Stability AI: Diffusion Design using DALL-E3

Stability AI Unveils Next-Gen Image Solutions: Business APIs & Innovative Features

Stability AI, a leader in generative artificial intelligence (GenAI) that creates and deploys its proprietary diffusion models to create realistic images using prompt texts, has unveiled significant enhancements to its text-to-image products. These advancements encompass enterprise-grade APIs, new image enhancement capabilities, and fine-tuning features, marking a significant advancement in creative storytelling and professional image manipulation.

Features

  1. Sky Replacer: This tool allows users to alter the sky in their photos, offering nine different sky options. It’s particularly beneficial for industries like real estate.
  2. Stable 3D Private Preview: An automatic process for generating concept-quality textured 3D objects, simplifying the creation of 3D content.
  3. Stable FineTuning Private Preview: Provides the capability to fine-tune pictures, objects, and styles rapidly, useful for industries reliant on visuals.
  4. Enterprise-Grade APIs: These APIs facilitate the integration of Stability AI’s advanced image generation and manipulation capabilities into business applications, allowing for seamless and scalable use of these tools in various professional settings.
  5. Content Credentials and Invisible Watermarking: Ensures authenticity and responsible use of AI-generated content.

Benefits

  • Enhanced Visual Appeal: Tools like Sky Replacer enhance the aesthetic quality of images.
  • Time and Resource Efficiency: Features like Stable 3D reduce the time and complexity involved in creating 3D content.
  • Customization and Creativity: Stable FineTuning allows for a high degree of customization in image transformation.
  • Scalability and Integration: The enterprise-grade APIs enable businesses to integrate advanced image generation and manipulation capabilities into their existing systems and workflows.
  • Authenticity and Responsibility: The integration of Content Credentials and invisible watermarking addresses the need for authenticity in AI-generated content.

Technical Details

  • Compatibility: The generated 3D objects are compatible with popular 3D tools and game engines.
  • Access: These features are currently in private preview, with access requests available through Stability AI’s contact page.
  • Industry-Specific Solutions: Stability AI is focusing on developing solutions tailored to specific industries, starting with Sky Replacer for real estate.

Conclusion

Stability AI’s enhanced image offerings, including enterprise-grade APIs and fine-tuning capabilities, represent a significant leap in the capabilities of generative AI for image and 3D content creation. These advancements streamline creative processes, offer new avenues for customization, and ensure responsible use of AI in various professional domains.

Microsoft Phi 1.5: Diffusion Design using Ideogram

Microsoft’s Phi 1.5 AI Model Advances with Enhanced Efficiency and Capabilities

Microsoft Research has made a significant leap in artificial intelligence by upgrading one of its smaller large language models, Phi 1.5, to possess multimodal capabilities, enabling it to interpret images. This advancement positions Phi 1.5 as a more economical alternative to OpenAI’s GPT-4, which is known for its extensive processing power and energy requirements. The open-source nature of Phi 1.5 democratizes access to advanced AI technology, potentially easing the demand for high-end graphics processors.

Features:

  • Multimodal Functionality: Phi 1.5 now shares the image interpretation ability of GPT-4, a feature that was previously exclusive to larger models.
  • Open Source: Phi 1.5 is freely available, allowing for widespread use and innovation without associated costs.
  • Parameter Efficiency: With only 1.3 billion parameters compared to GPT-4’s 1.7 trillion, Phi 1.5 operates with far fewer computational resources.

Benefits:

  • Cost-Effectiveness: The smaller size of Phi 1.5 translates to lower operational costs, making it an attractive option for companies and individuals who require AI capabilities without the financial burden.
  • Energy Efficiency: Reduced computational requirements mean that Phi 1.5 is less energy-intensive, aligning with sustainability goals and reducing greenhouse gas emissions.
  • Accessibility: As an open-source model, Phi 1.5 can be utilized by a broader range of users, from individual developers to large enterprises, fostering innovation and inclusivity in AI development.

