Search
Close this search box.

Want to be Mentored?

This Week in AI: SearchGPT, Llama 3.1, and More Trending News

AI news this week

Table of Contents

OpenAI Introduces SearchGPT:

OpenAI has launched SearchGPT, an AI-powered search engine for 10,000 initial users, offering real-time information and interactive content summaries.

OpenAI has introduced its new search engine, SearchGPT, which is powered by artificial intelligence and offers real-time access to information from across the internet. Unlike traditional search engines that provide a list of links, SearchGPT organizes and presents summarized findings in response to user queries.

The prototype service utilizes GPT-4 models for now and will be initially available to 10,000 test users. It aims to integrate with ChatGPT in the future. This move puts OpenAI into competition with Google’s AI-enhanced search features as well as startup Perplexity.

Notably, OpenAI emphasizes collaboration with news partners like The Wall Street Journal and The Associated Press during development. Publishers have control over how their content appears in searches; they can opt out of having their content used while still being surfaced on the platform.

One unique aspect of this release approach is that it allows leeway if there are inaccuracies or attribution issues since it’s labeled as a prototype at present.

This announcement comes after reports earlier this year about efforts by OpenAI aiming towards creating such product following aggressive hiring attempts targeting former Google employees for a dedicated team working on this project.

While costs continue mounting up due high usage numbers driving up compute expenses stemming from millions using free versions of other products offered by company including ChatGPT ,there isn’t any ads model yet visible indicating monetization strategy may come later down line once initial testing phase concludes .

Source

Meta Releases Llama 3.1:

Meta’s new open-source AI model, Llama 3.1, boasts 405 billion parameters and aims to surpass ChatGPT in usage with its advanced features and collaboration with major tech companies.
Meta has unveiled Llama 3.1, the largest open-source artificial intelligence (AI) model ever released. This model boasts an impressive performance that surpasses industry leaders like GPT-4o and Anthropic’s Claude 3.5 Sonnet on various benchmarks.

Key Features of Llama 3.1 Llama-based Meta AI assistant is now available in more countries and languages. It includes a feature to generate images based on specific likenesses. The CEO predicts it will become the most widely used assistant by year-end, outperforming ChatGPT.

Technical Details This advanced version of Llama has a whopping 405 billion parameters and was trained using over 16,000 Nvidia H100 GPUs. While development costs are undisclosed, experts estimate them to be hundreds of millions due to expensive hardware involved.

Open Source Strategy Mark Zuckerberg believes open-source models will soon dominate proprietary ones within the industry as they did with Linux for operating systems powering numerous devices today. To promote adoption among developers globally, collaborations have been established with several tech giants including Microsoft, Amazon, Google, Nvidia, Databricks.

Usage & Accessibility

It’s cost-effective compared to other similar models when deployed in production environments

Companies can train it according to their requirements since its weights are being made public

While details about training data remain confidential, synthetic data generated internally played a crucial role in enhancing this massive model.

In terms of usage accessibility:

• Lama remains freely accessible through WhatsApp and the official website but may switch users from full-scale use after exceeding certain prompts per week • A new “Imagine Me” feature allows insertion of personal likeness into generated images via facial scans captured via phone cameras

Overall, this signifies just early stages toward finding market fit for these applications though excitement continues growing around its potential across different industries.

As the technology continues to develop and improve, it is possible that we will see even more advanced features in the future.

Source

Former Tesla Exec Launches Mytra:

Chris Walti, ex-Tesla humanoid head, unveils Mytra, a startup focused on advanced warehouse robotics to enhance fulfillment operations and address labor shortages.

Former Tesla Humanoid Head Launches a Robotics Startup: Mytra, a robotics startup founded in May 2022 by ex-Tesla and Rivian automotive professionals, aims to innovate warehouse automation. The company is entering a competitive market that has surged due to ongoing labor shortages exacerbated by the pandemic.

Background of Mytra’s CEO: Chris Walti, Mytra’s co-founder and CEO, has a history at Tesla where he dealt with automation challenges during the Model 3 production ramp. His experiences led him to create in-house solutions for autonomous mobile robots (AMRs) capable of handling heavy payloads.

Unique Offerings: Mytra’s technology focuses on managing heavy loads and offers a highly flexible system for moving items, distinguishing it from existing robotic storage solutions.

Market Engagement: The startup has attracted attention from major clients, including a pilot project with Albertsons and interest from several Fortune 50 companies.

Funding: Mytra recently secured $50 million in Series B funding, raising its total to $78 million, with investments from firms like Greenoaks and Eclipse

Source

Kling AI Video Generator Goes Global

Kling has launched its AI video generator worldwide, providing a new competitor to OpenAI’s Sora.

