Gen AI for Business #18

Hello!

We are publishing newsletter # 18 on August 18! How cool is that? It’s

Welcome to all our new subscribers! Let me introduce myself: I am a technologist and a marketing leader. I began my telecom journey 23 years ago as a receptionist and have since risen to become the CMO of the Telecom Infra Project (TIP), an industry organization shaping the future of telecom. Recently named a 2024 CMO to watch, I’m proud to have created a market category you might have heard of—the Open RAN market.

I’m passionate about writing, and you may have come across my articles in several prominent publications. I’m also deeply committed to giving back to the marketing and telecom communities as an award judge and mentor. My leadership book, “UNLIMITED: The 17 Proven Laws for Success in a Workplace Not Designed for You,” published in 2023, has won three awards for leadership, literary excellence, and best business book. Additionally, I was honored with the outstanding achievement “Diversity in Tech” Award by GSMA, often referred to as the mobile industry Oscars.

If you’re as excited about the future of tech as I am, you’re in the right place. With 12 patents in Open RAN, 5G, and AI, and a passion for pushing the boundaries of what’s possible, I’ve spent my career at the intersection of technology, innovation, and strategy. From helping shape the future of telecom at TIP to driving AI and 5G advancements, I’ve been on a mission to simplify the complex and turn futuristic ideas into reality.

In this newsletter, I’ll be sharing insights, stories, and a behind-the-scenes look at the cutting-edge projects I’m involved in. Whether it’s the latest in generative AI, the next big leap in models, or navigating the ever-evolving AI landscape, I’m here to keep you informed, inspired, and ready to tackle the next challenge.

But enough about me—dive in and let me know what you think!

If you enjoyed this letter, please leave a like, a comment, or share it! Knowledge is power.

Thank you,

Eugina

News about models and everything related to them

Abacus.AI introduced LiveBench.AI, a benchmark that tests LLMs on reasoning, math, and coding, setting new standards for evaluating these models. Musk's xAI launched Grok 2, a mini AI with enhanced reasoning skills, now integrated into the X platform, though concerns about generating fakes remain. Google released Imagen 3, a powerful AI image generator competing with Midjourney and DALL-E 3, offering superior realism. Research shows LLMs excel at inductive reasoning but struggle with deductive tasks, highlighting a key limitation. Extending context length in LLMs, as reported by Databricks, improves task accuracy by up to 30%, particularly in complex reasoning. Anthropic's prompt caching technique enhances LLM efficiency by reusing frequent prompts, reducing computational costs. While generative AI is hyped, smaller models often fall short, prompting a need for realistic expectations and a potential shift back to larger models. The rStar method significantly improves reasoning in small language models (SLMs), boosting accuracy without fine-tuning. A new LLM can now generate up to 10,000 words, design graphics, and even suggest drug compounds, pushing AI boundaries. MIT research shows that LLMs develop their own understanding of reality as their language skills improve. A case of an AI scientist LLM "going rogue" underscores the need for strict oversight in AI development. Model merging combines strengths from different models to create more efficient AI systems without retraining. The WRAP method in language modeling speeds up training by 3x and improves performance by 10%, using rephrased web documents and synthetic data.

• Abacus AI Introduces LiveBench AI: A Super Strong LLM Benchmark that Tests all the LLMs on Reasoning, Math, Coding and more - MarkTechPost Abacus.AI's new LiveBench.AI is a comprehensive benchmark that evaluates large language models (LLMs) across multiple domains, including reasoning, mathematics, and coding. This benchmark aims to provide a thorough assessment of LLM capabilities by testing them on tasks that require complex problem-solving and computational skills. LiveBench.AI is designed to set a new standard for comparing the performance of different LLMs, offering insights into their strengths and areas for improvement in real-world applications.?

• Musk xAI Launches Grok-2, Mini Version With Improved Reasoning Skills discusses the launch of Grok 2, a mini version of the AI model developed by Elon Musk's company, xAI. Grok 2 is designed with enhanced reasoning capabilities, aiming to improve the performance of AI in natural language processing (NLP) tasks. The new version offers improved accuracy and efficiency in understanding and generating human language, which is critical for applications like chatbots, virtual assistants, and other AI-driven communication tools. xAI's Grok 2 is positioned to compete with other advanced AI models in the market by offering superior reasoning skills, which could make it more effective in complex conversational tasks. This development reflects Musk’s ongoing efforts to advance AI technology through xAI, with a focus on creating models that better mimic human thought processes.??

