Network Security Internet Technology Development Database Servers Mobile Phone Android Software Apple Software Computer Software News IT Information

In addition to Weibo, there is also WeChat

Please pay attention

WeChat public account

Shulou

GPT-4V multimodal capability is amazing! The screenshot of the formula goes straight to the code, and "Dragon and the Wizarding World" is generated instantly.

2025-02-27 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >

Share

Shulou(Shulou.com)11/24 Report--

After silently updating a number of functions, GPT-4 already had a variety of amazing new abilities. It was simply omnipotent!

Recently, GPT-4 has been quietly updated, adding multi-modal, voice interaction and other features.

However, compared with the new features released every time OpenAI was released at the beginning of the year, the volume of GPT-4 now seems to be much smaller.

I don't know if I'm worried that my product release is too dazzling, leading to constant supervision and lawsuits. In addition to updating the Wensheng graph model DALL-E 3 three weeks ago, OpenAI has not publicly released any new products and functions within seven months after the release of GPT-4.

But Greg Brockman, president of OpenAI, himself tweeted on X (Twitter), constantly retweeting all kinds of unconventional functions realized by the new version of GPT-4.

Take advantage of GPT-4V's multimodal capabilities and coding capabilities to directly translate a mathematical formula written in a book into a piece of Python code.

With the newly updated voice function of GPT-4, some people began to use it as a coach for business negotiations to conduct simulation exercises.

Just below this post, the learning designer at Imperial College Business School commented that they have started designing training courses for MBAs using the voice function of GPT-4.

Direct use of ChatGPT integrated DALLE 3 to generate world views and artwork settings for game designers.

Just a few lines of Prompt and a text description of the Flying Dragon World and an original drawing style diagram will appear.

Use DALLE 3 directly to generate the GIF file you want.

How a corn becomes popcorn.

A dancing dog.

Let's take a look at how to use GPT-4 to complete this series of functions.

GPT-4 users found that almost any mathematical formula can be converted into Python code through GPT-4 as long as it is captured.

Of course, because there was still the possibility of hallucinations in the model, all the results could not be used directly. It was still necessary to carefully check for errors and omissions.

For example, the code in the sixth line in the screenshot,"d_hat (i, j)" should be "d_hat (i)".

Although there are minor errors, netizens still give a very high rating to this feature.

Dr. MIT, founder of AI startup, thinks GPT-4 can't recognize this function without additional context, but it does know what it's doing…pretty cool.

Another developer who developed a financial AI tool called the use case awesome! There's no limit to imagination.

And he gave two specific use cases.

You can take screenshots of complex mathematical equations in research papers and run them quickly locally.

2. You can take screenshots (of anything) and have GPT generate code to implement the UI.

Similarly, in addition to mathematical formulas, it can also directly read molecular formulas and directly output preparation methods.

Feed it a circuit diagram of a headset and it will tell you the rough steps to assemble the device.

GPT-4V's good support for multi-modes, combined with its coding capabilities and extensive knowledge, can be combined into almost unlimited use scenarios.

Another netizen shared the process of creating a fantasy world related to dragons through ChatGPT.

GPT-4 generated concepts, anatomical structures, and even dragon habitats related to dragons.

Close-up of the dragon's head.

Dragon skeleton and solution map.

As well as dragon survival environment original painting and description.

First, you need to specify the image style you want.

The author wanted a technical infographic-style art style, and he used this Prompt, which is almost a plain English description.

「Can you generate me a technical engineer's drawing of a dragon, with labels of its various parts? Use a wide aspect ratio:」

The following results are obtained:

Next, generate a close-up of the faucet.

Then let him generate original drawings and descriptions of the habitat environment.

If you are not satisfied, refine your request further and let GPT-4 meet it.

As a game designer, if he wanted to design a scene related to dragons, he could directly create a usable result.

Another netizen generated an introduction related to saffron inspired by this use case.

「Can you generate me a technical engineer's drawing of a saffron, with labels of its various parts? Use a wide aspect ratio.」

Using this cue word generated a diagram of saffron.

A close-up of saffron bunches was generated.「Can you generate a close up of saffron strand in wide aspect ratio?」

Perspective picture of saffron field.「Please generate an aerial view of saffron field in wide aspect ratio.」

Finally, a profile of saffron was generated.「Anatomy of saffron strand in wide aspect ratio.」

A very complex submarine structure diagram!

Structural diagram of Gundam.

Detailed structural diagram of the head.

Detailed structure of the foot.

A schematic of the weapon.

The super detailed structure diagram of bread machine.

Netizens said they couldn't stop at all.

References:

https://twitter.com/gdb/status/1713301320961036466

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.

Views: 0

*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.

Share To

IT Information

Wechat

© 2024 shulou.com SLNews company. All rights reserved.

12
Report