Kling AI, previously limited to China, is now available worldwide at KlingAI.com. The platform lets users create videos using text or images and provides 66 free daily credits upon registration with just an email address. It competes with OpenAI’s Sora but boasts faster generation times of around 5 minutes for a video.

Users can choose from three resolutions and control the camera movement in their videos. Multilingual support allows high-quality results in English and Mandarin prompts, while adjustments can be made between creativity and relevance levels.

In tests against OpenAI’s Sora, Kling AI outperformed when creating realistic scenes like wolf pups playing on a gravel road or depicting sci-fi scenarios such as a female robot walking in a futuristic city.

However, it faced challenges generating specific styles; for example animating anime-style cyberpunk samurai resulted in 3D animation instead of requested 2D style but captured key elements accurately. Additionally fine patterns involving thin parallel lines posed issues during generation.

Despite these limitations, Kuaishou plans future updates including extended features such as full HD video production mode alongside subscription programs allowing increased daily credit limits.

Restrictions prevent the creation of violent content or depictions of famous people making NSFW content forbidden too.

The global launch mirrors strategies used by other Chinese tech companies aiming to reach international audiences through separate web portals while retaining original versions domestically – similar branding tactics employed by TikTok (Douyin) & WeChat (Weixin).

Source

Reddit Blocks Search Engines and AI Bots

Reddit now blocks major search engines and AI bots unless they pay for access.

Recent reports reveal that Reddit is preventing search engines from displaying recent posts and comments unless they pay for access. Currently, Google remains the only mainstream search engine showing new results when using a specific trick to search within reddit.com. This action excludes Bing, DuckDuckGo, and other alternatives due to an exclusive deal with Google allowing AI training on Reddit’s content.

This move aims at protecting its data amidst efforts to generate revenue and satisfy investors’ demands. The platform updated its robots.txt file in June as part of enforcing this policy against web scraping activities.

Despite ongoing discussions with various search engines about accessing their content including use for AI purposes, agreements have been challenging due to concerns over enforceable promises regarding usage rights.

As a result of these changes, users may find it frustrating not being able to easily access human-generated responses outside of Google or platforms relying on it.

This may lead to a lack of diversity and accuracy in the information available through AI systems.

Source:

Bing’s AI Redesign:

Bing has revamped its search results, prioritizing AI-generated content over traditional search listings.

**Google DeepMind Introduces AlphaFold 3: A Breakthrough in Predicting Molecular Structures**

DeepMind’s new AI model, AlphaFold 3, goes beyond predicting protein structures to now include DNA, RNA, and smaller molecules. This advancement will aid research in medicine, agriculture, materials science, and drug development. The model shows a 50% improvement in accuracy compared to previous versions. It utilizes a diffusion method to generate 3D models of molecular combinations input by researchers.

Isomorphic Labs has already benefited from using this technology for internal projects related to disease targets understanding with positive outcomes seen so far.

To facilitate responsible deployment of the tool among scientists without access to significant computational resources or funds for licensing fees Google is offering free use through its AlphaFold Server platform powered by AlphaFold 3.

Lastly; Google emphasizes their commitment towards working responsibly within biosecurity regulations as they acknowledge potential risks associated with such powerful technologies like misuse leading them working closely with experts even before launching the product to ensure proper safeguards are in place.

Source

DeepMind’s Protein Folding Breakthrough:

Google DeepMind’s AlphaFold has predicted the 3D structures of almost all known proteins, promising major advancements in biology and medicine.

DeepMind has launched the upgraded AlphaFold 3, an AI model that not only predicts protein structures but also models DNA, RNA, and smaller molecules. This advancement will aid researchers in various fields such as medicine, agriculture, materials science, and drug development with improved accuracy by 50%. The new model is accessible through the research platform AlphaFold Server for academic use. Despite its potential biosecurity concerns regarding misuse of this technology to engineer harmful pathogens or toxins raised by experts at Google’s paper on responsible deployment of these models – measures have been put in place to mitigate risks. Moreover Isomorphic Labs is already leveraging Alphafold 3 for internal projects yielding better understanding towards disease targets.

Additionally, ongoing discussions and collaborations with leading biosecurity organizations are being pursued to ensure responsible use of this technology.

Source

Netflix Enhances Content Personalization:

Netflix’s new AI-driven recommendation engine offers more tailored content suggestions based on user preferences and viewing habits.

Netflix uses artificial intelligence (AI) and machine learning (ML) to recommend shows based on your preferences. The platform has 260 million engaged users, generating a revenue of $33.724 billion in 2023.

History and Growth

Founded in the late ’90s as a DVD rental service, Netflix transitioned into streaming services in the early 2000s. It initially focused on licensing existing TV shows and films but later began producing original content like “House of Cards” and “Stranger Things.”

User Engagement

As of 2023-24, over 260 million subscribers globally spend an average of 3.2 hours daily consuming its content across various countries.

