#AI - Page 3
402 Stories
Meta claims that its largest Llama 3.1 405B model beats GPT-4o in several key benchmarks. However, in our reasoning tests, it remains behind GPT-4o. In fact, in my testing, Llama 3.1 405B ranks below Claude 3.5 Sonnet and Gemini 1.5 Pro. For detailed results, follow our article.
X (Twitter) is Training Grok AI with Your Data; Here’s How to Disable It
Arjun Sha
View quick summary
X (formerly Twitter) is secretly training its Grok AI on X posts by default and without seeking user consent. Thankfully, you can disable AI training from the web version of X. The dedicated setting is not available on the X app. You can follow our article and opt out of AI training on your X posts.
After months of anticipation, OpenAI has finally unveiled its AI-powered ChatGPT search engine, known as SearchGPT. It's designed to answer fast using authentic sources and clear attribution. Users can also find all the sources and engage with the web results. That said, it's available in a limited prototype which means users have to join the waitlist to receive access.
If you are on the lookout for the best AI tools for photo editing, follow our curated list of apps and services. With these new AI photo editors, you can remove or add new objects, change the background, isolate the subject, make fine adjustments, and do much more. You can even fix blurry images and upscale them to a higher resolution.
If you are looking for AI-powered tools for research purposes, we have curated the 10 best tools for academic research. These tools help you find relevant papers from a vast database, lets you extract key insights, find the general trend, and much more. From literature review to visualization of papers and chatting with local research papers, you can do all of it using these AI tools.
Llama 3.1 405B is the largest and most capable model from Meta. The company has also open-sourced it. You can start chatting with the model on Meta AI and WhatsApp if you are from the US. And for the rest of the regions, you can head to HuggingChat or Groq to access the 405B model for free.
Meta has introduced its new family of Llama 3.1 models which include the largest 405B model, 70B, and 8B models. The Llama 3.1 405B model outranks OpenAI's GPT-4o model in many key benchmarks. For the first time, we have an open-source model that competes with the closed ones. To learn more about the benchmarks, go through our article.
OpenAI Launches GPT-4o mini Model; Faster and Cheaper Than GPT-3.5 Turbo
Arjun Sha
View quick summary
OpenAI is back with a solid lightweight model that doesn't compromise on performance. The new GPT-4o mini model scores 82% on the MMLU benchmark and costs much less than GPT-3.5 Turbo, Gemini Flash, and Claude 3 Haiku. In fact, it outperforms some of the largest models such as Claude 3 Opus and GPT-4. You can start using the mini model on the free and paid version of ChatGPT starting today.
ChatGPT shows an 'Error in Moderation' and refuses to generate a response when the AI model thinks that your prompt is sensitive and might be used to generate harmful content. In such a scenario, it's recommended to tweak your prompt in such a way that it's not flagged by the AI model. Apart from that, you can re-login on ChatGPT to see if the problem has been fixed or not.
Vidnoz AI: The Effortless Tool to Create Stunning AI Videos in Minutes
Partner Content
View quick summary
If you are looking to create professional-quality AI videos at a fraction of the cost, I would recommend checking out Vidnoz AI. For budding content creators, Vidnoz AI offers hundreds of AI avatars, video templates, and AI voices. You can even generate AI scripts with a simple prompt. And if you want to personalize your video, you can create your own custom AI avatar and clone your voice as well.
If you are looking for an AI tool to upscale images, I would recommend Aiarty Image Enhancer. It's a fantastic program that generates more details to fix pixelation and removes blurriness from images. Overall, it's an image restoration program that actually works. Go through our guide to understand all the key features.
Google has released its latest open-source model called Gemma 2 27B. Until now, Meta ruled the open-source space with its Llama family of models. So in this article, I have compared Gemma 2 27B with Llama 3 70B to evaluate their prowess in a variety of tasks. Despite a much smaller size, Gemma 2 27B performs well, except in reasoning tests.
Gemini Now Supports File Uploads & Here’s How It Stacks Up Against ChatGPT
Arjun Sha
View quick summary
You can now upload documents to Gemini and ask questions or perform data analysis. Keep in mind, you need a Gemini Advanced subscription to be able to upload local files. In my testing, Gemini Advanced performed as good as ChatGPT. But ChatGPT offers file uploads for free and there is support for a wide variety of file formats as well.
OpenAI has finally released a statement saying that GPT-4o's Voice Mode is being delayed. It will be available by the end of the next month to a select group of paid users in Alpha. And by the end of this year, all ChatGPT Plus users will get the new feature. The company is also working to bring screen-sharing capability to ChatGPT.