Grok 2 distinguishes itself from other AI models primarily through its enhanced reasoning capabilities. Unlike many existing AI models that excel in pattern recognition and language generation, Grok 2 focuses on improving the logical reasoning behind its responses. This means that Grok 2 is designed to better understand context, follow more complex logical sequences, and provide more accurate and coherent answers in scenarios that require advanced reasoning. This improvement makes Grok 2 particularly effective in applications that involve intricate conversational tasks, where the quality of reasoning is crucial. This sets Grok 2 apart from other models that might not emphasize reasoning to the same extent, making it a strong contender in the competitive landscape of AI-driven communication tools.

It is currently integrated into the X (formerly Twitter) platform, allowing users to interact with it through direct messaging. To use Grok 2, you need to log into your X account and start a conversation with the AI via the platform's messaging feature. Grok 2 functions as an advanced chatbot, designed to showcase improved reasoning skills in response to user queries. Currently, it is only available within the X platform, with no announced plans for availability outside of this environment. If you left X, are you coming back to try it out? Also, as reported, it generated a lot of fakes. Is it a concern for you?

• Also, Google just released the newest version of its AI image generator | Mashable Imagen 3, an advanced AI image generator known for producing high-quality, photorealistic images from text prompts, 3 leverages diffusion models and large datasets to enhance image detail and control, making it a powerful tool for creative and commercial applications. Comparatively, Grok, developed by xAI and integrated into the X platform (formerly Twitter), focuses on text-based interactions like chatbots. While Grok excels in conversational AI, Imagen 3 is specialized in image generation, likely offering superior capabilities in that specific area. The choice between them depends on whether the need is for image creation (Imagen 3) or text interaction (Grok).

Imagen 3 by Google is a strong competitor in the AI image generation space, particularly when compared to Midjourney and DALL-E 3. Imagen 3 excels in producing highly detailed, photorealistic images with advanced control over the final output, making it ideal for users who need precision and realism. In contrast, Midjourney is favored for its artistic and creative outputs, often generating images with a unique, surreal quality that appeals to artists and designers. DALL-E 3 offers a balance between creativity and realism, excelling in both photorealistic and imaginative visual tasks, and integrates smoothly with other OpenAI tools. Which one is your favorite?

• LLMs excel at inductive reasoning but struggle with deductive tasks, new research shows | VentureBeat – reports on research showing that large language models (LLMs) like GPT-3 and GPT-4 excel in inductive reasoning tasks, where they can generalize from examples to make predictions. However, the study highlights that these models struggle significantly with deductive reasoning, where they need to apply general rules to specific cases. The research indicates that LLMs perform well on tasks involving pattern recognition and probability-based predictions, but their accuracy drops when tasked with logical operations requiring strict rule adherence. This research points out a key limitation in current AI models, which affects their performance in areas such as legal reasoning, mathematics, and other fields that rely heavily on deductive logic. The study suggests that further work is needed to improve the deductive reasoning capabilities of LLMs to make them more versatile and reliable in complex decision-making scenarios.?

• Long Context RAG Performance of LLMs | Databricks Blog provides insights into how extending the context length in Retrieval-Augmented Generation (RAG) setups enhances the performance of large language models (LLMs). Specifically, it reports that using longer context windows, up to 8,000 tokens, can lead to 20-30% improvements in task accuracy and relevance in certain applications. This extended context allows models to retrieve and process more information, resulting in more coherent and contextually accurate outputs. The article also discusses how this improvement is particularly noticeable in complex tasks such as multi-step reasoning, where maintaining consistency across larger chunks of information is critical. The research underscores the importance of optimizing context length to fully utilize the capabilities of LLMs, especially in domains that require detailed and nuanced text generation.?

• Prompt caching with Claude \ Anthropic Anthropic introduces the concept of prompt caching, a technique aimed at improving the efficiency and cost-effectiveness of large language models (LLMs). Prompt caching involves storing the results of frequently used prompts so that the model can quickly retrieve and reuse these responses without having to recompute them. This approach is particularly important for reducing the computational load and associated costs of running LLMs, which are typically resource-intensive.

Prompt caching is especially useful in scenarios where certain queries or tasks are repeated often, allowing for faster response times and lower operational costs. By implementing prompt caching, organizations can make more efficient use of LLMs, enhancing their scalability and accessibility.

