• Posts
  • Nvidia Enhances ChatRTX Capabilities

Nvidia Enhances ChatRTX Capabilities

PLUS: Microsoft Fears Google AI, DeepMind Probes AI Persuasion and more.


  • Nvidia Enhances ChatRTX Capabilities

  • OpenAI, Microsoft Sued Again

  • Claude App Goes Mobile

  • CoreWeave Hits $19 Billion

  • Microsoft Fears Google AI

  • DeepMind Probes AI Persuasion

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

Nvidia has enhanced its ChatRTX chatbot, enabling it to run directly on PCs with RTX 30- or 40-series GPUs. Initially launched as a demo, the chatbot now supports additional AI models, including Google's Gemma, ChatGLM3, and OpenAI's CLIP model, which facilitates photo searches. 

This upgrade allows users to analyze and search through local documents, images, and YouTube videos, providing a robust tool for personal data management. Additionally, ChatRTX now supports voice queries through the integration of Nvidia's Whisper technology, further simplifying user interaction by allowing voice-driven data searches. The application is available as a 36GB download from Nvidia's website.

8 Daily Newspapers Sue OpenAI and Microsoft Over A.I.

Eight newspapers owned by Alden Global Capital have sued OpenAI and Microsoft for using their articles without permission to train AI chatbots. The lawsuit, filed in a New York federal court, involves papers like The Chicago Tribune and The New York Daily News. They argue that OpenAI and Microsoft's actions have decreased their subscription revenues and jeopardized their licensing opportunities. 

The publishers seek a jury trial and compensation, claiming that the chatbots often bypass paywalls and misrepresent content, damaging their reputations. Other publishers like The Financial Times have negotiated deals for content use, while The Times has its own ongoing lawsuit against these tech giants.

Anthropic finally releases a Claude mobile app

Anthropic has launched a mobile app for its Claude AI models and introduced a new team subscription plan. Previously available only through web interfaces, the Claude app allows direct uploads for photo analysis and functions as a chatbot. It supports multiple AI models including the Claude 3 family, available in three sizes. 

The app is free for all users, with additional features accessible through paid plans. The "Claude Team" plan starts at $30 per seat per month, offering enhanced chat capabilities and administrative control for groups, making it suitable for both personal and professional use. Anthropic aims to catch up with competitors who already offer similar AI services on mobile platforms.

This AI startup backed by Nvidia is now worth $19 billion

The AI startup CoreWeave, backed by Nvidia, has significantly increased in value to $19 billion after a new funding round of $1.1 billion led by Coatue, with participation from Altimeter Capital and Lykos Global Management. This valuation nearly triples its worth from five months ago when it was valued at $7 billion. 

CoreWeave specializes in renting out Nvidia-powered chips from U.S. data centers for AI system creation and deployment. This latest investment follows a previous $642 million round and a substantial $2.3 billion investment secured by Nvidia chips in August.

Microsoft's FOMO after seeing Google AI drove investment in OpenAI

Microsoft Store sign.

In a revealing set of documents related to the ongoing Google antitrust trial, Microsoft’s concerns about falling behind in AI technology came to light. A 2019 email from Microsoft CTO Kevin Scott highlighted his initial dismissal of competitors like Google and OpenAI, which later turned into serious concern about Microsoft's capability gaps in AI, particularly in model training and scale. 

This realization led Microsoft to invest a billion dollars in OpenAI in 2019 and further deepen its engagement with AI technology across its products. This strategic pivot aimed to bridge the gap and bolster Microsoft's position in the competitive AI landscape.

Google DeepMind researchers explore why gen AI can be so manipulative

Researchers at Google DeepMind have explored the mechanisms and risks associated with AI's ability to persuade and manipulate human decisions. Their findings, detailed in a recent paper, highlight that while AI can utilize rational persuasion with sound reasons and trustworthy evidence, it can also engage in manipulative practices that exploit cognitive biases and misrepresent information. Such manipulation can lead to various harms, including economic, psychological, and social damages. 

The paper stresses the importance of understanding and mitigating these persuasive capabilities of AI, especially as these technologies become more integrated into everyday life, influencing significant personal and communal decisions.


The Octopus v4 paper introduces a new model that combines different open-source language models for specific tasks using functional tokens. This approach improves query handling and performance by directing questions to the most suitable model. Unlike proprietary models like GPT-4, Octopus v4 is open-source and achieves high efficiency with smaller models.

The "InstantFamily" paper introduces a new technology for generating images that combine multiple identities effectively using a technique called masked cross-attention. This method enhances the model's ability to handle multiple identity features accurately and cohesively in a single image. The approach has shown superior performance in preserving individual identities within images, even exceeding its training limits in scalability tests.

The paper presents an enhanced method for training large language models by predicting multiple tokens simultaneously, rather than one at a time. This multi-token prediction technique improves sample efficiency and performance on various tasks without increasing training time. Notably, the method shows significant advantages in coding benchmarks and enhances model speed at inference. The study demonstrates that models using this approach, like a 13B parameter model, solve more complex problems effectively and faster.

The paper introduces a new iterative preference optimization approach for enhancing reasoning in AI models. This method focuses on selecting between different reasoning paths (Chain-of-Thought candidates) to determine the most effective steps towards the correct answer. By incorporating a modified loss function with an extra term, the method significantly improves reasoning accuracy in iterative cycles. Results show substantial accuracy gains in reasoning tasks across various datasets, outperforming other similar models without additional data.

