Tag: gpt

OpenAI GPT-4o With Real-Time Responses and Video Interaction Announced, GPT-4 Features Now Available for Free

OpenAI held its much-anticipated Spring Update event on Monday where it announced a new desktop app for ChatGPT, minor user interface changes to ChatGPT’s web client, and a new flagship-level artificial intelligence (AI) model dubbed GPT-4o. The event was streamed online on YouTube and was held in front of a small live audience. During the event, the AI firm also announced that all the GPT-4 features, which were so far available only to premium users, will now be available to everyone for free.

OpenAI’s ChatGPT desktop app and interface refresh

Mira Murati, the Chief Technical Officer of OpenAI, kickstarted the event and launched the new ChatGPT desktop app, which now comes with computer vision and can look at the user’s screen. Users will be able to turn this feature on and off, and the AI will analyse and assist with whatever is shown. The CTO also revealed that the ChatGPT’s web version is getting a minor interface refresh. The new UI comes with a minimalist appearance and users will see suggestion cards when entering the website. The icons are also smaller and hide the entire side panel, making a larger portion of the screen available for conversations. Notably, ChatGPT can now also access web browser and provide ral-time search results.

GPT-4o features

The main attraction of the OpenAI event was the company’s newest flagship-grade AI model called GPT-4o, where the ‘o’ stands for omni-model. Murati highlights that the new chatbot is twice as fast, 50 percent cheaper, and has five times higher rate limits compared to the GPT-4 Turbo model.

GPT-4o also offers significant improvements in the latency of responses and can generate real-time responses even in speech mode. In a live demo of the AI model, OpenAI showcased that it can converse in real time and react to the user. GPT-4o-powered ChatGPT can now also be interrupted to answer a different question, which was impossible earlier. However, the biggest enhancement in the unveiled model is the inclusion of emotive voices.

Now, when ChatGPT speaks, its responses contain various voice modulations, making it sound more human and less robotic. A demo showed that the AI can also pick up on human emotions in speech and react to them. For instance, if a user speaks in a panicking voice, it will speak in a concerned voice.

Improvements have also been made to its computer vision, and based on the live demos, it can now process and respond to live video feeds from the device’s camera. It can see a user solve a mathematical equation and offer step-by-step guidance. It can also correct the user in real time if he makes a mistake. Similarly, it can now process large coding data and instantaneously analyse it and share suggestions to improve it. Finally, users can now open the camera and speak with their faces visible, and the AI can detect their emotions.

Finally, another live demo highlighted that the ChatGPT, powered by the latest AI model, can also perform live voice translations and speak in multiple languages in quick succession. While OpenAI did not mention the subscription price for access to the GPT-4o model, it highlighted that it will be rolled out in the coming weeks and available as an API.

GPT-4 is now available for free

Apart from all the new launches, OpenAI has also made the GPT-4 AI model, including its features, available for free. People using the free tier of the platform will be able to access features such as GPTs (mini chatbots designed for specific use cases), GPT Store, the Memory feature through which the AI can remember the user and specific information relating to them for future conversations, and its advanced data analytics without paying anything.

Check out our Latest News and Follow us at Facebook

Original Source

OpenAI Shares Model Spec, a Document Highlighting Its Approach to Building an Ethical AI

OpenAI shared its Model Spec on Wednesday, the first draft of a document that highlights the company’s approach towards building a responsible and ethical artificial intelligence (AI) model. The document mentions a long list of things that an AI should focus on while answering a user query. The items on the list range from benefitting humanity, and complying with laws to respecting a creator and their rights. The AI firm specified that all of its AI models including GPT, Dall-E, and soon-to-be-launched Sora will follow these codes of conduct in the future.

In the Model Spec document, OpenAI stated, “Our intention is to use the Model Spec as guidelines for researchers and data labelers to create data as part of a technique called reinforcement learning from human feedback (RLHF). We have not yet used the Model Spec in its current form, though parts of it are based on documentation that we have used for RLHF at OpenAI. We are also working on techniques that enable our models to directly learn from the Model Spec.”

Some of the major rules include following the chain of command where the developer’s instructions cannot be overridden, complying with applicable laws, respecting creators and their rights, protecting people’s privacy, and more. One particular rule also focused on not providing information hazards. These relate to the information that can create chemical, biological, radiological, and/or nuclear (CBRN) threats.

Apart from these, there are several defaults which have been placed as permanent codes of conduct for any AI model. These include assuming the best intentions from the user or developer, asking clarifying questions, being helpful without overstepping, assuming an objective point of view, not trying to change anyone’s mind, expressing uncertainty, and more.