Technical Details:

  • Scale Comparison: Phi 1.5’s parameters are akin to a footlong sub sandwich when compared to the Empire State Building-sized GPT-4, illustrating the stark contrast in model sizes.
  • Operational Efficiency: The smaller model requires less powerful processors and shorter response times, making it suitable for a variety of tasks.
  • Complementary Use: Microsoft researchers envision a future where small models like Phi 1.5 work alongside larger models, each serving different operational regimes and tasks.

Conclusion: Microsoft’s enhancement of Phi 1.5 to include multimodal capabilities is a testament to the potential of smaller AI models to perform complex tasks efficiently. This development not only offers a cost-effective solution for AI applications but also paves the way for responsible and sustainable AI use. As the AI research community continues to evolve, the proliferation of such capable and accessible models is likely to transform the landscape of AI technology and its application across industries.

Other AI News

  • OpenAI has introduced new beta features for ChatGPT Plus members, enhancing the tool’s capabilities.

These updates include the ability to upload and work with files, along with multimodal support, which allows the system to intuitively understand user requests without the need to select specific modes. This enhancement brings some of the functionalities of the ChatGPT Enterprise plan to individual subscribers.

One of the notable features is the Advanced Data Analysis, which enables OpenAI‘s ChatGPT to process uploaded files, summarize data, answer questions, and create data visualizations based on user prompts. The tool is not restricted to text files; it can also interact with images. For instance, users have successfully uploaded images and requested ChatGPT, in conjunction with DALL-E 3, to create modified versions of these images, showcasing its versatility.

  • Cowbell Secures Additional $25 Million Investment to Fuel Continued Growth and Expansion

Cowbell, a company specializing in cyber threat monitoring and insurance, has recently raised an additional $25 million from Prosperity7 Ventures, a subsidiary of Aramco Ventures. This funding comes amidst Cowbell’s impressive 49% year-over-year growth. The company, formerly known as Cowbell Cyber, offers a suite of products designed to meet the diverse needs of enterprises, ranging from small and medium-sized businesses to large multinational corporations.

The company’s adaptive cyber insurance aligns coverage and pricing with an organization’s evolving cyber risk profile, facilitated by continuous, automated risk assessment. Cowbell provides three main insurance plans: Cowbell Prime 100 for companies with up to $100 million in annual revenue, Cowbell Prime 250 for enterprises with up to $500 million in annual revenue, and Cowbell Prime Plus for multinational corporations. Utilizing AI and machine learning algorithms, Cowbell monitors over 38 million enterprises, processing vast amounts of data to assess and manage cyber risks effectively.

With the new investment, Cowbell aims to achieve operating profitability, focusing on profitable growth in the US and UK markets. The company plans to enhance its market differentiation and improve services for brokers and customers. Cowbell’s innovative approach to cyber risk management, including its effective reduction of ransom payments to an average of 26% of the initial demand, positions it as a leader in the cyber insurance industry.

  • Snowflake Introduces Cortex: A Comprehensive Managed Service for Developing LLM Applications in the Data Cloud

Snowflake, a Montana-based data-as-a-service and cloud storage company, has unveiled Cortex, a fully managed service designed to integrate large language models (LLMs) into its data cloud. Announced at the company’s annual Snowday event, Cortex provides enterprises with AI building blocks, including open-source LLMs, to analyze data and build applications for various business-specific use cases. Sridhar Ramaswamy, SVP of AI at Snowflake, highlighted that Cortex enables businesses to quickly tap into large language models, build custom LLM-powered apps, and maintain control over their data.

Cortex aims to simplify the process of building LLM applications, addressing challenges such as the need for AI talent and complex GPU infrastructure management. The service offers serverless specialized and general-purpose AI functions, accessible through SQL or Python code. These functions include language and machine learning models for tasks like data extraction, summarization, translation, forecasting, and anomaly detection. Additionally, Cortex provides vector embedding and search capabilities, allowing users to contextualize model responses based on their data and create custom applications.

