OpenAI secretly creates the 1st general AI equal to humans, according to a Twitter leak

While the eyes of the world are on GPT-4, OpenAI is reportedly secretly planning to complete GPT-5 training in December 2023. He would be convinced that he could achieve AGI: general artificial intelligence, capable of rivaling the human brain…

Just three months after the launch of ChatGPT, based on GPT 3.5, OpenAI launched GPT 4 : the fourth version of its deep learning Generative Pretraining Transformer.

This new, even more powerful iteration is capable of passing law and business school exams with flying colors. It already powers the paid version of the ChatGPT Plus service, as well as the Microsoft Bing search engine.

In addition, GPT-4 brings multimodal input handling allowing the AI to analyze text and images to generate a response. It is also possible to connect ChatGPT to the web via plugins.

However, it turns out that OpenAI already has its eyes on the next step. According to developer Siqi Chen, GPT-5 to complete training by end of 2023 and cause a veritable revolution.

What can the GPT-5 do?

It’s a question that interests us all. GPT-5 is the latest version of the famous GPT (Generative Pre-trained Transformer) AI model. promises even more advanced artificial intelligence that will exceed our expectations. Although it has not yet been officially announced, speculation about GPT-5’s capabilities is already circulating:

One of the most exciting possibilities is the potential integration of AGI (Artificial General Intelligence). This is aa form of AI capable of solving common tasks and problemslike the human brain. If the GPT-5 is indeed equipped with AGI capabilities, this would be a considerable development in the world of artificial intelligence.

In addition, the GPT-5 is also expected to explore new dimensions of development in the field of artificial intelligence. OpenAI, the company behind the GPT, attaches great importance to the development of artificial intelligence.continuous learning and innovation. With each new version of its model, OpenAI pushes the boundaries. It is leading the world towards an exciting future in artificial intelligence.
GPT-5 should also have an even better ability to understand text and to provide contextual responses. We can expect greater precision in the speech generated, resulting in more natural and fluent texts.

GPT-5: the first general-purpose AI to be launched in late 2023

On Twitter, Chen reveals that he has been informed that the end of training is… planned for December 2023. He also added thatOpenAI expects this model to reach “AGI”. general artificial intelligence.

If you’re not familiar with this term, you probably don’t realize the importance of this revelation…

An AGI is an artificial intelligence capable of understanding and learning any task or idea in the same way as a human.. In other words, such an AI will be indistinguishable from a human in terms of capabilities.

The developer’s statements are therefore of the utmost importance. Indeed, an AGI will opent a myriad of new possibilities. It could have an infinitely greater impact than ChatGPT or other tools based on previous versions of GPT.

General AI could massively increase process productivity, speed up human work and eliminate tedious tasks and tedious tasks. It will be able to handle multiple forms of input and output. generate even more accurate results. GPT-5 will also be able to perform more complex tasks.

However, it could also have harmful and unexpected effects that we haven’t even imagined yet. A new study by Goldman Sachs predicts the replacement of 300 million jobs worldwide by the current version of ChatGPT, so what’s next for an AI capable of competing with humans?

AGI, a limit not to be crossed?

No one can really predict all the implications of such a technology. An AGI could, for example, cause the proliferation of extremely credible robots on social networks such as Twitter and Facebook.

A wave of propaganda and misinformation could follow, with Fake News incredibly difficult to detect.

For the time being, OpenAI has officially confirmed the launch of GPT-4.5 in September or October 2023. This will be an intermediate version before GPT-5. Siqi Chen’s predictions are therefore entirely plausible…

On March 29, 2023, Elon Musk and over 1000 AI experts signed an open letter urgently calling for a moratorium of at least six months on the development of AI.

According to the signatories, it is imperative for researchers to pause before creating an artificial intelligence more powerful than GPT-4 to avoid a major threat to humanity and society.

If intelligent machines continue to evolve at the current rate, we simply risk lose control of this creature. The prospect of GPT-5’s launch in late 2023 seems to confirm this grave danger…

OpenAI denies GPT-5 rumors

At an event organized by MIT, OpenAI CEO Sam Altmanwas asked about the open letter signed by Elon Musk and thousands of experts calling for a halt to the development of new AI.

On this occasion, the company director discussed the rumours that OpenAI is currently developing GPT-5:we don’t and won’t for a long time to come.

Nevertheless, he confirms thatOpenAI is preparing other projects based on GPT-4and that the priority will be to solve security problems.

OpenAI still not training GPT-5 in June 2023

In March 2023, Sam Altman confirmed that that OpenAI had not begun training GPT-5 contrary to rumors.

Several months later, in early June 2023, at a conference organized by the Indian newspaper Economic Times, the CEO again declared that the training phase for the successor to GPT-4 has not yet begun.

In his words, ” we still have many of work to do before starting this model “. In particular, he explains that ” we’re working on new ideas that we think we’ll need for this, but we’re certainly not about to start “.

Also during this conference, Altman again insisted that his startup carefully assesses potential dangers with measures such as external audits and safety tests.

As he reminds us, ” when we finished GPT-4, it took us over six months. before we were ready to release it. “. This is one way of responding to criticism and concerns about artificial intelligence.

Furthermore, OpenAI’s CEO declares that he is against regulating smaller AI startups. According to him, ” the only regulation we have requested is for us and the biggest companies “.

This trip to India is part of an extensive to meet with legislators and industry players around the world to demonstrate OpenAI’s willingness to work with regulators.

Il urges legislators in particular to think very seriously to the risks of abuse and other dangers associated with the proliferation of AI, so that safeguards can be put in place to minimize unforeseen accidents…

OpenAI registers the GPT-5 trademark

Although Sam Altman has stated that GPT-5 was not yet in developmentthere are many indications to the contrary.

Recently, the July 18, 2023, OpenAI registered the trademark GPT-5 at USPTO of the United States. The application is currently being processed.

The document reveals several details about the company’s plans. It covers a wide variety of categories, including downloadable software related to language models, the artificial production of human speech and text, and the generation, understanding, analysis and natural processing of human language.

He also mentions software for machine learning-based language processing, translation, dataset sharing, predictive analysis and model building of language.

Also listed conversion of audio data files to textspeech recognition, text creation, and the development of algorithms capable of learning to analyze, classify and act in response to data exposure.

In addition, we learn thatOpenAI plans to offer Software as a Cloud Service (SaaS) for the development and implementation of artificial neural networks…

Accepted by the USPTO, the application now awaits assignment to an examiner. By way of comparison, the application for GPT-4 was filed on March 13, 2023.