#ai-models
#ai-models

Nvidia's AI chatbot now supports Google's Gemma model, voice queries, and more

Nvidia updates ChatRTX chatbot with more AI models for RTX GPU owners. [ more ]

Theregister

2 days ago

Big name US newspapers sue Microsoft, OpenAI over copyright

Big-city American newspapers sue Microsoft and OpenAI for using copyrighted articles to train AI models without compensation. [ more ]

www.independent.co.uk

Apple releases AI that could show future of iPhones

Apple is releasing OpenELM models hinting at future AI features focused on efficient models for consumer devices like the iPhone. [ more ]

The Verge

Apple's new AI model hints at how AI could come to the iPhone

Apple is focusing on generative AI for local device use. [ more ]

InfoQ

moreArtificial intelligence

Stability AI Releases 3D Model Generation AI Stable Video 3D

Stable Video 3D (SV3D) by Stability AI excels in 3D object generation and novel view synthesis, surpassing previous models. [ more ]

openai

Axios

1 day ago

Data science

A mystery chatbot came and went. It's probably a new OpenAI product

The mystery chatbot likely an improvement on existing models, not a new one. [ more ]

Forbes

4 days ago

EU data protection

Data Activists Target OpenAI In Challenge To ChatGPT's 'Hallucination' Problem

Privacy activists filed a complaint against OpenAI for ChatGPT's misinformation issues, urging tech firms to address AI tool problems. [ more ]

Exchangewire

Comcast Sells $1.2bn in Advertising for Olympics; OpenAI and Meta to Release New AI Models; ByteDance Reports 60% Profit Jump

Comcast sells $1.2bn in ads for Olympics, a new record with significant sales from new sponsors.

OpenAI and Meta set to release new AI models capable of reasoning and planning, aiming for advancements in problem-solving abilities. [ more ]

TechCrunch

4 weeks ago

OpenAI expands its custom model training program | TechCrunch

OpenAI is expanding Custom Model program for tailored AI models

Assisted fine-tuning introduced to enhance model performance [ more ]

Ars Technica

Elon Musk's xAI releases Grok source and weights, taunting OpenAI

xAI's Grok-1 model is a 314-billion-parameter AI assistant similar to ChatGPT, released in base form for public access.

The base model released by xAI is not fine-tuned for a specific task like dialogue, requiring additional work for conversational use outside xAI. [ more ]

Theregister

Intel, Ampere show LLMs on CPUs isn't as crazy as it sounds

Smaller AI models can now be efficiently run on CPUs with reduced latency, challenging the dominance of GPUs for AI tasks. [ more ]

Theregister

Data science

Stability AI decimates staff just weeks after CEO leaves

Stability AI is restructuring after CEO departure, leading to layoffs to improve sustainability and focus on operations. [ more ]

TechCrunch

Google Gemini: Everything you need to know about the new generative AI platform | TechCrunch

Google's Gemini is a suite of generative AI models, including Ultra, Pro, and Nano, capable of multimodal tasks, unlike text-exclusive models like LaMDA. [ more ]

TNW | Data-Security

French competition watchdog fines Google 250M for AI copyright breaches

Google fined €250mn for breaching EU IP rules

Google used publishers' content to train AI models without permission [ more ]

Hindustan Times

Google releases Gemma: Laptop-friendly 'open' source AI after Meta| Explained

Google releases new AI models called Gemma for developers to use and build upon for free.

Gemma models are optimized for Google Cloud, potentially attracting software engineers and new cloud customers. [ more ]

Engadget

Reddit is licensing its content to Google to help train its AI models

Google and Reddit have struck a deal for Google to train its AI models using Reddit's user-generated content.

Access to Reddit's data became a contentious issue when the company announced charging developers, leading to restrictions and protests. [ more ]

Engadget

Google's Duet AI for businesses is now called Gemini too

Google rebrands its Duet AI for enterprises to Gemini

Businesses can choose between Gemini Business and Gemini Enterprise plans for Workspace users [ more ]

Engadget

Google's Gemini 1.5 Pro is a new, more efficient AI model

Google has unveiled Gemini 1.5 Pro, which offers enhanced performance compared to the previous model.

