OpenAI released the most recent version of ChatGPT the artificial intelligence model for language that is making ripples in the tech industry on Tuesday.
GPT-4, the most recent model, can recognize images as input. That means it can analyze images and provide users general information about the photo.
The language model is also equipped with an extensive information database that allows it to give more precise information as well as write code in the majority of programming languages.
GPT-4 has the ability to be able to read, analyze or create up to 25000 words of text. It seems to be much more intelligent than the previous version. GPT-4 was able to score at the 90th percentile for the Uniform Bar Exam. The previous model was 10th as per OpenAI.
ChatGPT was made available a couple of months ago, is regarded as the fastest-growing application for consumers in the history of. It reached 100 million active users per month within a matter of months. TikTok was able to achieve this number of users, and Instagram took almost three years, as per the findings of a UBS study.
“While human beings aren’t as capable in many real-world situations [GPT-4] shows human-level performance in a variety of academic and professional benchmarks,” OpenAI wrote in its press release, noting that the model’s score was 700/800 in Math SAT.
While amazing, OpenAI acknowledged the program is “far from being flawless.”
“It isn’t perfect it is still in its infancy and appears more impressive at first than it will when you have spent more time with it.” OpenAI CEO Sam Altman tweeted.
Artificial intelligence models, such as ChatGPT have caused some controversy and even created controversial headlines in the last few months. In schools, students have been using these tools to complete their writing assignments, but teachers are divided over whether the systems can be disruptive or could be used to teach tools.
The systems are also at risk of generating incorrect information Google’s AI “Bard,” notably made an error of fact in its initial public demonstration. This is a problem OpenAI hopes to correct and fix. GPT-4 is 40 percent more likely to generate accurate data than its previous version, as per OpenAI.
MORE: What’s ChatGPT the bot with artificial intelligence that became through the internet?
Information that is inaccurate and misleading is a source of concern. AI language models have been trained using large data sets and can include biases in the form of gender, race, religion, and other. This can lead to the AI model creating biased or discriminatory responses.
Many have highlighted the ways that people can make use of misinformation via models like ChatGPT as well as scams that use phishing to distribute fake news to disrupt crucial occasions like elections.
OpenAI affirms that the company “spent months making ChatGPT more secure,” adding the company has partnered together with “over 50 experts to provide early feedback in areas like AI security and safety.”
GPT-4 is 82 percent less likely to provide customers with “disallowed material,” referring to illegal or morally objectionable material, According to OpenAI.