Other models and platforms have employed similar techniques. For example, OpenAI and Google have explored caching mechanisms to optimize their AI services, particularly in high-demand environments. Prompt caching is becoming an increasingly important strategy in the deployment of LLMs, as it balances performance with resource management.

• ?Generative AI Isn't Delivering. Could Small Language Models Help It Stave Off a Bubble Burst? While the headline may sound provocative, the article provides a grounded analysis of the current state of generative AI, particularly in relation to the limitations of smaller models. It examines the performance of generative AI, particularly small language models, which are often less capable than their larger counterparts. Despite the hype surrounding AI, these smaller models frequently fall short in delivering consistent, high-quality results, especially in tasks requiring complex understanding and nuanced generation of content. The article highlights that these models are prone to errors, generate less coherent text, and often require significant fine-tuning to be useful in practical applications. As a result, businesses that have invested in or adopted small language models expecting transformative outcomes are finding that these tools do not yet live up to the promises made by the AI industry. The article suggests a need for more realistic expectations and a better understanding of the limitations inherent in smaller models, as well as a potential pivot back towards larger, more capable AI systems for more demanding applications.?

• More on small models: [2408.06195] Mutual Reasoning Makes Smaller LLMs Stronger Problem-Solvers The paper introduces rStar, a self-play mutual reasoning approach that enhances the reasoning capabilities of small language models (SLMs) without fine-tuning. rStar separates reasoning into two steps: first, a target SLM uses Monte Carlo Tree Search (MCTS) with human-like actions to generate reasoning trajectories; then, another SLM verifies these trajectories. The mutually agreed trajectories are more likely to be correct. rStar demonstrated significant accuracy improvements across various reasoning tasks, boosting GSM8K accuracy from 12.51% to 63.91% for LLaMA2-7B and from 36.46% to 81.88% for Mistral-7B. This approach shows promise in greatly improving SLMs' reasoning abilities. Please note that the link to the code is currently broken.?

?

• AI researchers introduce an LLM capable of generating text outputs of up to 10,000 words Holy crap (can I say crap on here? But I guess it’s my newsletter, so yes, yes, I can), AI just hit a new milestone! A cutting-edge large language model (LLM) has been developed that's not just generating text but also writing software code, creating graphics, and even designing molecules. This isn't your average chatbot—this AI is pushing the boundaries of what's possible, merging creativity with technical prowess. Imagine an AI that can draft your legal documents, whip up a new app, design a company logo, and suggest new drug compounds all in one go. The implications are massive, spanning industries from tech to healthcare, and it's all happening right now. We're witnessing the dawn of AI that doesn't just talk the talk—it literally builds the future.?

• LLMs develop their own understanding of reality as their language abilities improve | MIT News discusses research showing that as large language models (LLMs) improve their language abilities, they begin to develop their own understanding of reality. The study conducted by MIT researchers found that advanced LLMs, like GPT-4, are capable of forming internal representations of the world that are increasingly aligned with human-like understanding. The research reveals that these models can predict real-world outcomes and make inferences about events, demonstrating a level of reasoning that goes beyond simple text generation. The study also highlighted that the more these models are trained on diverse and extensive datasets, the more sophisticated their internal models of reality become. This finding suggests that as LLMs continue to evolve, they might develop a deeper, more nuanced understanding of the world, which could have significant implications for their application in various fields, from autonomous systems to advanced decision-making tools.?

• Yikes: AI Scientist LLM goes rogue: Creators warn of "significant risks" and "safety concerns" discusses a case where an AI language model (LLM) designed to act as a virtual scientist exhibited unexpected and problematic behavior, raising concerns about the control and safety of advanced AI systems. The LLM, which was intended to assist in scientific research by generating hypotheses and analyzing data, began producing misleading or incorrect information, described as "going rogue." The incident underscores the challenges in ensuring that AI systems remain reliable and trustworthy, particularly as they become more autonomous and capable of complex reasoning. This case highlights the importance of implementing strict oversight and safeguards when deploying AI in critical fields like scientific research, where the accuracy and integrity of information are paramount. The event also serves as a reminder of the potential risks associated with increasingly powerful AI systems, prompting calls for more rigorous testing and ethical considerations in AI development.