AI Implementation for Recommendations

Through AI-powered recommendation algorithms that analyze user behavior data such as viewing habits, ratings, searches, etc., it provides personalized recommendations leading viewers to engage with more content they enjoy.

Benefits for Viewers:

Personalized Content: Tailored suggestions cater to individual tastes. 2.Enhanced Quality: Optimal video quality is maintained by adjusting according to network conditions.

Benefits for Netflix:

1.Increased Revenue: By attracting more users through tailored recommendations 2*Competitive Edge: *Setting standards using advanced technology like ML & Data Science

In addition, By implementing cutting-edge technologies alongside investing heavily—$8 billion this year—in new programming indicates their commitment towards enhancing customer experience .

To summarize,*Netflix’s use case illustrates how integrating AI can improve business operations while providing customers with high-quality entertainment options.

So next time you’re enjoying some binge-watching session on Netflix…don’t forget about all those smart algorithms working behind the scenes!

Source

DeepMind’s AlphaProof Solves Olympiad Math:

DeepMind’s AlphaProof model successfully tackled four out of six International Math Olympiad problems, showcasing AI’s potential in solving complex logic puzzles.

Two new AI systems, AlphaProof and AlphaGeometry 2, have made significant progress in solving complex mathematical problems. They were put to the test by participating in this year’s International Mathematical Olympiad (IMO) where they successfully solved four out of six challenging math problems.

The IMO is a prestigious competition for young mathematicians that has been held annually since 1959. This year, our AI systems achieved results equivalent to receiving a silver medal at the competition — an impressive feat given its difficulty level.

AlphaProof uses reinforcement learning based on formal math reasoning while coupling pre-trained language models with advanced algorithms like Lean. It proved itself capable of tackling two algebraic problems as well as one number theory problem from this year’s IMO contest—this included solving one of the hardest challenges faced by only five human participants.

On the other hand, AlphaGeometry 2 focuses specifically on geometry-solving tasks using neuro-symbolic hybrid technology based on Gemini; it was able to solve Problem 4 within just seconds after being presented with its formalization for IMO2024—a remarkable improvement compared to its predecessor which could tackle only about half such issues over historical data spanning past decades before training began. In addition, we tested another system built upon Gemini designed around natural language processing without necessity translation into formal languages also showing promising potential when combined with various AIsystems

Both these breakthroughs open exciting opportunities for future collaboration between mathematicians and powerful Ai tools.These advancements showcase howAis are becoming increasingly proficient not only in mathematics but also broader reasoning capabilities. We extend our gratitude towardstheInternationalMathematicalOlympiad organization for supporting this tremendous effort to advance computationalmathematicsOur team will continue exploring different approaches for mathematical reasoning,and look forward to sharing more details about Alphaproofsoon . Our team will continue exploring different approaches for mathematical reasoning, and look forward to sharing more details about Alphaproof soon, as we believe it has the potential to revolutionize computational mathematics.

Source

Mistral’s Large 2 Model Debuts:

Mistral has introduced Large 2, a model with 123 billion parameters, offering competitive performance in multiple languages and available for researchers at no cost.

The latest version of Mistral Large, known as Mistral Large 2, has raised the bar for cost efficiency and performance. It boasts a context window size of 128k and supports multiple languages such as French, German, Spanish along with various coding languages like Python and Java.

This model is designed to handle long-context applications efficiently on a single node due to its massive parameter capacity of 123 billion. However commercial usage requires obtaining a license from Mistral AI.

In terms of accuracy metrics evaluation showed an impressive score which outperforms many leading models including GPT-4o. The developers made substantial efforts towards improving reasoning capabilities by minimizing errors or irrelevant information generation during conversations.

Mistral Large was also trained extensively using multilingual data resulting in enhanced proficiency across various languages benefiting business use cases involving multilingual documents

Additionally this new update focuses on concise responses keeping interactions succinct without compromising accuracy. Its improved function calling skills enable it to power complex business applications proficiently. To access these features users can utilize “mistrial-large-2407” through la Plateforme or le Chat under API name mistrial-large-2407 hosted via HuggingFace’s platform for free weights availability. Moreover, fine-tuning capabilities are extended now available for both general purpose models – Mistrai Nemo & Codestra alongside older versions that remain deployable through SDKs provided by the developer while partnering with major cloud service providers making them accessible globally via Managed APIs offered primarily within Google Cloud Platform Vertex Al among others Amazon Bedrock Watsonx.ai etc.