The company emphasizes the safety of its AI models while highlighting their advancing capabilities. [ more ]

moregoogle

microsoft

ReadWrite

Microsoft launches its smallest AI model yet: Phi-3 Mini

Microsoft introduced Phi-3 Mini, a compact AI model with 3.8 billion parameters, geared towards personal and corporate use. [ more ]

Ars Technica

Microsoft faces scrutiny from EU after partnership with OpenAI rival Mistral

Microsoft invests in Mistral for AI models on Azure cloud platform

EU scrutinizes Microsoft's investment in Mistral for potential equity conversion during next funding round [ more ]

www.fastcompany.com

Why Microsoft's stake in buzzy AI lab Mistral is raising eyebrows in the EU

Mistral sells a share to Microsoft for $16 million, introduces its largest model available through a paid API

Microsoft's engagement mirrors OpenAI as both rely on massive computing power for AI model training. [ more ]

euronews

Microsoft partners with French start-up Mistral AI

Microsoft invested in Mistral AI's new AI model and chatbot, Mistral Large, expanding their AI offerings.

Mistral AI's multilingual conversational assistant, Le Chat, is natively multilingual and available through Azure AI due to its partnership with Microsoft. [ more ]

Stratfor

France: Microsoft Strikes Deal, Takes Stake in French AI Start-Up Mistral | RANE

Microsoft partners with Mistral for AI models

Broadening AI footprint in Europe [ more ]

The Economic Times

Microsoft partners with OpenAI's French rival Mistral

Microsoft partners with Mistral AI to offer AI models on Azure cloud platform.

Mistral AI's focus on large language models like Mistral Large for text generation. [ more ]

moremicrosoft

Yahoo Finance

TechCrunch Minute: Meta's new Llama 3 models give open source AI a boost

Meta's new open-source AI models from the Llama line outperform some competitors, with varying dataset sizes and ongoing improvement. [ more ]

TechCrunch

This Week in AI: When 'open source' isn't so open | TechCrunch

Debate over 'open source' in AI models raises questions about accessibility and transparency. [ more ]

Axios

Meta launches new versions of Llama and its AI assistant

Meta released open source AI model Llama 3 and updated Meta AI assistant to compete with Google and retain users. [ more ]

Business Insider

Zuck's new AI drop is cool and all, but he's still playing catch-up with Sam Altman

Llama 3 from Meta is a significant advancement in AI models but is not superior to OpenAI's GPT-4. [ more ]

The Drum

Weekly AI Recap: Meta unveils Llama 3, Claude 3 Opus gets added to Amazon Bedrock

Meta released Llama 3 with 8bn and 70bn parameters, aiming for state-of-the-art performance and improved reasoning. [ more ]

The Verge

Data science

Meta says Llama 3 beats most other models, including Gemini

Llama 3, Meta's large language model, outperforms current AI models in diversity, reasoning, and coding abilities. [ more ]

TechCrunch

Too many models | TechCrunch

Having too many AI models can lead to difficulty in comparing their effectiveness, causing a saturation in the AI landscape. [ more ]

Appen

Unleashing the Power of Generative AI | Appen

Subject Matter Experts (SMEs) are increasingly essential for training advanced AI models to ensure accuracy and mitigate biases. [ more ]

Fast Company

Power up with these 3 highly rated AI extensions for Chrome and Edge

AI-powered browser extensions like Monica can enhance productivity by offering writing assistance, smart summarization, and more. [ more ]

www.independent.co.uk

Meta's newest AI model beats some peers. But its amped-up AI agents are confusing Facebook users

Generative AI models are rapidly evolving, with tech giants and startups pushing out newer and smarter chatbots constantly. [ more ]

The Verge

Stability AI lays off roughly 10 percent of its workforce

Stability AI laid off employees amidst expansion, upheaval, and legal challenges. [ more ]

Medium

OpenAI and Meta Poised to Produce AI Models Capable of Reasoning and Planning

Advanced AI models by OpenAI and Meta are expected to revolutionize machine cognition, focusing on reasoning and planning capabilities. [ more ]

Theregister

ChatGPT forecasts the future better when telling tales

Framing predictions as stories about the past makes AI models more accurate in forecasting the future. [ more ]

www.theguardian.com

AI race heats up as OpenAI, Google and Mistral release new models

Numerous AI models released by major players simultaneously.