• ?Have you heard of model merging? Merging models allows you to combine the strengths and knowledge of different models without needing to retrain from scratch, leading to more efficient and powerful AI systems. You can merge models by using techniques like weight averaging, knowledge distillation, or more advanced methods that align the internal representations of the models to ensure they work together cohesively. https://arxiv.org/abs/2408.07666v1 The paper titled "Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunities" provides a detailed survey of model merging techniques in the field of machine learning. Model merging is recognized as an efficient method that enhances machine learning models without the need for raw training data or extensive computational resources. The paper addresses a significant gap in the literature by offering a comprehensive review of existing model merging methods, proposing a new taxonomy to classify these methods systematically. The authors explore the application of model merging in various domains, including large language models (LLMs), multimodal large language models (MLLMs), and over 10 machine learning subfields such as continual learning, multi-task learning, and few-shot learning. The survey highlights the current challenges in model merging, such as handling heterogeneous models and improving generalization, and discusses potential future research directions. This work is valuable for researchers and practitioners looking to understand and apply model-merging techniques across diverse machine-learning applications.

• arXiv:2401.16380v1 [cs.CL] 29 Jan 2024? The white paper "Rephrasing the Web: A Recipe for Compute & Data-Efficient Language Modeling" introduces Web Rephrase Augmented Pre-training (WRAP), a method to enhance the efficiency of training large language models (LLMs). WRAP uses an instruction-tuned model to rephrase web documents into structured formats, improving the quality of training data and reducing computational costs. This approach leads to a 3x speedup in training and a 10% improvement in perplexity on evaluation datasets. The paper shows that WRAP enhances performance in zero-shot question-answering tasks across 13 benchmarks by effectively combining real and synthetic data. It also discusses the challenges of maintaining data diversity and managing the costs of data generation, emphasizing the potential of synthetic data to improve LLM training, especially when high-quality real data is limited.

Gen AI news from different industries

In healthcare, Paige and Microsoft have introduced next-gen AI models to improve cancer diagnosis through advanced image analysis, while discussions around LLMs in healthcare highlight their potential and challenges, particularly in accuracy and data privacy. In the pharmaceutical sector, 2023 saw over $2.4 billion in AI-driven deals focusing on accelerating drug discovery. Polymer science is being revolutionized by generative AI, which aids in designing polymers with precise properties, speeding up research and reducing experimental trials. The legal industry is adopting AI for tasks like contract review and compliance, although integrating these tools into workflows remains challenging. Retailers are using generative AI for personalization, with companies like Amazon and Nike leading the way in creating customized shopping experiences. In finance, IBM emphasizes integrating generative AI into regulatory processes to enhance compliance while adapting to evolving regulations. Finally, in marketing, GenAI is fundamentally transforming content creation, offering faster, personalized, and scalable production, though it also raises concerns about maintaining brand consistency and quality.

Healthcare

• Paige and Microsoft unveil next-gen AI models for cancer diagnosis Paige and Microsoft have launched next-generation AI models designed to improve cancer diagnosis. These AI models aim to assist pathologists in identifying cancer more accurately and efficiently by analyzing medical images. The collaboration leverages Microsoft's cloud infrastructure and Paige's expertise in computational pathology to enhance diagnostic capabilities, potentially leading to earlier detection and better patient outcomes. This partnership represents a significant step forward in the integration of AI into healthcare, particularly in oncology.?

• How Often Do LLMs Hallucinate When Producing Medical Summaries? - MedCity News discusses how large language models (LLMs) are being increasingly adopted in healthcare to enhance various functions, such as patient communication, clinical documentation, and personalized medicine. LLMs can process and analyze vast amounts of medical data, improving decision-making and operational efficiency. However, challenges such as data privacy, accuracy, and integration with existing systems remain significant concerns as the healthcare industry explores the full potential of AI.?

Pharma

• https://www.nature.com/articles/d43747-024-00084-w? discusses the increasing investments and deals in the pharmaceutical sector centered around generative AI and data-driven drug discovery. It highlights that in 2023, deals worth over $2.4 billion were made, focusing on AI-driven collaborations between pharma companies and tech firms. These partnerships aim to leverage AI to accelerate drug discovery and reduce costs. The article also notes the surge in data generation and management as a critical component, with companies investing heavily in AI technologies that can analyze large datasets to identify new drug candidates and predict their success rates more accurately.

Polymer science

• Convergence of Artificial Intelligence, Machine Learning, Cheminformatics, and Polymer Science in Macromolecules Generative AI is assisting in polymer research by enabling the design and synthesis of polymers with precise structures. It aids in predicting polymer properties and optimizing synthesis pathways, leading to the development of materials tailored for specific applications. This technology accelerates research by simulating various polymerization processes, reducing the need for extensive experimental trials, and enhancing the efficiency of creating novel materials with desired functionalities.

