Good morning, dear Tecnogalaxy readers. Today, we will talk about the new ChatGPT 4.

After months of rumors and speculations, OpenAI has announced GPT-4: the latest AI language model powering applications like ChatGPT and the new Bing. The company claims that the model is “more creative and collaborative than ever” and “can solve challenging problems with greater accuracy.” It can analyze both text and image inputs, although it can only respond through text. OpenAI also warns that the systems retain many of the same issues as previous language models, including a tendency to make up information.

OpenAI states that they have already partnered with several companies to integrate GPT-4 into their products, including Duolingo, Stripe, and Khan Academy. The new model is available to the general public through ChatGPT Plus, OpenAI’s €20 monthly ChatGPT subscription, and it powers Microsoft’s Bing chatbot.

In a blog post, OpenAI mentioned that the distinction between GPT-4 and its predecessor GPT-3.5 is “subtle” in casual conversations. OpenAI’s CEO, Sam Altman, tweeted that GPT-4 is “still imperfect and limited,” but it “looks even more impressive at first use than it does after spending more time with it.”

The company claims that GPT-4’s improvements are evident in the system’s performance on various tests and benchmarks, including the Uniform Bar Exam, LSAT, SAT Math, and SAT Evidence-Based Reading & Writing exams, where GPT-4 scored around 88%.

Speculation about GPT-4 and its capabilities has been circulating over the past year, with many suggesting it would be a significant leap from previous systems. However, judging from OpenAI’s announcement, the improvement is more iterative, as the company had previously cautioned.

The rumor gained further traction last week when a Microsoft executive accidentally revealed in an interview with the press that the system would be launched this week. The executive also suggested that the system would be multimodal, meaning it can generate not only text but other types of content as well. Many artificial intelligence researchers believe that multimodal systems that integrate text, audio, and video offer the best path to creating more capable AI systems.

The research paper describing GPT was published in 2018, with GPT-2 announced in 2019 and GPT-3 in 2020. These models are trained on massive text datasets, primarily sourced from the internet. These models are then used to predict what word comes next. It’s a relatively simple mechanism to describe, but the end result is flexible systems capable of generating, summarizing, and rephrasing text, as well as performing other text-based tasks like translation or code generation.

OpenAI initially delayed the release of its GPT models out of fear that they could be used for harmful purposes like generating spam and disinformation. However, by the end of 2022, the company launched ChatGPT, a conversational chatbot based on GPT-3.5 that anyone could access. The release of ChatGPT triggered a frenzy in the tech world, with Microsoft soon following with its own AI chatbot, Bing.

In the GPT-4 announcement, OpenAI emphasized that the system had undergone six months of safety training and, in internal tests, was “82% less likely to respond to requests for prohibited content compared to GPT 3.5.”

However, this does not mean the system is free from errors or producing harmful content. For instance, Microsoft revealed that the Bing chatbot has always been powered by GPT-4, and many users have been able to bypass Bing’s safeguards in various ways, convincing the bot to offer dangerous advice, threaten users, and fabricate information. Furthermore, GPT-4 still lacks knowledge of events “that have occurred after September 2021.”

Read also:

Was this article helpful to you? Help this site to keep the various expenses with a donation to your liking by clicking on this link. Thank you!

Follow us also on Telegram by clicking on this link to stay updated on the latest articles and news about the site.

If you want to ask questions or talk about technology you can join our Telegram group by clicking on this link.

© - It is forbidden to reproduce the content of this article.