GPT-4.5 news: Everything we know so far about the next-generation language model – Digital Trends
OpenAI’s GPT-4 language model is considered by most to be the most advanced language model used to power modern artificial intelligences (AI). It’s used in the ChatGPT chatbot to great effect, and other AIs in similar ways. But that’s not the end of its development. As with GPT-3.5, a GPT-4.5 language model may well launch before we see a true next-generation GPT-5.
Here’s everything we know about GPT-4.5 so far.
GPT-4.5 is the theorized stop-gap next advance in OpenAI’s language models, following on from the highly successful GPT-4 model currently available to GPT Plus subscribers and in other online AI tools. It’s expected to debut before the release of any next-generation GPT-5 language models, and would likely bring certain advances to what is currently available.
The GPT-3.5 model is widely used in the free version of ChatGPT and a few other online tools, and was capable of much faster response speed and better comprehension than GPT-3, but still falls far short of GPT-4. GPT-4.5 would be a similarly minor step in AI development, compared to the giant leaps seen between full GPT generations.
At the time of writing, GPT-4.5 hasn’t been officially announced, so we don’t know for sure what it will be able to do. However, we can speculate based on past OpenAI development.
GPT-4.5 would likely be built using more data points than GPT-4, which was created with an incredible 1.8 trillion parameters to consider when responding, compared to GPT 3.5’s mere 175 billion parameters. GPT-4.5 would almost certainly factor more parameters and would be trained on more, as well as more up-to-date data.
At the time of writing, GPT-4 is restricted to data preceding the fall of 2021. Any future GPT-4.5 model would likely be based on information at least into 2022, but potentially into 2023. It may also have immediate access to web search and plugins, which we’ve seen gradually introduced to GPT-4 in recent months.
The launch of GPT-4 also added the ability for ChatGPT to recognize images and to respond much more naturally, and with more nuance, to prompts. GPT-4.5 could add new abilities again, perhaps making it capable of analyzing video, or performing some of its plugin functions natively, such as reading PDF documents — or even helping to teach you board game rules.
It’s possible that GPT-4.5 would also be able to remember more information, utilizing past conversations to build upon for its future responses. GPT-4.5 could also be more efficient, requiring less resources to run than existing GPT-4 models, potentially allowing it to run on smaller devices and respond faster.
There’s no timeline for GPT-4.5’s release. Initial estimations and rumors based on the early 2023 launch of GPT-4 targeted GPT-4.5 for a September/October 2023 release, but that seems unlikely now, considering how close we are and the lack of any kind of announcement to that effect.
For now, we have no idea when GPT-4.5 will be available to use, although some have suggested that if OpenAI were to take its plugins, web search, code interpreter, and custom instructions out of beta, that could constitute an upgrade to GPT-4.5 in itself.
Although there was a lot of hype about the potential for GPT-5 when GPT-4 was first released, OpenAI has shot down all talk of GPT-5 and has made it clear that it isn’t actively training any future GPT-5 language model. It claims that much more in-depth safety and security audits need to be completed before any future language models can be developed. CEO Sam Altman has repeatedly said that he expects future GPT models to be incredibly disruptive to the way we live and work, so OpenAI wants to take more time and care with future releases.
However, OpenAI has in more recent times filed a trademark for the name GPT-5, and has launched a web-crawling bot that will harvest information from the internet for training future language models — presumably GPT-5. This could prove problematic though, since OpenAI is being sued by several organizations and individuals for the unlawful use of their work in training GPT language models.
At the very least, GPT-5 appears to be a long way off from release. GPT-4.5 may not have been announced, but it’s much more likely to make an appearance in the near term.
The clumsy use of ChatGPT has landed a New York City law firm with a $5,000 fine.
Having heard so much about OpenAI’s impressive AI-powered chatbot, lawyer Steven Schwartz decided to use it for research, adding ChatGPT-generated case citations to a legal brief handed to a judge earlier this year. But it soon emerged that the cases had been entirely made up by the chatbot.
Microsoft’s Bing Chat is in a much better place than it was when it released in February, but it’s hard to overlook the issues the GPT-4-powered chatbot had when it released. It told us it wanted to be human, after all, and often broke down into unhinged responses. And according to a new report, Microsoft was warned about these types of responses and decided to release Bing Chat anyway.
According to the Wall Street Journal, OpenAI, the company behind ChatGPT and the GPT-4 model powering Bing Chat, warned Microsoft about integrating its early AI model into Bing Chat. Specifically, OpenAI flagged “inaccurate or bizarre” responses, which Microsoft seems to have ignored.
Despite all of the excitement around ChatGPT and similar AI-powered chatbots, the text-based tools still have some serious issues that need to be resolved.
Among them is their tendency to make up stuff and present it as fact when it doesn’t know the answer to an inquiry, a phenomenon that’s come to be known as “hallucinating.” As you can imagine, presenting falsehoods as fact to someone using one of the new wave of powerful chatbots could have serious consequences.
Upgrade your lifestyleDigital Trends helps readers keep tabs on the fast-paced world of tech with all the latest news, fun product reviews, insightful editorials, and one-of-a-kind sneak peeks.
