• Posts
  • Open Source AI's Impact on Image Generation

Open Source AI's Impact on Image Generation

PLUS: ChatGPT Homescreen Widget, Microsoft's Opening Access to AI Models for Global Development and more.


Open Source AI produces STUNNING Images | NSFW, Stable Diffusion 3, DashToon, Magnific & more.

In the recent AI developments, Sora and Google's image model made waves. Sora creates lifelike videos based on prompts, stirring attention in Hollywood and AI circles.

Open-source AI, like Stability AI, offers alternatives. Stability AI CEO, Iman Mustak, addresses concerns raised by Forbes and focuses on progress. Stable Diffusion, an open-source AI, promises advancements like video generation and upscaling. DashToon Studio, another open-source tool, aids comic creation. Overall, open-source AI offers flexibility and control, bridging gaps in AI content creation.

Microsoft partners with Mistral in second AI deal beyond OpenAI

Microsoft teams up with French AI startup Mistral, valued at $2.1 billion, for a multiyear partnership. Mistral's AI models, including Mistral Large, akin to OpenAI's GPT-4, will now be available on Microsoft's Azure AI platform. Unlike before, Mistral's models won't be open source, signaling a shift towards commercialization. 

Microsoft's investment in Mistral follows a turbulent period for its main AI partner, OpenAI, offering the tech giant more visibility into AI development without voting power. This move expands Microsoft's AI portfolio and strengthens its foothold in the evolving AI landscape.

MWC 2024: Microsoft to open up access to its AI models to allow countries to build own AI economies

At Mobile World Congress 2024, Microsoft laid out new guidelines for AI governance, emphasizing accessibility to its technology for global developers. Brad Smith, Microsoft's president, stressed the importance of democratizing AI and making development tools widely available. The company aims to expand access through its cloud computing infrastructure, fostering AI economies worldwide. 

Additionally, Microsoft partnered with Mistral AI, a French startup, to advance its AI market presence. This move aligns with Microsoft's commitment to open-source models and substantial investments in European AI infrastructure. Overall, Microsoft's initiatives signal a significant step towards democratizing AI technology on a global scale.

Mistral AI releases new model to rival GPT-4 and its own chat assistant

Paris-based startup Mistral AI unveils Mistral Large, a powerful language model challenging GPT-4. The company also introduces its chat assistant, Le Chat. Mistral AI, founded by ex-Google and Meta alumni, shifts from open source to a paid API model, competing with OpenAI. 

Mistral Large is cheaper than GPT-4 Turbo. The company collaborates with Microsoft to distribute its models via Azure. Le Chat offers free access with plans for enterprise versions. Mistral AI aims to rival industry giants in AI while navigating evolving pricing and performance landscapes.

ChatGPT gets a homescreen widget for Android

Tech enthusiast Mishaal Rahman reports an update for ChatGPT on Android, introducing a homescreen widget. This allows faster access to the chatbot's features like conversation and query modes.

To access, simply long-press an empty space on the phone's home screen. While it seems to be in beta, the latest version from Google Play has a lower version number.

SK Telecom partners with AI search startup Perplexity in Korea

SK Telecom, South Korea's top telco, teams up with U.S. AI startup Perplexity to offer users an AI search engine alternative to Google and Naver. The partnership grants SKT subscribers free access to Perplexity's paid version and enables SKT to utilize Perplexity's AI for personalized assistant products. The collaboration's specifics, including rollout timing, are pending. Perplexity's CEO sees it as mutually beneficial, leveraging SKT's distribution channels. 

Perplexity, challenging Google, employs generative AI models and recently secured significant investment. SKT's move aligns with its strategy of engaging with Silicon Valley startups for cutting-edge AI tech, following a previous investment in Anthropic.

NVIDIA RTX 500 and 1000 Professional Ada Generation Laptop GPUs Drive AI-Enhanced Workflows From Anywhere

NVIDIA launch RTX 500 and 1000 Ada Generation Laptop GPUs, catering to professionals needing AI-accelerated power on the go. These GPUs integrate neural processing units and Tensor Cores for efficient AI tasks. They significantly enhance AI-based workflows, offering up to 14x generative AI performance and 10x graphics performance compared to CPUs. 