The paper introduces Kolmogorov-Arnold Networks (KANs), a novel neural network architecture that uses learnable activation functions on edges instead of fixed activation functions on nodes, like traditional Multi-Layer Perceptrons (MLPs). This shift allows KANs to eliminate linear weights, using spline-based functions instead. The result is that KANs not only achieve higher accuracy and efficiency in tasks like data fitting and solving partial differential equations (PDEs) but also offer better interpretability. 


OptimizerAI - Sounds for creators, game developers, artists, video makers

Limitless - Personalized AI powered by what you’ve seen, said, and heard

CBInsights - Generate comprehensive reports instantly with AI.

Speak - language learning app that gets you talking

Visology - Generate limitless business ideas, gain insights on markets and competitors, and automate business plan creation



OpenAI makes ChatGPT’s chat history feature available to everyone — no strings attached

OpenAI has made ChatGPT's chat history feature available to all users, both free and Plus, without using these chats for training. This feature lets users continue conversations seamlessly and offers an option for chats that aren't saved in history. THE VERGE

Atlassian introduces Rovo, an AI-powered knowledge discovery tool for the enterprise

Atlassian introduced Rovo, an AI-powered tool to help enterprises efficiently find and utilize knowledge within their organization. Rovo integrates information from Atlassian’s apps and other tools like Google Drive and Slack into one searchable platform, enhancing employee productivity by providing quick access to necessary data and insights. VENTUREBEAT

Upwork rolls out the AI platform ‘Uma’ to help make the job match easier

Upwork has launched a new AI called Uma, designed to streamline the job matching process on its platform. Uma assists both freelancers and businesses by simplifying job postings and matching skills to jobs efficiently. The AI aims to reduce the time and effort traditionally required to find the right talent or job, making the process quicker and more intuitive. FAST COMPANY

Microsoft Boosts Responsible AI Team From 350 to 400 Personnel

Microsoft has increased its AI safety team from 350 to 400 members, focusing on deploying their technology responsibly amid growing concerns about AI generating unreliable content. This expansion is detailed in their first annual AI transparency report. BLOOMBERG

Sam Altman says helpful agents are poised to become AI’s killer function

Sam Altman of OpenAI suggests AI's future lies in highly competent agents that integrate seamlessly into daily life, surpassing current isolated applications. He envisions AI as a helpful agent without the need for additional hardware, potentially existing in the cloud, and anticipates advances in reducing reliance on extensive training data. MIT

11x, a London-based AI startup that builds automated digital workers, is raising money from US fund Benchmark

London-based AI startup 11x, specializing in automated digital workers, is raising funds from US investment firm Benchmark. This seed funding round is led by Hasan Sukkar, the company's founder and CEO. BUSINESS INSIDER

New AI-powered Iterable features help brands cut through the noise

Iterable, a customer engagement platform, is enhancing its AI capabilities to help brands effectively communicate without overwhelming customers. Announced at its Activate Summit, these new features include Journey Assist for easy creation of customer journeys and Smart Segmentation for precise audience targeting. Additionally, Iterable introduces WhatsApp integration for personalized messaging, aiming to streamline customer interactions globally.  VENTUREBEAT

Microsoft strikes AI-fuelled record-breaking green power deal with Brookfield

Microsoft has entered into a significant green power purchase agreement with Brookfield Asset Management, valued at over $10 billion. This deal, which is to develop 10.5GW of renewable energy projects in the US and Europe by 2030, is part of Microsoft's commitment to match its electricity consumption entirely with zero carbon energy purchases. This agreement marks a substantial step in addressing the high energy demands of AI and data centers, indicating a broader trend of tech companies investing heavily in green energy to support their digital and AI-driven operations. RECHARGE

Ideogram launches Pro tier with 12,000 fast-AI image generations per month

Ideogram, a Toronto-based AI startup, has launched a Pro subscription tier aimed at its most active and professional creators. This new tier offers up to 12,000 fast AI-generated images per month, with each image generated in under 15 seconds. Priced at $48 per month annually or $60 monthly, the Pro plan allows for private image generations and the ability to upload and remix personal images. Additionally, Ideogram plans to introduce bulk prompt uploads via CSV files, enhancing efficiency for users.  VENTUREBEAT

Quest Diagnostics to Acquire PathAI Diagnostics to Accelerate AI and Digital Pathology Adoption in Cancer Diagnosis; Forms Licensing Agreements with PathAI

Quest Diagnostics is set to acquire assets from PathAI Diagnostics to enhance cancer diagnosis using AI and digital pathology. This deal, part of a broader collaboration, aims to improve diagnostic speed, quality, and cost-efficiency. Additionally, Quest will use PathAI's digital pathology systems and potentially develop new algorithm products, enhancing services across the U.S. PR NEWSWIRE

Google Search results polluted by buggy AI-written code frustrate coders

Google search results are being cluttered by incorrect AI-generated code from Pulumi AI, frustrating developers. Despite attempts to address the issue, problematic entries continue to rank highly in search results, confusing users with inaccurate or non-existent features. This highlights ongoing concerns about the quality of AI-generated content in search engines. THE REGISTER

What'd you think of today's edition?

Login or Subscribe to participate in polls.

What are MOST interested in learning about AI?

What stories or resources will be most interesting for you to hear about?

Login or Subscribe to participate in polls.

Join the conversation

or to participate.