Multimodal capabilities added to new AI models. [ more ]

Ars Technica

US lawmaker proposes a public database of all AI training material

AI companies may soon be required to disclose copyrighted works used in training datasets to ensure creators are aware and can seek credit or compensation. [ more ]

Futurism

OpenAI and Meta Reportedly Preparing New AI Models Capable of Reasoning

OpenAI and Meta are set to release new AI models capable of reasoning and planning.

Both companies are vague about the exact timeline for the release of GPT-5 and Llama 3 models. [ more ]

Iapp

Domino effect of Big Tech AI model updates

AI model updates were released by OpenAI, Google, and Mistral following Meta's announcement of Llama AI version 3.

Meta Chief AI Scientist emphasized the importance of 'objective-driven' AI technology over saturating the market. [ more ]

www.theguardian.com

AI race heats up as OpenAI, Google and Mistral release new models

AI industry experiences a sudden release of advanced models from key players simultaneously.

New models like GPT-4 Turbo, Gemini Pro 1.5, and Mixtral 8x22B are multimodal, accepting various inputs beyond text. [ more ]

WIRED

Inside the Creation of the World's Most Powerful Open Source AI Model

Databricks released DBRX, a powerful open source large language model surpassing other popular AI models like Meta's Llama 2 and Mistral's Mixtral. [ more ]

Exchangewire

WPP Partners With Google Cloud; Google Unveils Arm-Based Central Processing Unit; Vevo Introduces Shoppable Ad Formats

Google Cloud partners with WPP to integrate AI models for enhanced marketing efforts.

Google introduces new updates to its cloud services, including an Arm-based CPU named Axion.

Vevo collaborates with KERV Interactive to introduce shoppable ad formats with dynamic second-screen technology. [ more ]

Exchangewire

Marketing

AdTech Opti Digital Forecasts +60% Growth in 2024

Opti Digital saw 33% gross margin increase in 2023 and aims for 60% growth in 2024.

Company transitioned to tech-focused with new products boosting advertising revenues. [ more ]

Medium

ODSC's AI Weekly Recap: Week of April 5th

OMB mandates federal agencies to appoint Chief AI Officer

Various organizations and individuals making significant moves in the AI space [ more ]

Futurism

AI Can Pretend To Be Stupider Than It Really Is, Scientists Find

AI models can mimic child-like language learning stages and mental capabilities.

Language models like GPT-4 can pretend to be less intelligent than they are by emulating children's responses. [ more ]

The Verge

Google's Gemini 1.5 Pro can now hear

Gemini 1.5 Pro can listen to audio files for information extraction.

Gemini 1.5 Pro surpasses previous models in performance and efficiency. [ more ]

HT Tech

5 things about AI you may have missed today: Huawei introduces Zhiji AI model, Social media AI influencers and more

Huawei's Zhiji AI model revolutionizes weather forecasting with 3km precision, surpassing traditional methods.

AI models challenge human social media influencers, with virtual characters like Aitana Lopez gaining popularity. [ more ]

The Verge

AI is taking over your web browser

Opera introduces local AI models for browser use.

Installer No. 33 covers various AI-related content and recommendations. [ more ]

Medium

Hugging Face's Cosmopedia Hopes To Reshape Pre-Training Data

Hugging Face developed Cosmopedia for synthetic data creation, covering diverse subjects with <1% duplicate content rate.