Snowflake is already leveraging Cortex to enhance its platform with native LLM experiences. The company has launched three Cortex-powered capabilities in private preview: Snowflake copilot, Universal Search, and Document AI. These tools offer functionalities like conversational assistance, LLM-powered search, and information extraction from unstructured documents. Snowflake’s introduction of Cortex marks a significant step in making generative AI more accessible and functional for enterprises.

  • Google to Invest $2 Billion in OpenAI’s AI Rival Anthropic

Alphabet’s Google has agreed to invest up to $2 billion in Anthropic, an artificial intelligence company and rival to OpenAI. The initial investment is $500 million, with an additional $1.5 billion planned over time. This move underscores Google’s efforts to compete with Microsoft, a major backer of ChatGPT creator OpenAI, as tech giants race to integrate AI into their applications.

Amazon has also expressed interest in Anthropic, planning to invest up to $4 billion to compete in the AI space. According to Amazon’s recent report to the U.S. Securities and Exchange Commission, the company invested in a $1.25 billion note from Anthropic that can convert to equity. Furthermore, Amazon has the option to invest up to $2.75 billion in a second note, expiring in early 2024.

These investments reflect the strategic maneuvers of cloud companies to forge ties with AI startups reshaping the industry. Anthropic, co-founded by former OpenAI executives Dario and Daniela Amodei, is actively seeking resources and influential backers to compete with OpenAI and position itself as a technology sector leader.

  • Siemens and Microsoft Collaborate on AI Project to Enhance Productivity and Innovation

Siemens and Microsoft have announced a joint project, the Siemens Industrial Copilot, which aims to use artificial intelligence to enhance productivity and human-machine collaboration. This initiative will apply generative AI in sectors such as manufacturing, transportation, and healthcare. German automotive supplier Schaeffler AG is among the early adopters of this technology.

The project’s goal is to create AI copilots that assist staff in designing new products and managing production and maintenance. These AI systems will analyze data collected by Siemens to help customers quickly develop, refine, and troubleshoot complex automation codes, significantly reducing simulation times in factories and other facilities.

Schaeffler is already using generative AI for programming industrial automation systems and plans to employ the Siemens Industrial Copilot to reduce production downtimes. Siemens highlights that tasks which previously took weeks can now be completed in minutes, indicating a potential revolution in how companies design, develop, manufacture, and operate, as stated by Siemens CEO Roland Busch.

  • Elon Musk Advocates for Independent ‘Third-Party Referee’ in AI Development

Elon Musk, speaking at the inaugural AI Safety Summit in Britain, emphasized the need for a “third-party referee” to oversee companies developing artificial intelligence. He suggested that this independent entity should have the capacity to sound the alarm if there are concerns about the AI being developed. Musk highlighted the importance of having insight into AI development before implementing oversight.

Musk’s remarks followed the publication of a declaration by Britain, signed by 28 countries and the European Union. This declaration outlines a dual agenda: identifying shared concerns related to AI risks, enhancing scientific understanding of these risks, and developing cross-country policies to mitigate them.

Addressing potential government intervention in AI regulation, Musk expressed concern that governments might prematurely impose rules without a thorough understanding of AI. However, he believes such premature regulation is unlikely to happen.

  • Dell and Meta Collaborate to Deliver On-Premises Llama 2 AI to Enterprise Users

Dell Technologies has announced its support for the open-source Llama 2 large language model (LLM) developed by Meta, marking a significant step in bringing this technology to enterprise users for on-premises deployments. This partnership differs from previous cloud provider support, as it focuses on on-premises applications. Dell is not only supporting Llama 2 for its enterprise users but is also utilizing it for its use cases.