Anthropic's new Claude 3.5 Sonnet model has performed remarkably well on benchmarks. So we have compared it with ChatGPT 4o and Gemini 1.5 Pro. In reasoning tests, Claude 3.5 Sonnet shows great intelligence and also does a great job at following user instructions. Even in coding, Claude 3.5 Sonnet generates bug-free code in comparison to ChatGPT 4o and Gemini 1.5 Pro.
Anthropic has announced a new Claude 3.5 Sonnet model which seems truly impressive. It's 2x faster than the largest Claude 3 Opus model and beats GPT-4o and Gemini 1.5 Pro in various benchmark tests. It's also a powerful vision model and can interpret graphs, diagrams, illustrations, and illegible handwriting perfectly.
While we are currently using AI chatbots like ChatGPT and Gemini, more powerful AI systems are on the horizon. AI agents, AGI, and Superintelligence are the next frontier of highly-powerful AI systems. Superintelligence, in particular, is said to be even more powerful than humans and AGI. To learn more about Superintelligence, go through our explainer.
OpenAI’s Former Chief Scientist Ilya Sutskever Launches a New AI Company
Arjun Sha
View quick summary
Ilya Sutskever, the brain behind ChatGPT at OpenAI, left the company last month and now, he has announced a new company called Safe Superintelligence Inc. (SSI). It aims to create a safe superintelligence to ensure safety always remains ahead of capabilities.
Google Gemini App Arrives in India with Support for 9 Regional Languages
Arjun Sha
View quick summary
Google has finally launched the Gemini app in India. It brings support for 9 Indian languages. Apart from that, Indian users can also access Gemini Advanced features such as file uploads and data analysis. In addition, you can chat with Gemini in the Google Messages app as well.
I Tested an Open-Source Alternative to Windows Recall That Works on Any PC
Arjun Sha
View quick summary
If you want to try out an alternative to Windows Recall, check out OpenRecall which is open-source and entirely free. It does all the processing locally and works on any x86 PC without an NPU. Keep in mind, the local vector database is not encrypted so that might be a major security concern.
Apple has developed two capable AI models to power Apple Intelligence features on iPhone, iPad, and Mac. For on-device processing, there is a small 3B model, and for complex tasks, there is a large, server-class model. The smaller 3B model performs better than Microsoft's Phi-3-mini and Google's Gemma 2B and 7B models. And Apple's large server model does better than GPT-3.5 Turbo.
To process AI requests on the cloud, Apple has built its own cloud server stack called Private Cloud Compute. Apple has built the cloud server using custom Apple silicon for quick AI inferencing. The Cupertino giant says that none of the personal user data is stored in the PCC system and not even Apple can access it.
Elon Musk Disses OpenAI-Apple Collab, Threatens Apple Device Ban at His Companies
Abubakar Mohammed
View quick summary
Elon Musk has slammed Apple's latest move to integrate OpenAI into Apple devices. Sharing his "completely unbiased" thoughts on the latest iOS release, he called Apple's integration of OpenAI an "unacceptable security violation," even threatening to ban Apple devices at his companies.
Anshuman Jain
In this piece, I discuss and highlight some of the key features of the Microsoft SwiftKey keyboard app, including clipboard sync with Windows devices, customizable AI stickers, a variety of themes, a built-in language translator, and a synced clipboard with Windows PCs.
At the WWDC 2024 event, Apple finally unveiled that ChatGPT integration is coming to iPhone, iPad, and Mac. It will be available to users for free. And if you have a ChatGPT Plus subscription, you can take advantage of those features on Apple devices.
Sagnik Das Gupta
ChatGPT offers you tons of GPTs to install and use from the GPT store. However, there are just way too many of them with similar ideas, and you might end up getting confused. So, we have handpicked the best ones to use, and the Canva GPT integration easily tops that list. Besides that, Scholar GPT takes care of your research needs. You also have super helpful educational GPTs like Khan Academy's Tutor Me tool and Code Copilot. In addition, you can also make use of the Mia AI GPT to get your very own therapist.
Data Science and Artificial Intelligence are closely related disciplines, but their difference lies in their scope, methodologies, and objectives. While Data Science is limited to data interpretation, AI models are aimed at creating intelligent systems that can perform tasks similar to humans.
A large language model is essentially a deep-learning algorithm that is designed to understand, process, and generate human language. It predicts the next word in a sentence based on the principles of probability. LLMs are trained on an extensive dataset of textual data from the internet, books, archives, etc.