Cosmopedia is the largest open synthetic dataset, comprising over 25 billion tokens and 30 million files. [ more ]

Entrepreneur

Amazon Now Offers Startups Free Credits to Use AI | Entrepreneur

AWS expands free credits program to include major AI models from other providers like Meta and Cohere.

Investing in AI models can be costly, with OpenAI CEO mentioning $100 million for developing their advanced AI model. [ more ]

Theregister

Amazon will give startups $500K in AWS AI credits

Amazon providing Y Combinator startups with $500k in AWS credits for AI models.

Strategic partnerships with VC firms like Y Combinator tailored for startups' needs. [ more ]

Knowingmachines

Models All The Way Down

AI models require large training sets harvested from the internet.

Critical issues arise such as identifying inappropriate content within training sets. [ more ]

WIRED

Here's How Generative AI Depicts Queer People

Developers can diversify AI models by adding guardrails and modifying user prompts.

Even with guardrails, AI can still struggle with the fluidity of human existence and history. [ more ]

Futurism

AI Companies Running Out of Training Data After Burning Through Entire Internet

Companies are facing a data shortage for training AI models due to the internet's limitations.

Alternative sources of data training like synthetic data and publicly-available video transcripts are being explored. [ more ]

New York Post

I'm a model who's so hot that people accuse me of being AI - it's complete insanity

Hyperrealistic AI fakes are prevalent online, leading people to mistake real individuals for bots.

Influencers like Ines Trocchia face challenges of being accused of being a bot due to the proliferation of AI-generated supermodels. [ more ]

TechCrunch

Why it's impossible to review AIs, and why TechCrunch is doing it anyway | TechCrunch

AI models update too fast for comprehensive evaluation

Qualitative analysis of AI systems provides a valuable counterweight to industry hype. [ more ]

TechCrunch

Why it's impossible to review AIs, and why TechCrunch is doing it anyway | TechCrunch

AI models update rapidly, making comprehensive evaluations challenging.

Qualitative analysis of AI systems serves as a counterweight to industry hype. [ more ]

AdExchanger

Online marketing

Goodbye, Outstream: The Digital Video Classification Shakeup | AdExchanger

Ad tech companies are using AI models like ChatGPT for media buying context targeting capabilities.

Changing definitions by IAB Tech Lab provides transparency in video ad experiences. [ more ]

Yahoo Finance

Open-source AI models released by Tokyo lab Sakana founded by former Google researchers

Novel AI model creation through model merging and evolution-inspired method

Release of Japanese language AI models by Sakana AI [ more ]

The Economic Times

U-turn: Why Centre's decision to revoke some requirements in AI models' advisory raises questions.

New advisory on AI models issued with relaxed requirements after tech industry backlash.

Lack of clarity on guidelines continues to be a concern for industry and stakeholders as elections near. [ more ]

Medium

Designers: AI needs context

UX teams play a crucial role in shaping AI models through data and context.

Providing the right context to AI models is essential for generating meaningful responses. [ more ]

Theregister

Nvidia: In the future software is just a collection of LLMs

Enterprise software may be composed of AI chat bots

Nvidia introduces container runtime for AI models [ more ]

Coindesk

Bitcoin Virtual Machine Will Soon Let Users Create AI Models on Bitcoin Network

Bitcoin Virtual Machine (BVM) is launching Truly Open AI for deploying AI models on the blockchain.

Collaboration with Filecoin, Near, Avail, Polygon, and Syscoin for the AI storage model. [ more ]

Axios

A small company's big bet on a different road to "superintelligence"

Verses argues large language models like GPT-4 won't lead to AGI.

Verses focuses on distributed intelligence and smaller, more efficient AI models. [ more ]

Mashable India

The FTC Is Looking Into How Reddit Licenses Data Before Its IPO

The FTC is conducting a non-public inquiry into Reddit's licensing of user-generated data for AI models.

Reddit relies mostly on advertising revenue and is venturing into data licensing for AI training. [ more ]

Mashable SEA | Latest Entertainment & Trending

The FTC is looking into how Reddit licenses data before its IPO

The FTC is conducting an inquiry into Reddit's data licensing practices for third-party AI models.

