GPT-4.5
The release of GPT-4.5 has been most awaited tech news of recent times, given the immense progress which Artificial Intelligence has made, since the onset of ChatGPT, which to start with was based on GPT-3.5. GPT technology of OpenAI has been at the core of bringing AI to the mainstream of today’s world. ChatGPT then moved to GPT-4 as its LLM Model. Officially, GPT-4 was released on 14th of March 2023. So, it has been nearly two years, since then the world has been waiting for the next LLM model, announced to be GPT-4.5.
It gained more limelight as because of the delay, there has been talk about GPT research running out of steam, that the LLMs will not be able to add commensurate improvement in performance with the newer GPT model. It was said that during trials, GPT-4.5 was falling short of the company’s expectations. Limitations of scaling and hallucination were some of the problems being talked about. Thus, post GPT-4 models were not likely to be conspicuously better. It was said that the future is about chain of thought and reasoning models and we can find enough evidence of the same.
With all the debates, reservations, debates, and delays, OpenAI has now unveiled GPT-4.5, its most advanced language model till date. Currently, the new model is available as a research preview for ChatGPT Pro users and developers worldwide. Sam Altman, OpenAI CEO and the “Brand Ambassador” says there is a magic to it. As officially stated, the model boasts as expected, of a broader knowledge base. What is also touted is that there is an improved alignment with user intent and heightened emotional intelligence. Given these facts it is expected to be more adept at tasks such as writing assistance, programming and practical problem-solving.
There are admissions and expressions of relevance to this release. Sam Altman makes it clear that this model won’t “crush” benchmarks. He also says that this is the first model where it feels like talking to a thoughtful person. Undeniably, It is a “giant expensive” model. As per OpenAI, this model offers a more natural and intuitive user experience. It is also expected to exhibit fewer instances of “hallucinations.” This model does not introduce any new capabilities compared to the previous models. How much this model has gained out of new supervision techniques is difficult to say. The product is out but the jury is still not out.
GPT-4.5 DOESN’T SEEM TO BE MOMENTOUS IN THE PATHBREAKING GPT’s ARTIFICIAL INTELLIGENCE JOURNEY.
Sanjay Sahay
Have a nice evening.