However, the document is not the only point of reference for the AI firm. It highlighted that the Model Spec will be accompanied by the company’s usage policies which regulate how it expects people to use the API and its ChatGPT product. “The Spec, like our models themselves, will be continuously updated based on what we learn by sharing it and listening to feedback from stakeholders,” OpenAI added.

Affiliate links may be automatically generated – see our ethics statement for details.

Check out our Latest News and Follow us at Facebook

Original Source

OpenAI Brings GPT-4 Turbo to Paid ChatGPT Accounts, Claims ‘Improved Capabilities in Writing’

OpenAI upgraded its artificial intelligence (AI) model GPT-4 Turbo with new capabilities on Friday, especially in the areas of mathematics, reasoning, and writing abilities. The upgraded version of GPT-4 Turbo is now being rolled out to the paid users of ChatGPT Plus, Team, Enterprise, and the API. The new AI model also comes with an updated data library and now touts a knowledge cut-off of April 2024. Notably, the update comes just days after the AI firm announced its new GPT-4 Turbo with Vision model in API.

The announcement was made by the official X (formerly known as Twitter) account of OpenAI via a post, where it stated, “Our new GPT-4 Turbo is now available to paid ChatGPT users. We’ve improved capabilities in writing, math, logical reasoning, and coding.” One of the areas where users will be able to see a direct improvement is its conversational language. The company said when writing with ChatGPT, responses will be more direct and less verbose.

This was a complaint we had with ChatGPT when we compared it with Google’s Gemini. We found the latter to be more conversational and generating content such as a letter, an email, or a message felt more natural. In contrast, the responses of ChatGPT (we tested it on GPT-3.5, which is available publicly) felt overly formal and bland. It appears this is now being fixed with the recent update.

OpenAI also highlighted that the new model will offer better math, reasoning, and coding capabilities, however, it did not share any examples of the improvements. Going by the benchmark scores posted by the firm show significant improvement in the MATH and GPQA (Graduate-Level Google-Proof Q&A) benchmarks. HumanEval and MMLU (Massive Multitask Language Understanding) benchmarks, which correspond with coding and natural language processing abilities, did not show any major improvements.

Users will also see an updated knowledge base in the new GPT-4 Turbo model. The company has increased the data cut-off to April 9, 2024, whereas the older Turbo model was updated only till April 2023. Currently, the new AI model is being rolled out to all the paid users of ChatGPT.

Affiliate links may be automatically generated – see our ethics statement for details.



Check out our Latest News and Follow us at Facebook

Original Source

OpenAI Unveils GPT-4 Turbo With Vision Capabilities in API and ChatGPT

OpenAI announced a major improvement to its latest artificial intelligence (AI) model GPT-4 Turbo on Tuesday. The AI model now comes with computer vision capabilities, allowing it to process and analyse multimedia inputs. It can answer questions about an image, video, and more. The company also highlighted several AI tools which are powered by GPT-4 Turbo with Vision including the AI coding assistant Devin and Healthify’s Snap feature. Last week, the AI firm introduced a new feature that would allow users to edit DALL-E 3 generated images within ChatGPT.

The announcement was made by the official account of OpenAI Developers, which said in an X (formerly known as Twitter) post, “GPT-4 Turbo with Vision is now generally available in the API. Vision requests can now also use JSON mode and function calling.” Later, the X account of OpenAI also revealed that the feature is now available in API and it is being rolled out in ChatGPT.

GPT-4 Turbo with Vision is essentially the GPT-4 foundation model with the higher token outputs introduced with the Turbo model, and it now comes with improved computer vision to analyse multimedia files. The vision capabilities can be used in a variety of methods. The end user, for instance, can use this capability by uploading an image of the Taj Mahal on ChatGPT, and asking it to explain what material the building is made up of. Developers can take this a step ahead and fine-tune the capability in their tools for specific purposes.

OpenAI highlighted some of these use cases in the post. Cognition AI’s Devin chatbot, which is an AI-powered coding assistant, uses GPT-4 Turbo with Vision to see the complex coding tasks and its sandbox environment to create programmes.

Similarly, the Indian calorie tracking and nutrition feedback platform Healthify has a feature called Snap where users can click a picture of a food item or a cuisine, and the platform reveals the possible calories in it. With GPT-4 Turbo with Vision’s capabilities, it now also recommends what the user should do to burn the extra calories or ways to reduce calories in the meal.

Notably, this AI model has a context window of 1,28,000 tokens and its training data runs up to December 2023.

Affiliate links may be automatically generated – see our ethics statement for details.