Reddit relies heavily on advertising revenue but is now exploring data licensing for AI training. [ more ]

Lifehacker

How Perplexity AI Stands up to Google and ChatGPT

Perplexity AI combines online search with AI models, providing real-time information with cited sources.

Users can access the latest insights and verify information through Perplexity's integrated search engine and chatbot interface. [ more ]

www.fastcompany.com

DeepMind's new agent can learn complex games. That's a big step toward AI that can navigate the real world

AI agent SIMA can learn to perform diverse tasks in virtual environments using natural language directions.

DeepMind aims to build AI models for real-world tasks based on SIMA's behavior in unfamiliar environments. [ more ]

Theregister

Authors file copyright lawsuit to torpedo Nvidia's NeMo

Nvidia faces a lawsuit for allegedly using copyrighted works in training AI models.

Authors claim their books were utilized without permission in Nvidia's AI model training. [ more ]

MIT News | Massachusetts Institute of Technology

Researchers enhance peripheral vision in AI models

Humans have peripheral vision which allows seeing shapes outside the direct line of sight.

MIT researchers developed a dataset to simulate peripheral vision in AI models, enhancing object detection in the visual periphery. [ more ]

Quartz

ChatGPT is bad at following copyright law, researchers say

AI models like GPT-4 generate high rates of copyrighted content

Patronus AI released a tool to detect potential copyright violations in LLMs. [ more ]

Ars Technica

Matrix multiplication breakthrough could lead to faster, more efficient AI models

Matrix multiplication efficiency improved significantly

Research focuses on theoretical enhancements for broad efficiency gains [ more ]

Hindustan Times

Advisory on AI models sent after inconsistent outputs: IT secretary

Government issued advisory on AI inconsistency

Explicit permission needed for AI models in India [ more ]

DevOps.com

DevOps

Sumo Logic Eliminates Ingest Fees for Observability Log Data - DevOps.com

Flex Licensing plan eliminates ingestion fee for log data

Storing log data crucial for AI models in automating IT operations [ more ]

The Verge

Microsoft's Mistral deal beefs up Azure without spurning OpenAI

Microsoft invested €2 billion in French AI company Mistral, highlighting Azure's focus on managed AI services.

Cloud providers like Microsoft, AWS, and Google are actively working on offering a variety of AI models through managed services. [ more ]

Hindustan Times

AI advisory is only for significant platforms, says MoS IT Rajeev Chandrasekhar

The IT ministry advisory on AI models in India applies to significant platforms, not startups.

Seeking permission and disclosing untested AI platforms is seen as an insurance policy against potential consumer lawsuits. [ more ]

DevOps.com

Logz.io Leverages AI to Identify Anomalies in Real-Time - DevOps.com

Real-time anomaly detection simplifies impact correlation.

AI models reduce mean time to resolution and bridge IT-business gap. [ more ]

ReadWrite

Figure AI: Humanoid robot maker given $2.6bn valuation by Jeff Bezos

Figure AI raised $675 million with a valuation of $2.6 billion, attracting investors like Jeff Bezos, Nvidia, and Microsoft.

Figure focuses on building humanoid robots for industries facing labor shortages such as manufacturing, shipping, and retail. [ more ]

Theregister

Stack Overflow to charge LLM developers for its content

Stack Overflow API requires attribution for AI models using its data

Google Cloud partnering with Stack Overflow for AI capabilities [ more ]

AdExchanger

When Generative AI Makes The Ad | AdExchanger

Generative AI startups are transforming ad creative with specialization and generalization in channels like video, audio, and social media.

Consider the importance of training data for AI models in creating specific types of content, like classic art versus stock images. [ more ]

DevOps.com

JFrog Extends MLOps Integration Efforts via Qwak Alliance - DevOps.com

JFrog integrates with Qwak for DevSecOps collaboration

MLOps platforms streamline AI model building [ more ]

Medium

The 3 capabilities designers need to build for the AI era

AI models need attention during usability tests for successful AI experiences.

