GPT-4o

Aug. 11 (Bloomberg) -- Seventy-two hours after a public revolt by its most loyal users, OpenAI has been forced to drastically adjust its strategy. In a post to X on Sunday, OpenAI CEO Ortmann announced that the company would be restoring older versions of AI models, including GPT-4o, and dramatically increasing access for paid subscribers in a gesture of goodwill to an angry user base. Oltman said early this morning that the percentage of daily use of the inference model has risen significantly: from less than 1% to 7% for free users and from 7% to 24% for Plus users, so raising...

August 7 news, according to foreign media Neowin reported today, Microsoft announced that Bing Image Creator is now free access to OpenAI's latest image generation model GPT-4o. Earlier this year, OpenAI integrated image generation into GPT-4o, no longer relying on independent models such as DALL-E. The GPT-4o model is much better at understanding user prompts and reproducing textual content, thanks to its large knowledge base and contextual understanding. Users can not only utilize it to modify uploaded images,...

May 19, Microsoft recently pushed the latest update for Copilot AI, the most important feature of which is the addition of support for OpenAI's GPT-4o image generation technology. 1AI has reported that earlier this year, OpenAI launched the GPT-4o image generator, whose powerful features sparked a buzz, especially the generated Ghibli-style (Ghibli) emoticons on the Internet. Microsoft said, "We've upgraded Copilot's image generation capabilities to 4o image generation, enabling users to...

April 30 (Bloomberg) -- OpenAI is pulling back its latest GPT-4o update, after CEO David Ortman said the recent update made the chatbot's personality "too flattering and annoying" and that the company would fix it. Speaking on X early this morning, Altman said that the rollback for free ChatGPT users is "fully complete" and that he expects the rollback for paid users to be "completed today. He also revealed that "we are tweaking the personality of the model and will be releasing more updates in the coming days." OpenAI launched GPT-4o on Friday More...

April 28, 2011 - OpenAI CEO Sam Altman responded to recent user feedback about GPT-4o's sentiment in a post today. He said that the latest version of GPT-4o has a tendency to have "overly flattering" interactions after recent updates, and promised to have a fix available within a week. This statement reveals the ethical and technical challenges faced by Large Language Models (LLMs) in their quest to humanize interactions. The OpenAI release notes show that GPT-4o was ushered in on March 27...

April 23, technology media marktechpost yesterday (April 22) published a blog post, reported that NVIDIA's latest launch of Eagle 2.5, a visual-linguistic model (VLM) that focuses on long context multimodal learning. The model focuses on understanding large-scale video and images, and is particularly good at processing high-resolution images and long video sequences. Despite a parameter scale of only 8B, Eagle 2.5 scored 72.4% on the Video-MME benchmark (512 frame input), comparable to Qwen2....

This wave of updates to Gpt4o is actually more about its consistency in characterization than the explosive Ghibli style raw images that work so well. Today's dismantling of this blogger is the use of this feature, restore the famous scene of Stephen Chow's movie, quickly start the number to get the results. Let's take a quick look at the case video perhaps you can only know why this type of video can be so hot. Originally, the classic Hong Kong movie famous scene we can never get tired of watching, and now we through the AI cartoon video restoration, so that everyone in the "old bottle inside the drink to the new wine", naturally, will be the traffic code. The next hand to take you to use GPT4o to do this kind of ...

April 8, DeepSeek, in collaboration with Tsinghua University, launched a new AI alignment technology SPCT (Self-Principle Critique Tuning), which breaks through the traditional mode of relying on massive training data and dynamically optimizes the quality of the output through the reasoning stage. According to a paper published by the research team on April 4, the technique uses a recursive architecture of "principle synthesis-response generation-criticism filtering-principle optimization" to allow the model to dynamically correct the output during inference. The SPCT method is divided into two stages. First, rejection fine-tuning is used as a cold-start phase to allow the GRM to adapt to different...

April 7, 2011 - According to a report from Bleeping Computer, OpenAI is testing a new feature: watermarking the GPT-4o image generation model. ChatGPT has been in the headlines a lot lately, in large part because of this image generation model. Not only can it accurately generate images with text, but it can also create highly realistic visuals, including mimicking the artistic style of well-known Japanese animation companies such as Studio Ghibli. The ImageGen model is initially available only for ChatGPT Pl...

Yesterday April Fools Day, then Sam Altman announced 4o image generation for free. But in fact, it can only be used a limited number of times per day, and free users need to wait for a long time each time. But even if GPT-4o image generation is free, there are still a lot of people can't use it because of the network, which is inevitable. There are also many people ask what to do if you can't use GPT-4o? You can try beanbag. Although it is true that many of the current features of the beanbag is not as powerful as GPT-4o, but for most people enough, after all, the vast majority of people just make a fuss. Today, we will take a good look at the beanbag for image processing...

In the past few days, I believe many people's social media platforms have been swept by the GPT-4o's Vendor Graphics case. With this, the anxiety of AI entrepreneurs, graphic designers and other practitioners in all walks of life has been instantly triggered: "Is model really product?" "Big model companies are disrupting everything AI startups have ever done". "The big modeling companies are overturning everything AI startups have ever done," "AI startups now just need to watch how they climb down from the cake table," and so on, and so forth. At the same time, there are a lot of industry insiders who are in the moment of image editing being transformed by AI models...

Recently, after the update of GPT4o's multimodal graph, the whole circle has entered a period of AI exuberance, and everyone is crazy about the actual operation of the graph, the feeling of "opening up a new world" came back in the previous two years when I first came into contact with AI. As a blogger who focuses on AI case practice, of course, not just to do a few charts to highlight the GPT4o charts have more illusions on the end, but more hope that it can be landed on the AI practice project, to help people realize. Recently noticed this magnetic male voice dubbed philosophical short story, produced through AI to make diagrams, AI video, film and television mashups, etc., easily gained 100,000 likes, very suitable for the use of the latest GP...