Professionals, including video editors and graphic designers, benefit from streamlined tasks and higher-quality outputs. The GPUs boast advanced features like DLSS 3 and AV1 encoder support. Available this spring in laptops from major manufacturers, they promise superior performance and portability for diverse professional applications.

Intel unveils Sierra Forest and Granite Rapids-D Xeon processors to support ‘AI Everywhere’

Intel has announced new processors, Sierra Forest and Granite Rapids-D, alongside an Edge Platform, aimed at bolstering AI capabilities and enhancing infrastructure. Granite Rapids-D, slated for 2025, targets virtual radio access network workloads, promising significant performance gains. Sierra Forest, set for later this year, boasts up to 288 cores, promising a 2.7-times performance boost for 5G workloads. The Edge Platform facilitates AI deployment at the network edge, offering cloud-like simplicity and enabling real-time AI inferencing optimization. 

With partnerships with major players and a focus on AI-driven modernization, Intel aims to make "AI Everywhere" a reality in 5G, edge, and enterprise infrastructures.


"Genie" is a groundbreaking project introducing the first generative interactive environment trained from internet videos without supervision. With 11B parameters, it forms a foundation world model, enabling users to prompt diverse virtual worlds through text, images, and sketches. It allows interactive exploration and training of agents without specific labels.

"Orca-Math" introduces a 7-billion-parameter small language model (SLM), achieving 86.81% accuracy on the GSM8k math problem-solving benchmark without multiple model calls or external tools. It utilizes a synthetic dataset and iterative learning techniques, outperforming larger models like LLAMA-2-70B and ChatGPT-3.5 while using smaller data.

"MobileLLM" addresses the demand for efficient large language models (LLMs) on mobile devices by designing high-quality models with fewer than a billion parameters. By focusing on model architecture, leveraging deep and thin designs, embedding sharing, and grouped-query attention mechanisms, MobileLLM achieves significant accuracy boosts over previous state-of-the-art models. Additionally, block-wise weight sharing further enhances accuracy without increasing model size or latency. MobileLLM demonstrates improvements in chat benchmarks and API calling tasks, showcasing the effectiveness of small models for on-device use cases.

"Gen4Gen" addresses challenges in personalizing text-to-image diffusion models for multiple concepts. It introduces a dataset creation pipeline, Gen4Gen, combining personalized concepts into complex compositions along with text descriptions. A comprehensive metric evaluates performance, leading to improved multi-concept personalized image generation without altering model architecture or training algorithms.

This study investigates the impact of input length on Large Language Models (LLMs) reasoning performance. Using a novel QA reasoning framework, it reveals notable degradation in performance at shorter input lengths than expected. Traditional perplexity metrics do not correlate with long input reasoning tasks. The findings highlight important considerations for future LLM research.


Graphy - a tool designed to quickly visualize and share data. It offers an intuitive interface for creating dashboards, reports, and charts, catering to users who need to analyze and present data effectively. 

Dopt - AI-powered platform for in-app assistance, enhancing user success with personalized and contextual experiences. 

TechInterviewer AI - AI-driven platform for conducting technical interviews, offering realistic simulations to evaluate candidates' skills effectively and efficiently.

VectorShift -  AI platform offering end-to-end automations for businesses, streamlining processes like data analysis, decision making, and workflow management with AI technology.

PhotoRoom - AI photo editor that simplifies photo editing. It offers background removal, object removal, instant background creation, and batch editing. 