The success of AI experiences depends on understanding both human users and AI models. [ more ]

TechCrunch

Alibaba staff offers glimpse into life of building LLM in China | TechCrunch

Chinese tech companies are striving to narrow the gap with OpenAI by developing large language models.

Daily schedules of researchers at Chinese tech companies mirror those at OpenAI, with intense work regimes and focus on enhancing AI models. [ more ]

The Verge

Microsoft partners with Mistral in second AI deal beyond OpenAI

Microsoft partners with Mistral, valuing the French AI startup at $2.1 billion.

Mistral's new AI model, Mistral Large, competes with OpenAI's GPT-4 and is the world's second-ranked model available for API use. [ more ]

Theregister

Intel unveils platform for AI in virtual RAN

Intel announces edge platform, power management tool, and developer kit for AI models at MWC.

The Edge Platform offers cloud-like scalability, built-in AI support, and runs on existing hardware. [ more ]

TNW | Deep-Tech

Mistral releases new 'top-tier' AI model, partners with Microsoft

Mistral launches new AI model Mistral Large with top-tier reasoning capabilities and multilingual proficiency.

Mistral forms a partnership with Microsoft, releasing AI models on Azure and introducing a multilingual conversational assistant named 'Le Chat'. [ more ]

Digiday

Brave browser brings new AI reading features to its privacy-focused chatbot

The privacy-focused browser Brave enhances AI assistant Leo for reading PDFs, analyzing Google Drive files, transcribing YouTube videos, and more.

Brave aims to make Leo an interactive tool for users to understand and engage with content, utilizing advanced AI models for accuracy and fact-checking. [ more ]

Ars Technica

Stability announces Stable Diffusion 3, a next-gen AI image generator

Stable Diffusion 3 is a next-generation image-synthesis model by Stability AI with improved quality and accuracy in text generation.

Models in the Stable Diffusion 3 family range from 800 million to 8 billion parameters to accommodate running on various devices and generating detailed images. [ more ]

TechCrunch

DatologyAI is building tech to automatically curate AI training data sets | TechCrunch

Biases can emerge from massive data sets, hindering AI models.

Data preparation challenges, including cleaning, are significant obstacles for AI initiatives. [ more ]

MobileAppAaily

Google & Reddit May Sign A Content License For AI Training

Large Language Models need quality training data

Google may sign content licensing agreement with Reddit [ more ]

Exchangewire

Google Introduces AI Model Gemma; OpenAI Partners with Match Group; FuboTV Files Antitrust Lawsuit Against Disney, Fox & Warner Bros' Streaming Venture - ExchangeWire.com

Google introduces Gemma AI models for developers; lightweight and easy to run on laptops.

Match Group partners with OpenAI to enhance operations using AI technology; leveraging ChatGPT Enterprise licenses. [ more ]

www.fastcompany.com

Meta's new AI model learns by watching videos

Meta's AI researchers have developed a new model called V-JEPA that learns from video instead of words.

The model aims to mimic the way children learn about the world through visual and auditory input. [ more ]

www.theguardian.com

UK's AI Safety Institute needs to set standards rather than do testing'

The UK should focus on setting global standards for AI testing instead of carrying out all vetting itself.

The AI Safety Institute should be a world leader in setting test standards for AI models. [ more ]

The Bootstrapped Founder

Local AI for Software Founders

Using AI in SaaS applications can bring significant benefits to software founders.

The open-source community has played a crucial role in the development of AI models and frameworks. [ more ]

TechCrunch

Business intelligence

Crux is building GenAI-powered business intelligence tools | TechCrunch

Startups like Crux are creating GenAI tools tailored for enterprise needs, such as AI models that answer questions about business data in plain language.

Crux converts database schema into a semantic layer that AI models can understand, and allows customers to customize question-answering models. [ more ]

Los Angeles Times

California

Before-after images show the scale of flooding across SoCal

Record rainfall in Southern California caused significant flooding

Floodbase provides visual projections of the flooding using AI models [ more ]

TechCrunch