In the midst of the current ChatGPT boom , his company, OpenAI, has released GPT-4 , a more powerful upgrade to its previous model that can even analyze images and has better reasoning capabilities. OpenAI is so confident in the revolution that GPT-4 entails that it has not hesitated to declare that it ” exceeds ChatGPT in its advanced reasoning capabilities “.
GPT-4 is the latest milestone in OpenAI’s effort to extend deep learning into Artificial Intelligence. GPT-4 is a large multimodal model (accepting image and text input, and emitting text output) that, while less capable than humans in many real-world scenarios, exhibits human-level performance at various points of view. academic and professional reference.
GPT-4 is the new revolution
GPT stands for generative pre-trained transformer, and GPT-4 is part of a series of fundamental language models dating back to the original GPT in 2018 . After the original release, OpenAI announced GPT-2 in 2019 and GPT-3 in 2020. Another refinement called GPT-3.5 arrived in 2022. In November, OpenAI released ChatGPT , which at the time was a refined chat model based on GPT- 3.5.
Over time, OpenAI has increased the size and complexity of each GPT model, resulting in generally better performance, model over model, compared to how a human would complete text in the same scenario, though varies by task. GPT-4 is a great multimodal model capable of processing image and text input and producing text output. It has the potential to be used in a wide range of applications, such as dialog systems, text summarization, and machine translation.
Open AI@OpenAIAnnouncing GPT-4, a large multimodal model, with our best-ever results on capabilities and alignment: https://t.co/TwLFssyALF https://t.co/lYWwPjZbSgMarch 14, 2023 • 18:0058.2K
4.9K
One of the main goals of developing this new version is to improve its ability to understand and generate natural language text , particularly in more complex scenarios. To test its capabilities in such scenarios, GPT-4 was even tested on a variety of tests originally designed for humans. It performs quite well on these assessments, often outperforming the vast majority of human test takers, usually the top 10% of test takers (this is in contrast to GPT-3.5, the version on which ChatGPT was based, which scored at bottom 10%).
GPT-4 is more creative and collaborative than ever. You can generate, edit, and iterate with users on creative and technical writing tasks, such as composing songs, writing screenplays, or learning a user’s writing style. In casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle. The difference arises when the complexity of the task reaches a sufficient threshold: GPT-4 is more reliable, creative, and capable of handling much more concrete instructions than GPT-3.5.
“We spent 6 months making GPT-4 more secure and more reliable. GPT-4 is 82% less likely to respond to requests for disallowed content and 40% more likely to produce trusted responses than GPT-3.5 in our internal evaluations.”
Despite this revolution, limitations are still recognized . “GPT-4 still has many known limitations that we are working to address, such as social bias, hallucinations, and conflicting indications. We encourage and facilitate transparency, user education, and increased AI literacy as society embraces these models. We also intend to broaden the avenues of entry for people to shape our models.”
At the moment, GPT-4 is available in ChatGPT Plus and as an API for developers to build applications and services. Some of the brands with which alliances have already been announced include Duolingo, Be My Eyes, Morgan Stanley or Khan Academy.
What you can do thanks to GPT-4
Greg Brockman, president and co-founder of OpenAI, showed in the GPT-4 presentation some of the uses that this new version of the model can be put to. For example, thanks to image recognition , it allows you to analyze them and start conversations about them. If you upload a photo with what you have in the fridge and ask “what can I do with these ingredients?” it should give you back a full prescription.
Also through Be My Eyes you can describe and understand what is the point of interest in an image, even divided into different boxes or sections. Be My Eyes uses GPT-4 to transform visual accessibility . The difference between GPT-4 and other language and machine learning models, explains Jesper Hvirring Henriksen, CTO of Be My Eyes, is both the ability to have a conversation and the greater degree of analytical prowess that the technology offers.
“Basic image recognition apps only tell you what’s in front of you. They can’t have a discussion to understand if the noodles have the right kind of ingredients or if the object on the floor is not just a ball, but a tripping hazard, and communicate that.”
In one of the demos shown by OpenAI, the ability to even transform simple inputs into code and create a web page from a simple image was shown. An image with a very basic sketch of a web page that the user wanted to create with HTML and JavaScript was displayed, and GPT-4 returned the code needed to create that web page with everything in order as shown in the sketch.
Rowan Cheung@rowancheungI just watched GPT-4 turn a hand-drawn sketch into a functional website.This is insane. https://t.co/P5nSjrk7Wn
March 14, 2023 • 21:4710.7K
690
If you are a fan of the “lots of text” meme, now GPT-4 is a great ally to help you understand and synthesize large strings of text. GPT-4 is capable of handling over 25,000 words of text , enabling use cases such as long-form content creation, extended conversations, and document search and analysis.
When it comes to creative writing , you can achieve some really amazing things, which would cost a human quite a few hours, in a matter of minutes. An example that OpenAI wanted to highlight shows the response to the following request: “explain to me the plot of Cinderella in a sentence where each word has to start with the next letter of the alphabet from A to Z, without repeating any letter. The result (in English) could not be more perfect. «A beautiful Cinderella, dwelling eagerly, finally gains happiness; inspiring jealous kin, love magically nurtures opulent prince; quietly rescues, slipper triumphs, uniting very wondrously, xenial youth zealously».
Linus (●ᴗ●)@LinusEkenstamHere are some incredible things people are already doing with GPT-4It’s been less than 3.5 hours
🧵 A thread
March 14, 2023 • 22:2729.5K
968
The leap in terms of writing code is also considerable. In the Twitter thread that you have above, which includes some of the first GPT-4 applications, we can see how simple games like Snake or Pong have been recreated, providing the code to people with no programming knowledge.
For example, GPT-4 is what’s powering the new Bing , running on this new version that Microsoft has customized for search. If you’ve used the new Bing preview at any time in the last five weeks, you’ve already experienced a previous version of this powerful model. As OpenAI makes updates to GPT-4 and later, Bing will benefit from those improvements along with its own updates based on community feedback. In fact, if you want to experience GPT-4, sign up for the new Bing preview , and once you’re in, you’ll be able to use the new Bing to search, reply, chat, and create.