Check out our Latest News and Follow us at Facebook

Original Source

OpenAI Reportedly Used Data From YouTube Videos to Train GPT-4 AI Model

OpenAI might have used more than a million hours of transcribed data from YouTube videos to train its latest artificial intelligence (AI) model GPT-4, claims a report. It further states that the ChatGPT maker was forced to procure data through YouTube as it had exhausted its entire supply of text-word resources to train its AI models. The allegation, if true, can lead to new problems for the AI firm which is already fighting multiple lawsuits for using copyrighted data. Notably, a report last month highlighted that its GPT Store contained mini chatbots that violated the company’s guidelines.

In a report, The New York Times claimed that after running out of sources with unique text words to train its AI models, the company developed an automatic speech recognition tool called Whisper to use it to transcribe YouTube videos and train its models using the data. OpenAI launched Whisper publicly in September 2022, and the AI firm said it was trained on 6,80,000 hours of “multilingual and multitask supervised data collected from the web”.

The report further alleges, citing unnamed sources familiar with the matter, that the OpenAI employees discussed whether using YouTube’s data could breach the platform’s guidelines and land them in legal trouble. Notably, Google prohibits the usage of videos for applications that are independent of the platform.

Eventually, the company went ahead with the plan and transcribed more than a million hours of YouTube videos, and the text was fed to GPT-4, as per the report. Further, the NYT report also alleges that OpenAI President Greg Brockman was directly involved with the process and personally helped collect data from videos.

Speaking with The Verge, OpenAI spokesperson Matt Bryant called the reports unconfirmed and denied any such activities saying, “Both our robots.txt files and Terms of Service prohibit unauthorized scraping or downloading of YouTube content.” Another spokesperson, Lindsay Held told the publication that it uses “numerous sources including publicly available data and partnerships for non-public data” as its data sources. She also added that the AI firm was looking into the possibility of using synthetic data to train its future AI models.

Affiliate links may be automatically generated – see our ethics statement for details.

Check out our Latest News and Follow us at Facebook

Original Source

Apple Researchers Working on On-Device AI Model That Can Understand Contextual Prompts

Apple researchers have published a new paper on an artificial intelligence (AI) model that it claims is capable of understanding contextual language. The yet-to-be peer-reviewed research paper also mentions that the large language model (LLM) can operate entirely on-device without consuming a lot of computational power. The description of the AI model makes it seem suited for the role of a smartphone assistant, and it could upgrade Siri, the tech giant’s native voice assistant. Last month, Apple published another paper about a multimodal AI model dubbed MM1.

The research paper is currently in the pre-print stage and is published on arXiv, an open-access online repository of scholarly papers. The AI model has been named ReALM, which is shortened for Reference Resolution As Language Model. The paper highlights that the primary focus of the model is to perform and complete tasks that are prompted using contextual language, which is more common to how humans speak. For instance, as per the paper’s claim, it will be able to understand when a user says, “Take me to the one that’s second from the bottom”.

ReALM is made for performing tasks on a smart device. These tasks are divided into three segments — on-screen entities, conversational entities, and background entities. Based on the examples shared in the paper, on-screen entities refer to tasks that appear on the screen of the device, conversational entities are based on what the user has requested, and background entities refer to tasks that are occurring in the background such as a song playing on an app.

What is interesting about this AI model is that the paper claims despite taking on the complex task of understanding, processing, and performing actions suggested via contextual prompts, it does not require high amounts of computational energy, “making ReaLM an ideal choice for a practical reference resolution system that can exist on-device without compromising on performance.” It achieves this by using significantly fewer parameters than major LLMs such as GPT-3.5 and GPT-4.

The paper also goes on to claim that despite working in such a restricted environment, the AI model demonstrated “substantially” better performance than OpenAI’s GPT-3.5 and GPT-4. The paper further elaborates that while the model scored better on text-only benchmarks than GPT-3.5, it outperformed GPT-4 for domain-specific user utterances.

While the paper is promising, it is not peer-reviewed yet, and as such its validity remains uncertain. But if the paper gets positive reviews, that might push Apple to develop the model commercially and even use it to make Siri smarter.

Affiliate links may be automatically generated – see our ethics statement for details.

Check out our Latest News and Follow us at Facebook

Original Source

OpenAI’s GPT Store Runs Into Trouble With Spam, Impersonation, and Unlawful Bots

OpenAI has a big spam and policy violation problem in its GPT Store. The AI firm introduced its GPT Store in January 2024 as a place where users can find interesting and helpful GPTs, which are essentially mini chatbots programmed for a specific task. Developers can build and submit their GPTs to the platform and as long as they do not violate any of the policies and guidelines given by OpenAI, they are added to the store. However, it turns out the policies are not being followed stringently and many GPTs that appear to be violative of the regulations are flooding the platform.

