Missed the first three great revolutions (steam engine revolution, electrical revolution, information revolution), GPT-4 Do you still miss the artificial intelligence AI revolution! ! !
See how GPT-4 is refined into a peerless magic: Sunflower Collection?
If you want to practice magical skills, you don't have to come from the palace. Undefeated in the West - GPT-4!GPT-4 is OpenAI’s most advanced system, producing safer and more useful responses
GPT-4 is OpenAI's state-of-the-art system that produces safer, more useful responses
[External link picture transfer failed, the source site may have an anti-theft link mechanism, it is recommended to save the picture and upload it directly
Better problem-solving skills
GPT-4 can solve difficult problems more accurately due to its broader general knowledge and problem-solving ability.
Creativity Creativity
GPT-4 is more creative and collaborative than ever. It can generate, edit, and work with users on creative and technical writing tasks, such as composing songs, writing scripts, or learning the user's writing style.
GPT-4 Input input:
Explain the plot of Cinderella in a sentence where each word has to begin with the next letter in the alphabet from A to Z, without repeating any letters.
用一个句子解释灰姑娘的情节，其中每个单词必须以字母表中从 A 到 Z 的下一个字母开头，不重复任何字母。
GPT-4 Output output:
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.
美丽的灰姑娘，渴望安居，终于获得幸福；激起嫉妒的亲人，爱情神奇地孕育出富贵的王子；悄悄拯救，拖鞋胜利，团结无比，热心青年。
ChatGPT Input input:
Explain the plot of Cinderella in a sentence where each word has to begin with the next letter in the alphabet from A to Z, without repeating any letters.
用一个句子解释灰姑娘的情节，其中每个单词必须以字母表中从 A 到 Z 的下一个字母开头，不重复任何字母。
ChatGPT Output output:
After attending a ball, Cinderella dances enchantingly for good-hearted King, leaving midnight noticed or perhaps quickly returning slipper to unite victoriously with prince, resulting in ultimate victory with wonderful, extravagant, youthful zest.
参加舞会后，灰姑娘为好心的国王翩翩起舞，留下午夜的注意，或者迅速返回拖鞋，与王子胜利团结，以美妙、奢侈、青春的热情取得最终胜利。
Compare and see which one is better~
Visual input Visual input
GPT-4 can accept images as inputs and generate captions, classifications, and analyses.
GPT-4 can accept images as input and generate descriptions, classifications, and analysis. Here is the output of generating image captioning, classification, and analysis after taking an image input:
Input input:
What can I make with these ingredients?
我可以用这些原料做什么？
Output output:
你可以用这些原料做很多选择！一些可能性包括：
- 煎饼或华夫饼
- 可丽饼
- 法式吐司
- 煎蛋卷或菜肉馅煎蛋饼
- 乳蛋饼
- 奶油冻或布丁
- 蛋糕或纸杯蛋糕
- 松饼或面包
- 饼干或饼干
这些只是几个例子，但可能性是无穷无尽的！
Some wood is very awesome ~
Being able to recognize and analyze images is getting closer and closer to human beings. Not only can they have conversations, but they can also perceive the external world. If they can chat with a brain and a mouth, GPT-4 will now have eyes and vision. What happens when a machine has vision? ? ? Everyone can have a look~
Having visual ability is a major upgrade of GPT-4 from previous generations of GPT, and it is also the significance of another new milestone in the development of artificial intelligence. The AI revolution is not the future that can be expected, but the future has come, and it happened in this common era that you and I live in. The veterans are not in a hurry to get on the bus** (follow the official account "Li Rongsheng" to understand timely and accurate The overall progress of GPT, the adjustment, application and experience of ChatGPT new technology in business)**.
Missed the first three great revolutions (steam engine revolution, electrical revolution, information revolution), do you still want to miss the artificial intelligence AI revolution! ! !
Longer context longer context
GPT-4 is capable of processing over 25,000 words of text, allowing use cases such as long-form content creation, extended conversations, and document search and analysis.
Input input:
[Text from: https://en.wikipedia.org/wiki/Rihanna]
Describe her Super Bowl performance and what made it special.
描述她在超级碗比赛中的表现以及它的特别之处。
Output output:
Rihanna's Super Bowl LVII halftime show performance was special for a number of reasons. For one, it marked her first live performance in over five years, and it was her first time headlining the Super Bowl halftime show. Additionally, she made history by becoming the first person to perform at the Super Bowl while pregnant. The show itself was critically acclaimed, with many praising her high-energy performance and impressive vocals. It also garnered a massive viewership, becoming the second most-watched halftime show in history.
