academic and professional benchmarks.”

The detail that is multimodal is key, as it brings it closer to systems like DALL-E, without turning into an AI that generates images from text instructions. The key factors to understand GPT-4 are these:

•Being a multimodal AI now it can also understand text, images and videos.

•Your system has been fine-tuned to customize your response style, ranging from formal to handwriting that could be mistaken for any real human being.

•It has a greater capacity for multiple tasks. GPT-4 could streamline and speed up processes for businesses and organizations that rely on AI technology. With such qualities, it could give rise to new employment profiles that help improve the AI ​​data validation process.

In fact, OpenAI claims that in some specific work cases, such as work carried out in any call-center, GPT-4 could save up to 500 hours of work per day in call centers.

•Improved precision. An improvement is designed for implementations of this model in search engines such as Bing. The idea is that GPT-4 can provide more accurate and reliable results, reducing the spread of misinformation.

•Improved security: GPT-4, in theory, is 82% less likely to “respond to requests for disallowed content” and 40% more likely to “produce factual responses than GPT-3.5″.

The bad news is that GPT-4 is already available but its integration, known as ChatGPT-4, is still under development. But developers can have access to API closed betaalthough it is necessary to sign up for a waiting list.

As the image above shows, upon entering the current version of ChatGPT we found that it is still running version 3.5 released in November of last year. But it’s only a matter of time before it arrives.

We have already seen what this system has done in the hands of criminals, and now uncertainty is growing about how they could exploit this development.