The collaboration aims to bring AI closer to enterprise data, with Dell’s Matt Baker, senior vice president of AI strategy, emphasizing the importance of having sophisticated AI models like Llama 2 run on-premises alongside company data. This integration is expected to enable the development of powerful applications. Dell is guiding its enterprise customers on the necessary hardware for deploying Llama 2 and assisting in building applications that leverage the open-source LLM.

For Meta, the partnership with Dell provides valuable insights into how enterprises use Llama, aiding in the expansion of Llama’s capabilities. Joe Spisak, head of generative AI open source at Meta, highlights the importance of on-premises deployment options for considerations like data privacy. The partnership is expected to enhance the Llama development community’s understanding of enterprise requirements, contributing to the development of future Llama models and a safer, more open AI ecosystem.

  • Canva Introduces ‘Classroom Magic’ AI Tools for the Education Sector

Canva, the Australian company known for its cloud-based graphic design tools, is expanding its AI offerings with the launch of “Classroom Magic,” a version of its Magic Studio tailored for the education sector. This new feature, part of the Canva for Education product launched in 2019, brings AI tools to over 50 million students and teachers globally. Classroom Magic aims to assist teachers in creating engaging content and help students embrace creativity.

Classroom Magic includes AI tools like “Magic Write,” which offers quick actions such as summarizing text, expanding short text, rewriting, and changing tone. “Magic Animate” enables the transformation of static text into moving text and adds transitions to presentations. “Magic Grab” detects elements within an image, allowing users to manipulate them, while “Magic Switch” transforms projects across formats. Canva emphasizes that these tools are designed to develop comprehension skills and encourage creativity.

Addressing concerns about AI in the classroom, Canva has introduced Canva Shield for Education, providing strict controls to ensure safety. This includes advanced educator controls, automatic reviews, blocked terms, and reporting options. Canva’s survey of U.S. teachers indicates a strong interest in AI, with benefits such as boosting productivity, creativity, and personalized learning. Jason Wilmot, Canva’s Head of Education, emphasizes the goal of creating a safe environment for learning about AI capabilities within Canva.

  • AMD forecasts $2 billion sales of AI chips

Advanced Micro Devices (AMD) has forecasted that its new chip designed to compete in the artificial intelligence (AI) market will generate $2 billion in sales by 2024. The chip, MI300X, is positioned to rival Nvidia‘s offerings in the data center AI chip sector. This projection was made despite AMD’s quarterly revenue and gross margin estimates being below expectations, affected by a weak gaming market and a decline in demand for programmable chips in some industries.

During a conference call, AMD’s CEO Lisa Su announced that several significant hyperscale customers, typically large tech and cloud computing firms, have committed to using the MI300 chips. The company has adjusted its fourth-quarter revenue expectations for the chip to $400 million, up from the $300 million previously forecasted in August.

Despite the optimistic outlook for the MI300 chip, AMD acknowledges challenges in other areas of its business. The PC market is experiencing a slow recovery, and there’s a noted slowdown in demand for programmable chips used in various industries, including wireless communications, healthcare, and automotive.

  • Alibaba Enhances Tongyi Qianwen AI, Reveals Sector-Specific Models

Alibaba has announced significant updates to its artificial intelligence (AI) model, Tongyi Qianwen, enhancing it to version 2.0 with “hundreds of billions of” parameters, positioning it as one of the most powerful AI models globally. Alongside this update, Alibaba has released eight specialized AI models tailored for industries such as entertainment, finance, healthcare, and legal. This rapid development, occurring just six months after the initial release, underscores the swift progress in the competitive AI market in China.

The company’s advancements come amid a broader AI expansion in China, described by rival Tencent as a “war of a hundred models,” indicating a surge of over 130 AI models in the market. Tencent’s own AI, Hunyuan, boasts over 100 billion parameters, claiming to surpass OpenAI‘s GPT-4 in processing Chinese language tasks.