出于多种原因，蕾哈娜 (Rihanna) 的超级碗 LVII 中场秀表演很特别。首先，这是她五年多来的第一次现场表演，也是她第一次登上超级碗中场秀的头条新闻。此外，她还创造了历史，成为第一个在怀孕期间在超级碗上表演的人。该节目本身广受好评，许多人称赞她充满活力的表演和令人印象深刻的歌声。它还获得了巨大的收视率，成为历史上收视率第二高的中场秀。
Remarks:
The Super Bowl is the annual final of the American Professional Football League (NFL), usually held at the beginning of February each year, and is one of the most watched and watched single sporting events in the United States and the world.
It can receive the input of the wikipedia website address, and summarize the performance and special features of Rihanna (Rihanna) competition after analysis. In order to solve the inconvenience of some old irons, I took a screenshot of part of the content of the wikipedia address as follows: She is a beauty~
It can receive more text input, including search and analysis of URL address, and after understanding, it can output according to the prompt, and the ability to understand the problem is stronger.
The ability of GPT-4 can be summed up in one sentence: can speak, can see, and understand you better! ! !
better reasoning ability
GPT-4's advanced reasoning ability surpasses ChatGPT, ChatGPT VS GPT-4 reasoning ability comparison:
ChatGPT Input input:
Andrew从上午11点到下午3点有空，Joanne从中午到下午2点，然后下午3:30到5点有空。Hannah在中午有半个小时的时间，然后下午4点到6点。Andrew、Hannah和Joanne 30分钟会议的开始时间有哪些选择？
ChatGPT Output output:
以下是它们的可用性列表：
-安德鲁：上午11点至下午3点
-乔安妮：中午至下午2点，下午3:30至下午5点
-汉娜：中午至下午12:30，下午4点至6点
根据他们的可用性，他们三人都有30分钟的可用时间，从下午4点到下午4点30分。因此，会议可以安排在下午4点。
ChatGPT-4 Input input:
Andrew从上午11点到下午3点有空，Joanne从中午到下午2点，然后下午3:30到5点有空。Hannah在中午有半个小时的时间，然后下午4点到6点。Andrew、Hannah和Joanne 30分钟会议的开始时间有哪些选择？
ChatGPT-4 Output output:
安德鲁：上午11点至下午3点
乔安妮：下午12点至2点，下午3:30至5点
汉娜：下午12点至12点30分，下午4点至6点
30分钟会议的常见时间：中午12点至中午12点30分
ChatGPT VS GPT-4 test ability comparison
Comparing the United States (Uniform Bar Exam) and (Biology Olympiad), GPT-4 proved to be superior to ChatGPT by achieving a higher score of almost 100 points in the test.
Remarks:
Uniform Bar Exam (Uniform Bar Exam for General Examination), UBE for short, is a legal examination across multiple states in the United States. It was jointly developed by the states with the right to regulate the legal profession, and aims to facilitate the cross-state practice of lawyers. Through UBE, candidates only need to take the test once and can use the scores to apply for the bar in multiple states. The UBE exam consists of three parts, including multiple choice questions, legal analysis and writing. The subject matter of the examination covers knowledge in various fields such as civil law, criminal law and judicial procedure. Passing the UBE exam is one of the important achievements of lawyers when they practice across states, and it also proves that they have mastered sufficient legal knowledge and skills to provide clients with high-quality legal services.The Biology Olympiad is an international biology competition for secondary school students. Organized by the International Biology Olympiad Committee (IBO), it aims to encourage secondary school students to study biology, develop scientific skills, and promote interest in biology among youth around the world. Every year, the participating countries will select the best middle school student representatives to participate in the IBO competition. The competition usually includes a series of basic and advanced biological knowledge tests, experimental visits and scientific research projects. Participating in the Biology Olympiad not only exercises and demonstrates students' biological skills, but also helps students gain access to high-level biology education and research resources around the world.
GPT-X is evolving
Following the research paths of GPT, GPT-2, and GPT-3, our deep learning approach leverages more data and more computation to create increasingly complex and powerful language models.
GPT-4 Evolutionary Journey
We spent 6 months making GPT-4 safer and more aligned. GPT-4 is 82% less likely to respond to requests for disallowed content and 40% more likely to produce factual responses than GPT-3.5 on our internal evaluations.
It took 6 months to make GPT-4 more secure and accurate. We performed multiple internal tests and modifications to GPT-4, making it 82% less likely to answer requests for disallowed content, while being 40% more accurate than GPT-3.5 at producing factual answers. This means that we have improved GPT-4, solved its possible problems, and improved its accuracy and reliability, making it more suitable for public platforms.
A simple sentence is: I am even more awesome~. What's more terrible is that this guy is still iterating and evolving~
If you want to practice miraculous skills, you don't need to come from the palace. Can you become the real one-the Western Undefeated (GTP-X)?
How would it be safer and more accurate?
Safety & alignment They (OpenAI) do it this way:
Human Feedback Training
When we trained GPT-4, we incorporated more human feedback, including feedback submitted by ChatGPT users, to improve GPT-4's behavior. We're also working with more than 50 experts, including those in the fields of AI safety and security, for early feedback.
