OpenAI's Next Move: Launching GPT-5?

2023-07-17

"Within days of the release of GPT-4 in March, AI experts united to oppose OpenAI's continuation of massive AI experiments, urging them not to train anything beyond GPT-4. These experts may either fear AI or simply want to hinder Microsoft-backed AI companies, as they struggle to catch up with OpenAI. OpenAI faces numerous competitors. Although the founders believe that no one can replicate what they have built, as they mentioned during their visit to Israel, many companies now have the ability to snatch OpenAI's clients. For instance, Anthropic recently launched Claude-2. If we compare the capabilities of GPT-4 and Claude-2, in many cases, Anthropic's product outperforms OpenAI's. Twitter is filled with people showcasing the ability of LLM to write code and understand the meaning of uploaded documents. With a context length of 100K, lower pricing, and compared to GPT-4's 32K, Claude-2 offers better value for some customers. OpenAI has wasted too much time. With GPT, they left other companies in the dust. Everyone is trying to replicate their success. But it seems like OpenAI wants them to catch up a bit. After Elon Musk and Steve Wozniak issued a pause letter, Altman decided to "temporarily" not train the successor to GPT-4, stating that the company has a lot of work to do before starting to build the model. Fast forward to June, Altman said that OpenAI has not yet started training GPT-5 and is instead focusing on building new ideas. Now, even Musk himself is entering the competition with his own AI company, xAI. "When we finished GPT-4, it took us over six months to get ready to release it," Altman told the Economic Times during his visit to India. This means that the company started training GPT-4 even before releasing GPT-3.5. After releasing GPT-3.5 in November 2022, it took OpenAI another five months to release GPT-4. Since then, OpenAI has been progressively updating its technology without any substantial advancements. Interestingly, the progressive update of their software is a shared approach by founders Altman and Greg Brockman. In Lex Fridman's podcast, Altman mentioned that although many may disagree with the company's approach, he believes that progressively releasing AI models helps control AI. GPT-4.5 is here! Some claim that the next version of GPT will be superintelligent. In a recent blog post about AI model adjustments, OpenAI stated that superintelligence could be achieved within four years. This timeline does not align. If the company plans to compete with other companies' increasingly powerful AI models, they might start building GPT-5 immediately. Furthermore, the only comparable leap in capabilities from GPT-3.5 to GPT-4 is the introduction of a code interpreter. In the Latent Space podcast, Simon Wilson, Alex Wolf, Aravind Srinivas, and Alex Graves suggest that the code interpreter is essentially GPT-4.5. Considering the current opposition in society, OpenAI may simply not want to use this term. OpenAI is likely training GPT-4 in a different way, similar to what they did from GPT-3 to GPT-3.5. Plugins, APIs, multimodality, all hint at the training of GPT-5. Additionally, Musk's xAI has entered the scene, and OpenAI needs to act fast. Source: https://analyticsindiamag.com/its-time-openai-launched-gpt-5/"

RECENT AI NEWS

RECENT AI TOOLS