OpenAI, the developer of ChatGPT, one of the most popular artificial intelligence tools of recent times, has officially introduced its most advanced language model, GPT-4. Upgrading the features that ChatGPT can offer, GPT-4 will be “more creative and collaborative than ever before,” according to the company’s statement. OpenAI states that GPT-4 is much better than ChatGPT at tackling difficult problems.
GPT-4 is an advanced language model in almost every area from GPT-3.5, which formed the basis of ChatGPT. The company says that the new language model takes valid exams such as the Law School Admission Test (LSAT) that is being applied in the USA, and its scores are much better than GPT-3.5. When we look at the LSAT results, we see that the GPT-3.5 scored 149 and the GPT-4 scored 163. The new artificial intelligence tool has performed better than GPT-4 in all tests it has entered.
Photos can now be uploaded!
ChatGPT’s biggest shortcoming was undoubtedly that it was text-only. In other words, it was not possible to process photos in ChatGPT. GPT-4 will solve this problem, because users will be able to give commands to GPT-4 by uploading photos. Looking at the example above, we see that GPT-4 interprets this photo on request: The unusual thing about this image is a man ironing on an ironing board attached to a moving taxi.
This shows that GPT-4 actually has visual interpretation capability.
It would be unfair if we describe the capabilities of the GPT-4 regarding photos as photo interpretation. I wish you have a dream website but you don’t know coding. You just draw a sketch image of that site on a photo and upload it to the system. GPT-4 will provide you with the source code in HTML, CSS and JavaScript that will enable you to create that site. Of course, you can design the same interface and functions with languages such as React, as it is today.
GPT-4 will also have banned topics like ChatGPT. In the statements made by the company, it is stated that the probability of the new language model to respond to the issues that are not allowed decreased by 82 percent. GPT-4’s probability of producing reality-based text has been increased by 40 percent. From here; Again, we can conclude that GPT-4 is much better than GPT-3.5 at filtering and maintaining its integrity.
GPT-4 has known flaws!
Although GPT-4 is much more advanced than GPT-3.5, it is not much different from ChatGPT in terms of known issues. OpenAI states that this artificial intelligence tool can provide “fake” information. Also, ChatGPT’s known issues, such as violent content and harmful text generation, are largely present in GPT-4.
So how will GPT-4 be used?
OpenAI has made GPT-4 available to everyone. However, consumers who want to use GPT-4 must pay $ 20 per month and become a “ChatGPT Plus” subscriber. If you want to start using GPT-4 right away, you can reach the official page using the link here.