In addition to the upgraded Tongyi Qianwen model, Alibaba has launched its suite of industry-specific AI models, which became operational on Tuesday. These models provide dedicated tools for various applications, including image creation, computer code writing, financial data analysis, and legal document search. Alibaba’s Chairman Joe Tsai highlighted the company’s cloud dominance, stating that about half of the large-language AI models in China are now running on Alibaba Cloud.

  • Cohere introduces Embed V3 for enterprise LLM applications

Cohere, a Toronto-based AI startup, has released Embed V3, the newest version of its embedding model tailored for enterprise applications that utilize large language models (LLMs). Embed V3 is designed to enhance semantic search capabilities and offers improved data compression to reduce operational costs for businesses. The model competes with OpenAI’s Ada and other open-source alternatives, promising better performance in matching documents to queries by providing more accurate semantic representations.

Embed V3 plays a crucial role in retrieval augmented generation (RAG), which is essential for providing context to LLMs during runtime. It enables the AI system to retrieve relevant information from various sources, such as documents and histories, that were not part of the original training data. This process involves creating embeddings of documents and comparing them to user prompts to provide the necessary context to the LLM.

The model addresses challenges in enterprise AI by offering more precise document matching and reducing the likelihood of generating false information. Cohere claims that Embed V3 outperforms other models in standard benchmarks and is available in multiple embedding sizes, including a multilingual version. It also enhances reranking capabilities in search applications and reduces the costs associated with running vector databases through a specialized compression-aware training process, maintaining high search quality while cutting expenses.

  • SAP Advances Developer Tools with Generative AI Integration for Custom Applications

At the TechEd conference in Bengaluru, SAP announced a suite of AI tools aimed at enhancing developer productivity and enabling the creation of custom AI-powered applications. The German software giant introduced AI-infused pro-code tools within its Business Technology Platform (BTP), which now includes vector database capabilities in the HANA Cloud and an AI foundation hub. These tools are designed to streamline application development and facilitate the building of large language model (LLM) applications for various business scenarios.

SAP’s BTP, a cornerstone for cloud business applications, has evolved to include Build Code, a solution that embeds AI to generate code, data models, app logic, and test scripts, potentially increasing developer productivity by 40-60%. Additionally, SAP is moving towards empowering developers to build custom AI applications with the integration of vector database capabilities in HANA Cloud, simplifying the data preparation process for AI model training.

The company is also developing its own foundation model to enhance the relevance and quality of AI-driven business applications, complementing the general-purpose models from partners like OpenAI, Anthropic, and Meta. This proprietary model aims to deeply understand business processes and context, providing a robust starting point for enterprises. SAP’s approach remains partner-centric, with a commitment to compliance with AI safety and security standards as outlined in recent executive orders.

  • Cloudera Revamps Technology to Capture Corporate AI Market Share

Cloudera, a data analytics firm, has successfully revamped its core technology to target the growing artificial intelligence market, achieving profitability, according to the company’s CEO, Charles Sansbury. The firm, established in 2008 and known for its data analysis capabilities, faced challenges in profitability, leading to its privatization in a $4.7 billion deal by Clayton Dubilier & Rice and KKR.

Under the new leadership of CEO Sansbury since August, Cloudera has not only restructured its systems but also reported over $1 billion in annual revenue with operating profits in the hundreds of millions. This financial milestone places Cloudera at about half the size of its competitor Snowflake in terms of revenue.

Cloudera, which also competes with the highly valued startup DataBricks, is positioning itself to meet the needs of large corporations that prefer to maintain some data on-premises for privacy and security reasons. The company is focusing on assisting customers, especially in regulated sectors like finance, to prepare their data for in-house AI operations, aligning with the trend of leveraging AI technology while retaining data control.

