Days after the release of GPT-4 in March, AI experts rallied against OpenAI to pause giant AI experiments, which basically meant to not train anything beyond the company’s latest model. The experts were probably scared about AI or just wanted to stop the Microsoft-backed AI company midway as it was becoming hard for them to catch up. Interestingly, now the tables are possibly turning against OpenAI.

There is a lot of competition coming up against OpenAI. Even though the founders believe that no one can replicate what the company has built as mentioned during their visit to Israel, a lot of other startups and companies are possibly capable of stealing OpenAI’s customers. One of the most recent examples for this is Anthropic introducing Claude-2.

Read: Claude-2 vs GPT-4

If we compare the capabilities of GPT-4 against Claude-2, in a lot of cases, Anthropic’s offerings outperform OpenAI’s. Twitter is filled with people using the LLMs capabilities of writing code and uploading documents to make sense of it. The large context length of 100K against GPT-4’s 32K, which is also at much lower price also makes it suitable for a lot of companies and their use cases. 

OpenAI wasted enough time already

OpenAI with GPT had given other companies’ a run for their money. Everyone was trying to replicate the success. Seems like OpenAI decided to let them play catch up. 

After the pause letter by the likes of Elon Musk and Steve Wozniak, Altman decided to not train the successor to GPT-4 “for some time,” saying that the company anyway has a lot of work to do before starting to build the model. This was in April. Cut to June, Altman said that OpenAI has not yet started training GPT-5 and is only going to focus on building new ideas. Now, even Musk is here with xAI, his own AI company to compete against OpenAI.

“When we finished GPT-4, it took us more than six months until we were ready to release it,” Altman told Economic Times when he visited India. This means that the company had already started training GPT-4 while they were about to release ChatGPT based on GPT-3.5.

After releasing ChatGPT based on GPT-3.5 in November 2022, it took OpenAI five months more to release GPT-4. It’s been five months now since the release of GPT-4. Ever since then, OpenAI has been introducing incremental updates to their technology, nothing substantial one might say. 

Interestingly, incrementally updating its software is an approach that is shared between the founders Altman and Greg Brockman. At the Lex Fridman podcast, Altman said that even though many people might not agree with the company’s approach, he believes that releasing AI models incrementally makes sense to keep AI under control. 

Similarly in April, Brockman tweeted about his views about the pace of AI progress where he said, “it’s easy to create a continuum of incrementally-better AIs (such as by deploying subsequent checkpoints of a given training run), which presents a safety opportunity very unlike our historical approach of infrequent major model upgrades.”

GPT-4.5 is already here!

There have been claims that the next version of GPT would be superintelligent. In a recent blog about alignment of AI models, OpenAI said that superintelligence might be achieved within four years. The timeline does not match. If the company plans to compete with the rising capabilities of other companies’ AI models, it might ditch the halt on training GPT-5 and actually build it right away. 

Moreover, the only thing close to the jump from GPT-3.5 to GPT-4 in terms of capabilities can be compared with the introduction of Code Interpreter. 

In a podcast on Latent Space, Simon Willison, Alex Volkov, Aravind Srinivas, and Alex Graveley argue that Code Interpreter is actually GPT-4.5. It might be just that the company does not want to use the terminology as of now given all the backlash of the pause letter.

It is possible that OpenAI is currently training GPT-4 with different modalities, similar to what the company did with GPT-3 to GPT-3.5. The introduction of plugins, API, multimodality, all hint towards the training of GPT-5. Moreover, Musk’s xAI is here as well. OpenAI needs to be quick!