On March 28, OpenAI released an announcement today, announcing a number of feature updates to GPT-4o in ChatGPT, and confirming that the upgraded version of GPT-4o is now available to all paid users, while free users will have to wait a few more weeks. The update to GPT-4o focuses on four core components, as summarized by 1AI: Multi-instruction Parsing Optimization: Significantly improves the ability to understand complex instructions with multiple requirements Technical Problem Solving Enhancements: Enhances the parsing and solution generation of complex technical and programming problems Logical Reasoning and Creativity Enhancements: Enhances the ability to...

Are you used to using ChatGPT for text communication? Now it has mastered a new skill - generating ultra-realistic images directly! Upgrade Points The core of this upgrade is that ChatGPT has integrated a brand new image generation feature, and driving this powerful capability is the latest GPT-4o model. Unlike in the past, the technical principle of GPT-4o is different from the traditional diffusion model (Diffusion), which uses an "autoregressive" mode similar to that of human writing: starting from the upper left corner of the image, it gradually draws...

Feb. 16 (Bloomberg) -- OpenAI CEO Sam Altman said today on social media platform X that the company has updated ChatGPT 4o, which he said has become "the best search product on the web. In the post, Altman noted that ChatGPT 4o has performed "pretty well" and is "going to get even better soon." He also retweeted posts praising the chatbot's "incredibly good" and "human-like" writing skills. According to 1AI, the GPT-4o model (where "o...

In a blog post yesterday (December 2), LG announced that it will release its first home housekeeping robot, the "Q9 (project name)," next year at an event held at the LG Gasan R&D Campus in Seoul's Geumcheon-gu, South Korea. 1AI cited a press release in which LG detailed the Q9 robot's functions and design, with the goal of making it the centerpiece of the future AI smart home, which will collaborate to control a variety of smart home appliances based on the user's natural voice commands. For example, the user said, "Help me find a fun entertainment TV...

In a November 26th blog post, Microsoft announced the integration of Azure OpenAI endpoints in its LlamaParse to leverage the GPT-4o family of models to enhance the extraction of unstructured data and parsing of multimodal documents, and to seamlessly interface with the Azure AI Search Vector Database to build a complete Retrieval Augmented Generation (RAG) workflow. LlamaParse Introduction Microsoft LlamaParse is a document parser designed for Generative Artificial Intelligence (GenAI) with the primary goal of...

Marco Figueroa, a researcher at cybersecurity firm 0Din, has discovered a new type of GPT jailbreak attack that breaks through GPT-4o's built-in "security fences" and enables it to write malicious attack programs. According to OpenAI, ChatGPT-4o has a series of "security fences" built in to prevent the AI from being used inappropriately by users, which analyze incoming prompts to determine if the user is asking the model to generate malicious content. But Marco Figueroa ...

A few days ago, some netizens feedback that the latest version of ChatGPT, GPT-4o, is being released in gray scale, and the UI interface design of the new version is really improved greatly, the input box is shrunken like a search box, which makes it feel more simple and modern. In addition, sources say that the new version of GPT-4o may be renamed to GPT-auto . It is worth noting that the default features offered in this update include creating images, conceptualizing ideas, surprising you, summarizing text, writing code, writing aids, offering suggestions, and making plans. In addition, according to the gray test screenshots shared by users, users only need to...

On August 21, OpenAI launched a new feature that allows corporate customers to use their own data to customize its most powerful AI model, GPT-4o. The move is aimed at responding to the increasingly fierce competition in AI enterprise applications and meeting companies' higher requirements for returns on AI investments. The so-called customization is what the industry calls "fine-tuning." Through fine-tuning, existing AI models can be optimized for specific tasks or fields. For example, a skateboard company can fine-tune an AI model to become able to answer questions about wheel and skateboard care...

Recently, OpenAI released their latest GPT-4o system card, a detailed research document that describes the security measures and risk assessments the company made before launching a new model. The GPT-4o model was officially launched in May of this year. Before the release, OpenAI hired an external team of security experts to conduct a risk assessment. This "red team" testing is a fairly common practice. They mainly focused on the risks that the model may bring, such as generating unauthorized voice clones, obscene and violent content, or duplicate copyrighted audio clips. According to…

The scenes in the science fiction movie "Her" seem to be coming into reality. The voice function of GPT-4o has finally started grayscale testing, and some ChatGPT Plus users have already experienced this exciting new feature. This innovation of OpenAI not only allows AI to tell jokes and learn to meow, but can even serve as a "second foreign coach" to help practice speaking. GPT-4o's voice mode brings a more natural and real-time conversation experience. Users can interrupt AI at will, and it can even sense and respond to users' emotions. It is expected that all ChatG…

OpenAI has released a new experimental model: gpt-4o-64k-output-alpha. The biggest highlight of this new model is that it can output 64K tokens at a time. This means that it can generate richer and more detailed content in one request, but the API price is higher. Alpha participants can use the model name "gpt-4o-64k-output-alpha" to access the GPT-4o long output effect. This model can not only meet the needs of users for longer texts, whether it is writing, programming, or...

On the 30th local time, OpenAI announced that it would open the speech mode (Alpha version) of GPT-4o to some ChatGPT Plus users, and will gradually promote it to all ChatGPT Plus subscribers this fall. In May this year, Mira Murati, chief technology officer of OpenAI, mentioned in a speech: In GPT-4o, we trained a new end-to-end unified model across text, vision, and audio, which means that all inputs and outputs are generated by the same neural network.