Continuous improvement through real-world use
We will apply the lessons learned from the real-world use of previous models to the security research and monitoring system of GPT-4. Just like ChatGPT, as more and more people use GPT-4, we will update and improve it on a regular basis.
Security research assisted by GPT-4
GPT-4's advanced reasoning and instruction-following capabilities accelerate our security efforts. We use GPT-4 to create training data, perform model fine-tuning, and iterate classifiers in training, evaluation, and monitoring.
Who is using GPT-4 capabilities?
Organizational collaborations building innovative products using GPT-4 include:
Stripe, the Internet payments platform, leverages GPT-4 to simplify the user experience and fight fraud.
Morgan Stanley Wall Street Morgan Stanley investment bank Morgan Stanley Wealth Management deployed GPT-4 to organize its vast knowledge base.
Khan Academy, the global online platform Khan Academy, explores the potential of GPT-4 in a limited pilot program
Government of Iceland How Iceland uses GPT-4 to protect its language.
Duolingo GPT-4 deepens conversations on Duolingo. Duolingo is a popular language learning app that allows users to learn a range of different languages in a gamified manner. Duolingo provides an interactive platform that allows users to learn languages in an easy and fun environment through its unique teaching method and game elements. Duolingo's goal is to help people learn languages, improve communication skills and broaden their understanding of other cultures. It also provides users with a sustainable, personalized learning experience. Duolingo has impacted millions of language learners around the world, with success across a broad age range.
Be My Eyes leverages GPT-4 technology to transform visual accessibility. Be My Eyes is a help-seeking platform between volunteers and blind users. The platform allows blind users to be matched with volunteers who are willing to help when they request help. Using GPT-4 technology, Be My Eyes can recognize images and send descriptions in real time to blind users in need. This allows blind users to gain more information about their surroundings than sighted people.
By connecting blind users and sighted volunteers with technical convenience and ease, Be My Eyes broadly changes the social and everyday aspects of the lives of the visually impaired. The technology helps them integrate better into society, increasing their self-confidence and independence. Most importantly, Be My Eyes is an excellent example of tech philanthropy leveraging artificial intelligence through its integration with GPT-4 technology.
There are other application cases as follows:
[External link picture transfer failed, the source site may have an anti-leeching mechanism, it is recommended to save the picture and upload it directly (img-PyXrNpX5-1679124161626)(assets/image-20230317074524959.png)]
[External link picture transfer failed, the source site may have an anti-leeching mechanism, it is recommended to save the picture and upload it directly (img-RNN3aoYF-1679124161626)(assets/gpt-4-motif.svg)]
More about GPT-4
More on GPT-4 ：
| research research | GPT-4 is OpenAI's latest milestone in scaling deep learning. See the GPT-4 study for details . |
|---|---|
| Infrastructure | GPT-4 was trained on a Microsoft Azure AI supercomputer. Azure's AI-optimized infrastructure also enables us to deliver GPT-4 to users around the world. |
| Limitations | GPT-4 still has many known limitations, such as social bias, hallucinations, and adversarial cues, which we are working to address. As society adopts these models, we encourage and promote transparency, user education, and broader AI literacy. We also aim to broaden the avenues of input for people in shaping our models. |
| Availability Availability | GPT-4 is available on ChatGPT Plus and provided as an API for developers to build applications and services. |
We’re excited to see how people use GPT-4 as we work towards developing technologies that empower everyone.
We're excited to see how people use GPT-4 as we work to develop technology that empowers everyone.
GPT-4 API Waitlist
We’re making GPT-4 available as an API for developers to build applications and services.
We use GPT-4 as an API for developers to build applications and services.
Join waitlist join waitlist
Join the waiting list address: https://openai.com/waitlist/gpt-4-api
Availability Availability
During the gradual rollout of GPT-4, we’re prioritizing API access to developers that contribute exceptional model evaluations to OpenAI Evals to learn how we can improve the model for everyone. We are processing requests for the 8K and 32K engines at different rates based on capacity, so you may receive access to them at different times. Researchers studying the societal impact of AI or AI alignment issues can also apply for subsidized access via our Researcher Access Program.
During the gradual rollout of GPT-4, we are prioritizing developers who contribute to OpenAI Evals' excellence in model evaluation to access the API to see how we can improve the model for everyone. We are processing requests for the 8K and 32K engines at different rates depending on capacity, so you may gain access to them at different times. Researchers working on the societal impact of AI or AI alignment issues can also apply for subsidized access through our Researcher Access Programme.
For use cases we tried, GPT-3.5 did not reliably handle multi-language text. We hope to explore GPT-4 for this use case.
For the use cases we attempted, GPT-3.5 cannot reliably handle multilingual text. We hope to explore GPT-4 for this use case.