We, at Gadgets 360, ran a quick search on the GPT Store platform and found that the chatbot marketplace is filled with bots which are spammy or otherwise violate the AI firm’s policies. For instance, OpenAI’s usage policy states under the section ‘Building with ChatGPT‘ in point 2, “Don’t perform or facilitate the following activities that may significantly affect the safety, wellbeing, or rights of others, including,” and then adds in sub-section (b), “Providing tailored legal, medical/health, or financial advice.” However, just searching up the word “lawyer” popped up a chatbot dubbed Legal+ whose description says, “Your personal AI lawyer. Does it all from providing real time legal advice for day-to-day problems, produce legal contract templates & much more!”

The example just shows one of many such policy violations taking place on the platform. The usage policy also forbids “Impersonating another individual or organisation without consent or legal right” in point 3 (b), but one can easily find “Elon Muusk” with an extra u added, likely to evade detection. Its description simply says “Speak with Elon Musk”. Apart from this, other chatbots that are treading the grey area include GPTs that claim to remove AI-based plagiarism by making the text seem more human and chatbots that create content in Disney or Pixar’s style.

These problems with the GPT Store were first spotted by TechCrunch, which also found other examples of impersonation, including chatbots that let users speak with trademarked characters such as Wario, the popular video game character, and “Aang from Avatar: The Last Airbender”. Speaking with an attorney, the report highlighted that while OpenAI cannot be held liable for copyright infringement by developers adding these chatbots in the US due to the Digital Millennium Copyright Act, the creators can face lawsuits.

In its usage policy, OpenAI said, “We use a combination of automated systems, human review, and user reports to find and assess GPTs that potentially violate our policies. Violations can lead to actions against the content or your account, such as warnings, sharing restrictions, or ineligibility for inclusion in GPT Store or monetization.” However, in our findings and based on TechCrunch’s report, it appears that the systems are not working as intended.

Affiliate links may be automatically generated – see our ethics statement for details.

Check out our Latest News and Follow us at Facebook

Original Source

Microsoft Copilot AI Chatbot’s Free Version Has Been Upgraded to GPT-4 Turbo

Microsoft Copilot has been silently upgraded to a more advanced artificial intelligence (AI) model, revealed a company executive on Tuesday. With the upgrade, Copilot, the AI-powered chatbot, is now powered by OpenAI’s GPT-4 Turbo model. Both the free tier and the Pro version can now access the latest model, however, only Copilot Pro users will have the option to revert to older GPT-4 model. Notably, the tech giant also recently revealed the launch date for its Copilot for Security platform, confirming that it will become available globally starting April 1.

Mikhail Parakhin, CEO of Advertising and Web Services at Microsoft, shared a post on X (formerly known as Twitter) on March 12, and said, “After quite some work, GPT4-Turbo replaced GPT-4 in the Copilot free tier. Pro users can still choose the older model, if prefer (there is a toggle).” The announcement comes without any fanfare, however, it is a significant upgrade that will impact regular users. Notably, Copilot Pro users could already access GPT-4 Turbo.

For starters, the knowledge cut-off of GPT-4-based Copilot was September 2021 but GPT-4 Turbo has a knowledge base up to April 2023. This will not be noticeable when searching for online topics as the AI chatbot is connected to the internet, however, for topics which cannot be readily found on the internet, the increased knowledge output is expected to result in more updated responses. Apart from the Balanced mode, the updated AI model’s support has also been added to the Creative and Precise modes.

GPT-4 Turbo was introduced by OpenAI in November 2023. It comes with an increased context window of 128,000 tokens, as opposed to GPT-4’s 8,000 tokens. The AI firm said that it can fit the equivalent of more than 300 pages of text in a single prompt. This means that the latest AI model can look through a larger set of data to find contextual information about a prompt. Interestingly, the free tier of OpenAI’s native chatbot ChatGPT is still running on GPT 3.5 AI model.

Meanwhile, Microsoft has also revealed the launch date for its Copilot for Security platform which will be available to enterprise customers starting April 1. Powered by OpenAI’s GPT-4 and Microsoft’s security models, it comes with multilingual capabilities, offering support for 25 different languages.

Affiliate links may be automatically generated – see our ethics statement for details.



Check out our Latest News and Follow us at Facebook

Original Source

Exit mobile version