  • Brave Introduces Leo: A New Privacy-Focused AI Chatbot

Brave, known for its privacy-centric web browser, has launched Leo, an AI assistant chatbot that emphasizes user privacy. Leo, which is free for all Brave desktop users on version 1.60, will soon be available on Android and iOS. Unlike other AI chatbots, Leo does not record user interactions or use them for AI training, ensuring a private user experience. It offers functionalities similar to competitors, such as translation, answering queries, summarizing content, and content generation, but users are advised to be cautious of potential inaccuracies.

The standard version of Leo is powered by Meta’s Llama 2 large language model and is available at no cost. For those seeking alternative AI models, Brave introduces Leo Premium, a subscription service at $15 per month featuring Anthropic’s Claude Instant. This premium option promises faster responses, cost-efficiency, and access to Anthropic’s Claude 2 large language model.

Brave’s commitment to privacy extends to Leo, with plans to offer additional AI models for premium users, including higher-quality interactions, priority queuing, and early feature access. Brian Bondy, Brave’s CTO and co-founder, highlights the company’s dedication to combining AI utility with stringent privacy standards, offering users a secure and personalized AI experience within their browser.

  • Shopify’s Strategic AI Integration and Cost Management Drive Profitable Quarter

Shopify, the Canadian e-commerce giant, reported a return to profitability in the third quarter, attributing its success to stringent cost management and the introduction of artificial intelligence tools aimed at attracting and retaining merchants. The company has been proactive in its approach, rolling out AI-driven solutions like the Shopify Magic suite and the Sidekick app, which have contributed to its competitive edge in the bustling online retail market. Additionally, Shopify has been focusing on enhancing its delivery speeds.

The firm, known for providing a platform for businesses to establish online stores, announced a net income attributable to shareholders of 55 cents per share, marking a significant turnaround from a loss in the same period last year. High-profile brand collaborations have also played a role in Shopify’s growth, with notable launches such as Taylor Swift’s Eras Tour merchandise and Drake’s new store, Drake Related, both of which have seen substantial sales and traffic.

Shopify’s strategic partnerships extend beyond celebrity collaborations. In a notable development, Amazon announced an app within Shopify’s ecosystem, granting Shopify’s U.S. merchants the ability to offer the “Buy with Prime” feature directly on their sites, a move that analysts believe is particularly advantageous for Shopify as it allows merchants to retain full control over their brand and customer data.

  • UK Boosts Investment to 300 Million Pounds for AI Supercomputing Research

The UK government has announced a significant increase in funding for the development of two supercomputers dedicated to AI research, raising the investment to 300 million pounds. This move, revealed during an AI safety summit, aims to ensure that the country’s scientific community has the necessary resources to advance AI technology safely and responsibly. Prime Minister Rishi Sunak emphasized the importance of this investment in supporting Britain’s scientific talent to work on making sophisticated AI models secure.

Scheduled to be operational by next summer, these supercomputers will be located in Cambridge and Bristol. They will provide researchers with access to computing power over thirty times greater than the UK’s current largest public AI computing resources. The focus will be on analyzing and testing the safety of advanced AI models, with additional applications in drug discovery and clean energy initiatives.

The Bristol-based supercomputer, named “Isambard-AI,” will feature 5,000 advanced AI chips from Nvidia and is being constructed by Hewlett Packard Enterprise. The Cambridge counterpart, “Dawn,” will be developed in collaboration with Dell and UK SME StackHPC, powered by over 1,000 Intel chips. These investments underscore the UK’s commitment to leading in the safe development and application of AI technologies.

About The Author

Bogdan Iancu

Bogdan Iancu is a seasoned entrepreneur and strategic leader with over 25 years of experience in diverse industrial and commercial fields. His passion for AI, Machine Learning, and Generative AI is underpinned by a deep understanding of advanced calculus, enabling him to leverage these technologies to drive innovation and growth. As a Non-Executive Director, Bogdan brings a wealth of experience and a unique perspective to the boardroom, contributing to robust strategic decisions. With a proven track record of assisting clients worldwide, Bogdan is committed to harnessing the power of AI to transform businesses and create sustainable growth in the digital age.