#AI - Page 3
486 Stories
OpenAI has finally launched Sora, its text-to-video AI tool. You can start using Sora at sora.com with your ChatGPT Plus or Pro subscription. You can generate videos up to 20 seconds at 1080p resolution.
If you are unable to scroll down on ChatGPT, follow our tutorial to fix the issue. You can update your browser to the latest version or change the Overflow property from the Inspect window.
Elon Musk-led xAI has made the Grok AI chatbot free for everyone. You can chat with the Grok 2.0 model, generate AI images, and upload files and images too. Free users on x.com can send up to 10 messages every two hours.
Meta’s New Llama 3.3 70B Model Matches 405B’s Performance at a Lower Cost
Arjun Sha
View quick summary
Meta has launched a new cost-efficient text-only model called Llama 3.3 70B. It nearly matches the performance of the larger 405B model. It achieves 86.0 on the MMLU benchmark.
I Asked Google’s AI Podcast Hosts to Roast Spotify Wrapped and It Was Hilarious
Anshuman Jain
View quick summary
Spotify Wrapped 2024 featured two AI podcast host as its key highlight this year but due to limited availability it didn't turn out too well. It was only available in 7 countries leaving out the rest of the audience so in this read, we are using the same Notebook LM to voice our frustration for Spotify.
OpenAI has introduced a new subscription tier called ChatGPT Pro that costs $200 per month. It hosts the full o1 model by replacing the o1-preview model. ChatGPT Pro also gets an exclusive 'o1 pro mode' that uses more compute resources to solve harder problems.
Gemini on Android now lets you perform local actions quickly, thanks to the Utilities extension. You can also perform multiple local actions at once. For example, you can say, "Turn off battery saver and take a selfie with a 10-second timer."
Speaking at the NYT's DealBook summit, OpenAI CEO Sam Altman said AGI will matter much less and the safety concerns don't come at the AGI moment. He also said OpenAI will achieve AGI much sooner than most people think.
12 Days of OpenAI: ChatGPT Maker to Launch Sora and New Features Starting Tomorrow
Arjun Sha
View quick summary
OpenAI has confirmed that in the next 12 days, it's going to release new AI products and features non-stop. The company is likely to launch Sora and a reasoning model, among other things.
Amazon has released Nova AI models that cover several Generative AI use cases. Nova Pro is a multimodal model and it competes with Claude 3.5 Sonnet (New), GPT-4o, and Gemini 1.5 Pro. Nova Reels can generate AI videos and Nova Canvas can produce AI images.
An X user claimed that Microsoft is using Word documents to train its AI models. We took a closer look and found that it was not the case. Microsoft also refuted the claim on X.
You can easily run AI models on your Raspberry Pi locally. There are many small models that work decently well on Raspberry Pi. The best part is that you don't need additional hardware to run LLMs.
A Chinese AI lab, DeepSeek, has released a powerful reasoning model called "DeepSeek-R1-Lite-Preview." It works similarly to OpenAI's flagship o1 models. It performs reasoning steps before giving a final answer. In key benchmarks such as AIME 2024 and MATH, the DeepSeek model outperforms OpenAI's o1 model.
Microsoft is Launching Copilot Actions to Automate Routine Tasks Using AI
Arjun Sha
View quick summary
Microsoft has announced Copilot Actions that allow users to automate repetitive tasks. It can help you generate weekly reports or create a summary of the meeting.
Google’s Experimental Gemini Model Tops the Leaderboard, But Stumbles in My Tests
Arjun Sha
View quick summary
Google's upcoming Gemini-exp-1114 model has topped the LMArena leaderboard, outranking OpenAI and Anthropic models. However, in my reasoning tests, the model fails to answer tricky questions. In my book, it's still behind OpenAI's o1 models.
ChatGPT Desktop App For Windows Now Available to All Users; Download Here
Arjun Sha
View quick summary
OpenAI has finally opened access to the ChatGPT desktop app on Windows 11 and 10 to all users. You no longer need a paid subscription.
The latest ChatGPT desktop app on macOS can see what is on your screen and offer contextual suggestions. That said, it only works with coding editors such as VS Code, Xcode, Terminal, and more.
OpenAI is reportedly planning to release its AI agent codenamed 'Operator' in January next year. The AI agent can perform actions in a web browser based on user input.
In the AI field, it was long believed that scaling and making larger models would lead to greater intelligence. However, Ilya Sutskever now says that LLM scaling has reached its limits. OpenAI, Google, and Anthropic are now focused on inference scaling to improve model performance.
ChatGPT Search is a promising alternative to Google Search, but it doesn't cover all types of search queries. For informational queries, ChatGPT Search is good, but for local, navigational, and shopping queries, it falls short.
Perplexity is working on an AI-powered shopping tool to make online shopping hassle-free. It searches the web to find the best product based on your prompt and you can quickly place the order without leaving the platform. It seems Perplexity is trying to compete with Google Shopping.
You can move away from Google and switch to ChatGPT Search to find the latest information from the web. OpenAI has released an official ChatGPT Search Chrome extension that replaces Google with ChatGPT Search. You can install the extension on any Chromium-based browsers.
Google Vids is a new AI-powered video marketing tool for Google Workspace users. It allows users to create engaging videos by typing a prompt, with AI handling the rest. The tool includes features like voiceovers and image generation, making it a lucrative option for businesses looking to streamline their video content creation process.