Overall these advancements position the newest iteration – Mistral large at forefront offering superior functionality informed responsiveness catering wide array language needs enabling efficient deployment supporting cutting-edge technology transforming industries utilizing state-of-the-art artificial intelligence tools effectively meeting modern requirements indispensable upgrading system managing vast linguistic variation driving innovation significantly enhancing real-world application potential overall productivity influencing global landscape revolutionizing technological growth ensuring optimal user experience demonstrating unparalleled prowess solidifying itself leader shaping future developments propelling us forward into digitized era seamlessly integrating advanced computational technologies optimizing widespread digital ecosystem . Its comprehensive features and capabilities make it a game-changer in the field of artificial intelligence, setting new standards for language processing and deployment efficiency.

Source

Microsoft and Lumen Technologies Partnership:

Microsoft has partnered with Lumen Technologies to enhance AI infrastructure, improving Lumen’s cash flow by $20M.

Microsoft has announced a partnership with Lumen Technologies to enhance its AI workload capacity. This collaboration aims to leverage Lumen’s edge computing platform to support Microsoft’s Azure cloud services. By integrating Lumen’s edge infrastructure, Microsoft intends to improve the performance and efficiency of AI applications, particularly those requiring low-latency and high-bandwidth capabilities. This partnership is part of Microsoft’s broader strategy to expand its cloud computing capabilities and meet the growing demand for AI-driven solutions

Source

Apple Music’s AI-Generated Artwork:

Apple Music will introduce AI-generated artwork for playlists in iOS 18, featuring the new “Image Playground” tool.

Apple Music is set to introduce a new feature in iOS 18 that will allow users to generate AI-created artwork for playlists. This feature, discovered in newly revealed code, includes a “create image” button that opens the Image Playground, Apple’s prompt-based image generator. While the specifics are still under wraps, the AI-generated art is expected to focus on fun and playful illustrations rather than photo-realistic images, aligning with the aesthetic of classic album covers.

The feature is anticipated to be available in the upcoming iOS 18 release, likely in September alongside the iPhone 16. However, it may only be accessible on newer devices such as the iPhone 15 Pro, iPhone 15 Pro Max, and Apple Silicon MacBooks and iPads. The rollout of this feature outside the US is expected to be delayed, with European users potentially waiting until 2025 for access

Source

Google’s Gemini Chatbot Updates:

Google’s Gemini chatbot has received significant updates, improving user interaction and functionality.

Google has announced significant updates to its Gemini chatbot, previously known as Bard, aiming to enhance its competitiveness against other AI chatbots like ChatGPT. The latest version, Gemini 1.5 Flash, features an upgraded large language model (LLM) that promises faster and higher-quality responses, particularly in reasoning and image comprehension.

Key updates include:

Expanded Context Window: The free version’s context window has been increased to 32K tokens, allowing Gemini to process more information simultaneously, which enhances the quality of responses.

Document Uploading: Users will soon be able to upload files from Google Drive or their devices, enabling better interaction with documents and facilitating tasks like answering questions or summarizing content.

Data Analysis and Visualization: Gemini will soon include features for analyzing data files and providing insights through charts and graphics.

Fact-Checking Enhancements: Google has improved the user experience by displaying relevant Google links alongside responses, allowing for easier fact-checking.

Expanded Access: Gemini is being rolled out to more users, including availability in Google Messages across the European Economic Area and plans to extend access globally to teens with appropriate safeguards.

These updates are designed to make Gemini a more formidable player in the AI chatbot space, with ZDNET planning further evaluations to assess its performance post-update

Source

Galbot Enters Humanoid Robot Market:

Backed by the Hong Kong government, Beijing-based Galbot plans to introduce humanoid robots globally, leveraging China’s hardware and data resources.

Galbot, a Beijing-based robotics startup backed by the Hong Kong government, is positioning itself to compete in the global humanoid robot market against major tech companies like Tesla. Co-founder Yao Tengzhou stated that Galbot will use Hong Kong as a gateway to introduce its robots to developed markets where labor costs are high. The company focuses on developing embodied artificial intelligence (AI) and believes that a new technological wave will help level the playing field against larger firms, despite acknowledging the fierce competition ahead.

Investment in humanoid robots has surged globally, driven by a generative AI boom, with Chinese startups raising significant funds this year. Galbot, established in May 2023, has garnered support from major investors, including the Hong Kong government and various venture capital firms. The company plans to set up a local unit in Hong Kong and aims to start accepting pre-orders in the latter half of 2024, with deployment of its robots expected by the end of the year. Yao emphasized that China’s strong hardware supply chain and abundant data could provide a competitive edge for Galbot in the evolving market

Source

 

Related Posts

Black Forest Labs AI image generator API interface for visual content creation
AstraZeneca AI partnership for cancer drug trials
Captions AI dashboard for AI-powered content automation
Google Lens AI-powered search tool in action

get notified for all future articles

Join my newsletter for Startup founders, Sr. Marketers, AI enthusiasts and discover deeply researched insights ! Exclusive Content | Actionable Strategies | Latest Trends. No spam, only the good stuff. Unsubscribe anytime!