The Recall feature is being called a security nightmare because of its poor implementation. Hence, users are looking to disable the Recall AI feature right away. We have added three different ways to turn off Recall on Copilot+ PCs. You can disable it from the Windows Settings, Registry or Group Policy Editor.
C2PA has developed a powerful tool called Content Credentials to detect AI-generated images. If the images have been modified or metadata has been removed, Content Credentials can still detect AI images and their source. Other than that, you can find inconsistencies in AI images and check for watermarks.
A Windows dev has managed to run the headline AI feature called Recall on unsupported Windows 11 PCs. If you have an older ARM PC, you can run Recall without any issues. x86 users can't run Recall yet as Microsoft has not released the ML models for this architecture. However, you can emulate it via a VM, but the experience is painfully slow.
This AI Platform Lets You Create New Episodes for Your Favorite TV Shows
Aparna Ukil
View quick summary
AI technologies haven't left any stone unturned to surprise us over the years and now, the award-winning startup Fable has come up with an AI streaming platform. So, let us find out more about it.
In its blog, OpenAI has revealed that the company is training its next big frontier model and it will unlock the next level of AI capabilities, bringing us closer to AGI. The company has also formed a Safety and Security Committee to evaluate upcoming models and look into critical safety and security decisions.
Meta Trains Its AI on Your Instagram and FB Photos; Here’s How to Opt Out
Arjun Sha
View quick summary
Meta is using images and other data from Instagram and Facebook to train its AI models. It's opt-in by default and users are being notified now. You can ask Meta to stop training on your personal data, but it will only apply to data gathered from third-party services. Meta has also made it harder to opt out by asking for evidence to further process the request.
Google’s AI is Losing It! Asks Users to Eat Rocks, Add Glue to Pizza & More
Arjun Sha
View quick summary
Google's new AI Overview experience in Search is rolling out to users in the US. People are complaining about the misinformation AI Overview is generating. We have collated some of the replies generated by Google's AI Overview. Essentially, Google has redefined the relationship of a search engine provider and taken the role of a publisher.
You can easily install the new Copilot app on Windows 11, 10, and older builds as well. I have extracted the new Copilot app from the 24H2 build and installed it on my PC running the 23H2 build. The new Copilot app is actually a web-based app and not a native Windows app.
Google and Microsoft are Already Testing Ads in AI Overview and Copilot
Arjun Sha
View quick summary
Google and Microsoft are now testing to display ads in AI Overview and Copilot. Google is initially rolling out ads in AI-generated results in the US. Microsoft is also adding ads to its generated responses on Copilot.
For the first time, a mega-corporation like Microsoft has released an open-source multimodal model. The new Phi-3 Vision model is trained on 4.2B parameters and brings a context length of up to 128K tokens. Microsoft has also released Phi-3 Small (7B) and Phi-3 Medium (14B) models.
In a major push to kickstart the new era of AI PCs, Microsoft has announced a new Recall AI feature on Copilot+ PCs. It actively takes screenshots of your PC every few seconds to create a local semantic index. The processing is done locally and the information is encrypted. However, would you be comfortable actively sharing your screen for the Recall feature? We find out.
Last week, OpenAI demoed the 'Sky' voice during the ChatGPT 4o launch. The voice seemed straight out of the movie 'Her', where Scarlett Johansson lent her voice to the AI character 'Samantha'. Now, it turns out Sam Altman, the CEO of OpenAI, was in contact with Scarlett Johansson to voice the new ChatGPT 4o system. However, Johansson had refused the offer, way back in September 2023. Now, Johansson is seeking clarity and transparency as to how the 'Sky' voice was developed. Meanwhile, OpenAI has paused the 'Sky' voice on ChatGPT.
Gemini 1.5 Flash is an Underrated Gem You Need to Try Right Now: Here’s How
Arjun Sha
View quick summary
At the I/O 2024, Google unveiled many AI models, but Gemini 1.5 Flash remained under the radar. It's a lightweight AI model that delivers remarkable speed and efficiency with support for multimodal reasoning and a large context window of 1 million tokens. It's also very cheap to run. You can try the model on Google AI Studio for free and without any waitlist.
If you don't like the new AI Overview on Google Search, you can turn it off by following our guide. We have added two extensions to entirely block AI-generated answers on Google Search. Apart from that, we have added three other ways to remove AI Overviews on desktop, Android, and iOS.