Jarvis is Almost Here; Google Accidentally Leaks Its New AI Extension
Sagnik Das Gupta
View quick summary
According to a new report, Google accidentally posted an internal preview of its Jarvis AI web companion. However, they quickly removed the listing, although a couple of users managed to download it. But, it wasn't that easy. Read it all here!
OpenAI has released ChatGPT Search to challenge Google in the search domain. It can browse the web and find current information on all the latest events. You can also install the Chrome extension to replace Google Search with ChatGPT Search.
Meta has reportedly been at work, eight months in, working on its very own AI-powered search engine to reduce its dependence on Google and Bing for current event queries on Meta AI chatbot. Read on to know more!
Microsoft Sidesteps OpenAI Again; Integrates Claude and Gemini into GitHub Copilot
Arjun Sha
View quick summary
Microsoft has brought Claude 3.5 Sonnet and Gemini 1.5 Pro to GitHub Copilot in VS Code. For the first time, the company has integrated proprietary models from rival AI firms, overlooking OpenAI.
Similar to ChatGPT's Code Interpreter which runs Python code, users can now run JavaScript code in Claude, and it runs in the browser itself. It can come in handy for performing data visualizations, number crunching, and more. Users can also use it to run various mathematical calculations.
Google’s New Online Course Teaches You How to Effectively Talk to AI Chatbots
Sagnik Das Gupta
View quick summary
Google has now released an official Prompting Essentials course in partnership with Coursera to help users learn everything about efficiently crafting prompts to interact with AI chatbots. Check out all the course details right here!
Apple Intelligence Features: All New AI Features Coming to iPhone, iPad & Mac
Kanika Gogia
View quick summary
The Cupertino giant has revamped Siri and now it comes with on-screen awareness. It can perform many on-device tasks with ease. It also creates a semantic index on your device to pull out relevant information and perform tasks intelligently. There are also other Apple Intelligence features and experiences in the Mail app, Notes app, and other first-party apps.
Sam Altman Denies Reports of OpenAI Launching GPT-5 ‘Orion’ by December
Arjun Sha
View quick summary
According to a new report, OpenAI is planning to release its next big model, GPT-5 aka Orion by December this year. Microsoft is also readying its Azure infrastructure to host the Orion model and it will be completed by November.
The wait is over! Google has finally open-sourced its AI text watermarking and detection tool called SynthID. Announced last year, the tool can be an absolute game-changer in standardizing the detection of AI-generated content. More details here!
Miles Brundage, former Head of Policy Research and Senior Advisor for the AGI Readiness team has quit OpenAI. Brundage says he wants to work in the non-profit sector for greater freedom to publish.
Beebom Staff
Those looking for an AI video generator can check out Wondershare Virbo which delivers studio-quality AI videos in minutes. You can generate a script using AI and add natural speech using the text-to-speech model. It offers a friendly interface and you can even use it on your desktop and smartphone.
Character.ai Issues Apology After Teen Commits Suicide Over AI Chatbot Obsession
Arjun Sha
View quick summary
In a shocking incident, a teenager from Florida committed suicide after engaging with an AI character on Character.ai. The 14-year-old teen stopped interacting with people around him and isolated himself.
Anthropic’s Claude AI Can Now Control Your Computer Without Any Help
Sagnik Das Gupta
View quick summary
Anthropic's new Claude 3.5 Sonnet model now has the ability to glance at the computer screen and carry out human actions like clicking, typing, and moving the cursor. Here's all you need to know in detail!
Microsoft AI Chief Apparently Yelled at OpenAI Employees; Cracks Starting to Appear?
Arjun Sha
View quick summary
A recent report by The New York Times reveals that OpenAI and Microsoft are at loggerheads. According to the report, Microsoft AI CEO Mustafa Suleyman yelled at an OpenAI employee over the delayed transfer of advanced AI technologies to Microsoft. OpenAI, on the other hand, wants to set itself free from Microsoft's control, and looking to invoke the AGI clause.
Meta’s New Spirit LM Open-Source Model Can Mimic Human Expressions
Sagnik Das Gupta
View quick summary
Meta takes multimodality seriously and launches Spirit LM, that can mimic human-like vocal expressions. It uses both speech and text to generate human-like expressions.
NotebookLM Now Lets You Customize AI Podcasts and the Results are Unreal
Arjun Sha
View quick summary
You can now customize AI podcasts on NotebookLM. Google has added an option to specify your preferences which will guide the AI hosts. It's like passing a note to the AI hosts. Besides that, you can listen to Audio Overview in the background while working on your notebooks.
OpenAI has finally released the ChatGPT app for Windows 11 and 10. You can download the app right now from our article. Unlike Copilot which is a web app, the ChatGPT app seems to be built using the native WinUI framework.
This High Schooler’s App Brings Apple Intelligence-like Writing Tools to Windows 11
Arjun Sha
View quick summary
A high schooler from Bangalore has developed an app that brings Apple's Writing Tools to Windows 11. It works amazingly well and supports all apps and windows. The tool uses Google's Gemini 1.5 Flash model to process the text quickly.
Nvidia Releases Nemotron 70B Model; Claims to Beat GPT-4o and Claude 3.5 Sonnet
Arjun Sha
View quick summary
Nvidia's latest Nemotron 70B model claims to beat GPT-4o and Claude 3.5 Sonnet models. It's trained on top of Llama 3.1 70B model using RLHF. Nvidia says it can correctly answer the strawberry question without using reasoning tokens.