The 9-Month-Old AI Startup Challenging Silicon Valley’s Giants

Arthur Mensch, a former Google employee, launched an AI startup a year ago in Paris. The company specializes in generative AI, valued at over $2 billion. Mensch's venture challenges tech giants like Microsoft, showcasing AI's rapid evolution from sci-fi to mainstream tech. THE WALL STREET JOURNAL

Former Twitter engineers are building Particle, an AI-powered news reader

Particle.news, founded by ex-Twitter engineers, introduces an AI-powered news reader. The startup aims to offer personalized news summaries while compensating publishers fairly. Concerns arise about AI-summarized news reducing website traffic for publishers. Backed by notable investors, Particle strives to revolutionize news consumption amidst a changing media landscape. TECHCRUNCH

Former Google video generation expert joins TikTok owner ByteDance amid generative AI race

Former Google video generation expert Jiang Lu has joined ByteDance, the parent company of TikTok, amidst the race to advance generative AI tools. This move comes after OpenAI's Sora made waves in AI competition. Tech giants are ramping up efforts in the next generation of applications, with ByteDance investing in talent like Jiang Lu. SCMP

Google puts controversial Gemini AI into Messages

Google introduces Gemini, its controversial AI assistant, to its Messages app despite recent backlash. Users can now chat with Gemini directly within the app, though it's currently only available on Android phones for beta testers. This move comes amidst Google's broader focus on AI enhancements in its Android updates, including features for Android Auto and accessibility improvements. INDEPENDENT

UK start-up raises $14mn to develop AI-boosted adaptable vaccines

A UK start-up has secured $14 million in funding to develop AI-enhanced adaptable vaccines. The company aims to leverage artificial intelligence to create vaccines that can adapt to new variants and emerging threats. This investment underscores the growing interest in AI-driven solutions for healthcare challenges. FINANCIAL TIMES

Meta to set up team to counter disinformation, AI abuse in EU elections

Meta, owner of Facebook, will establish a team to combat misinformation and misuse of AI ahead of European Parliament elections in June. Concerns about election interference and deceptive AI content prompt action. The team will monitor threats and expand fact-checking partnerships to safeguard election integrity. REUTERS

Google expects its more advanced large language models to arrive on Android phones next year

Google anticipates integrating its advanced large language models (LLMs) into Android phones from next year, enhancing on-device AI capabilities. Currently, Google offers Gemini Nano on compatible Android devices. Brian Rakowski, from Google’s Pixel unit, foresees these more sophisticated AI models being accessible directly on smartphones, a move likely to reshape mobile technology. CNBC

Microsoft Is Using AI to Stop Phone Scammers From Tricking You

Microsoft introduces Azure Operator Call Protection, a new AI-powered service aimed at detecting and alerting users to suspicious phone interactions, particularly fraudulent calls. Piloted with BT Group, the service analyzes conversations in real time, identifying potential scams and empowering users to make informed decisions during calls. CNET

DeepMind Chief Says Google’s Bungled AI Faces Feature Is Returning Soon

Google's AI chief, Demis Hassabis, announces plans to reintroduce a paused AI feature that generates images of people "in the next couple of weeks." The decision follows criticism over the feature's inaccurate depiction of race. Hassabis acknowledges the challenge of creating a "multi-modal" system. BLOOMBERG

Disney activist Blackwells demands AI strategy to boost shares

Disney shareholder Blackwells Capital urges an AI strategy, suggesting it could boost Disney's stock by 129%. Blackwells proposes potential changes including a breakup and spinning off park assets. Disney refutes, highlighting its long-term tech investments and commitment to shareholder value. Blackwells seeks three board seats to drive technological transformation. REUTERS

AI chatbots show human-like personality and decision-making traits, study finds

The study in PNAS assesses AI chatbots' human-like behaviors and personality traits using the OCEAN Big Five questionnaire and behavioral games. ChatGPT-4 closely mirrors human scores across personality dimensions and exhibits altruism and cooperation akin to humans. Context influences AI decisions, suggesting their potential to complement human abilities with proper training. NEWS MEDICAL LIFE SCIENCES

China’s Oppo unveils prototype augmented reality glasses with voice assistant as AI hype hits gadgets

At the Mobile World Congress, Oppo unveiled the prototype Oppo Air Glass 3, augmented reality glasses with a voice assistant, emphasizing the integration of AI across devices. Tethered to Oppo smartphones, these glasses feature touch sensors for control and AndesGPT-powered voice assistance for tasks like travel planning. While commercialization remains uncertain, Oppo aims to demonstrate its tech capabilities in a potentially burgeoning market. CNBC

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.