GPT-4 vs. ChatGPT: how much better is it? | digital trends

GPT-4 is the latest language model for the ChatGPT AI chatbot, and even though it has only just been released, it is already making a splash. The new model is smarter in several interesting ways, most notably its ability to understand images, and it can also process more than eight times as many words as its predecessor. It’s a lot harder to fool now too.

However, you will need to pay to use the new version as it is locked behind the ChatGPT Plus subscription for now.

How are GPT-4 and ChatGPT used?


The easiest way to access ChatGPT is through the official OpenAI ChatGPT website. There’s a lot of interest in it right now, and OpenAI’s servers regularly reach capacity, so you may have to wait for a place to open up to use it, but just update a few times and you should be able to get access.

If you don’t want to wait, you can sign up for a subscription to ChatGPT Plus. That gives you priority access, and you should be able to use ChatGPT whenever you want if you’re a paid member. However, there is a waiting list for new subscribers right now, so you may have to wait a bit anyway.

You will also need to register if you want to use GPT-4. The default free version of ChatGPT currently runs GPT 3.5, a modified version of the GPT3 model that has been in use since 2020. GPT-4 is, for now, a subscriber-only feature, though as it sees further development, it may well be more widely available.

What can GPT-4 do better than ChatGPT?

GPT-4 is a next-generation language model for the AI ​​chatbot, and while OpenAI doesn’t specify what changes it has made to the underlying model, it wants to highlight how much it’s improved over its predecessor. OpenAI claims it can process up to 25,000 words at a time, eight times more than the original GPT-3 model, and can understand much more nuanced instructions, prompts, and questions than GPT-3.5, the model used in the existing model. ChatGPT AI.

OpenAI also assures us that GPT-4 will be much harder to fool, won’t tell falsehoods as often, and is more likely to reject inappropriate requests or queries that could lead to harmful responses.

But GPT-4 also has some exciting new abilities that early adopters are already taking advantage of.

GPT-4 can understand images

GPT-4 is an AI multimodal language model, which means it can understand text as well as other media, such as images. This may sound familiar if you are tried Stable Diffusion AI art generation, but it’s more capable than that, as it can respond to images and queries. This has led to some interesting uses, such as creating a GPT-4 website from a quick sketch… or being able to suggest recipes to a user after analyzing a picture of the ingredients on hand. .

He’s getting a lot better at programming.

ChatGPT has already shown itself to be a capable programmer, but GPT-4 takes it to a new level. Early adopters have managed to turn it into basic games in just a few minutes. Both Snake and Pong were recreated from scratch, despite users having almost no programming experience.

Can pass exams

ChatGPT was good at acting like a human, but it stressed him out and you could often see the cracks and seams. But with GPT-4, that’s much less likely to happen. In fact, it can perform so well in human tests that GPT-4 was able to pass the even bar exam in the 90th percentile of test takers. It also passed the Biology Olympiad test at the 99th percentile. By comparison, ChatGPT could only do so at the 31st percentile.

GPT-4 can create your own demands

The combination of improved reasoning and text comprehension holds a lot of potential for the DoNotPay team. He is working on using GPT-4 to generate “one-click lawsuits,” in which robocallers would be sued if they spammed him. Such a system could also be used to scan medical bills and identify errors, or compare prices with other hospitals to help reduce bills. Then you could even draft a legal defense using the No Surprises Law.

You can understand humor.

GPT-4 is much better for understanding what makes something fun. Not only can he tell better jokes when asked, but if you show him a meme or other funny image and ask him to explain what’s funny about it, he can understand what’s going on and explain it to him.

GPT-4 Limitations

Like ChatGPT before it, GPT-4 isn’t perfect. It certainly is a worthy competitor to Google Bard, but still has a ways to go before it doesn’t make any mistakes and can do almost anything.

As of this writing, GPT-4 is trained on data collected up to August 2022, so it has no knowledge beyond that date. That creates severe limitations on what the AI ​​can do and means that as time goes by, it becomes less accurate due to a lack of the most up-to-date information.

Like its predecessor language models, GPT-4 is also prone to “hallucinations”, in which it claims that inaccurate information is fact. This reportedly happens much less with this model, but it’s not immune, raising concerns about its use in precision-sensitive environments. You also have a rather limited ability to learn from experience, so you can keep making the same mistakes, even when they are pointed out to you.

Currently, GPT-4 is limited to 100 messages every four hours, even for ChatGPT Plus subscribers, and if you’re not already a member, you’ll need to join the waiting list and explain your reasons for wanting to use it.

Editors’ Recommendations

Source link

James D. Brown
James D. Brown
Articles: 9119