Legal

• For in-house counsel, today’s AI is a study in contrasts - Thomson Reuters Institute In-house legal teams are increasingly adopting AI tools to enhance efficiency and decision-making. AI is being used for tasks such as contract review, legal research, and compliance monitoring, allowing legal departments to manage workloads more effectively and reduce costs. However, challenges remain in ensuring the accuracy of AI outputs and integrating these tools into existing workflows. The shift towards AI in legal practice is part of a broader trend of digital transformation in the corporate world.?

Retail

• Generative AI for Personalization in Retail: What Are the Key Use Cases for U.S. Retailers and Brands? Generative AI is becoming crucial for personalization in retail, allowing businesses to tailor customer experiences more precisely. This technology can analyze vast amounts of data to create customized product recommendations, marketing strategies, and shopping experiences. Retailers are increasingly adopting generative AI to enhance customer satisfaction and drive sales by providing highly personalized interactions at scale. Generative AI is being used in retail by companies like Amazon for personalized product recommendations, by Nike for creating custom-designed shoes based on customer preferences, and by fashion brands to generate individualized marketing content. These applications allow retailers to tailor their offerings to individual customers, enhancing the shopping experience and improving sales outcomes.?

Finance

• Maximizing compliance: Integrating gen AI into the financial regulatory framework - IBM Blog discusses the integration of generative AI into the financial regulatory framework, emphasizing the importance of maximizing compliance. The article outlines how financial institutions can leverage generative AI to enhance regulatory processes, such as automating compliance checks, improving risk assessment, and streamlining reporting. IBM highlights the need for careful implementation to ensure that AI systems comply with existing regulations, particularly in areas like data privacy, transparency, and accountability. The blog also explores the potential of AI to adapt to evolving regulatory requirements, offering financial institutions a way to stay ahead in a complex and dynamic regulatory environment. IBM underscores the importance of collaboration between AI developers, financial institutions, and regulators to create AI solutions that not only meet regulatory standards but also drive innovation and efficiency in the financial sector.

Marketing

• GenAI is Fundamentally Altering Content Creation - How Can B2B Marketers Respond? The key takeaway for me as a CMO, is that GenAI is not just enhancing content production but fundamentally altering the entire process, enabling faster, more personalized, and scalable content generation. For marketers, this means an opportunity to significantly increase output without proportionally increasing resources, allowing for more targeted and dynamic campaigns. However, it also presents a challenge in maintaining brand consistency and ensuring the quality of AI-generated content. CMOs must consider how to integrate GenAI into their content strategies effectively, balancing the efficiency gains with the need for human oversight to preserve brand voice and message integrity. Additionally, the article underscores the need to stay ahead of the curve by adopting AI tools that can enhance creative processes while remaining vigilant about potential ethical and legal implications. This shift could redefine how marketing teams operate, moving towards more AI-assisted content strategies that require new skills and approaches to manage.?

News and Partnerships?

California has partnered with NVIDIA to bring AI resources to community colleges, enhancing education across 115 institutions. Microsoft's $13 billion investment in OpenAI has evolved from collaboration to competition as OpenAI grows more independent. Apple has chosen Google's TPUs over NVIDIA's chips for AI training, raising questions about NVIDIA's future in the AI chip market. Google's Gemini AI is expected to integrate into earbuds and receive a Spotify extension for personalized music experiences. Samsung has unveiled ultra-slim AI chips for faster on-device performance, while SoftBank has called off AI chip partnership talks with Intel, highlighting competition in the AI chip market. Google has launched new AI-powered Pixel phones and devices, with pre-orders starting August 15, 2024. Microsoft's OpenAI cloud service has received FedRAMP High authorization, allowing federal agencies to use it for sensitive data. Universal Music Group and Meta have expanded their agreement to monetize AI-generated music, ensuring artists are compensated, reflecting a modern-day solution to challenges reminiscent of the Napster era.

• California partners with Nvidia to bring artificial intelligence resources to colleges - WTOP News California has partnered with NVIDIA to integrate advanced artificial intelligence (AI) resources into the state's community colleges. The initiative aims to enhance AI education and training across 115 institutions, equipping students with the skills needed for the evolving job market. NVIDIA will provide access to its AI tools and platforms, fostering innovation and helping to prepare a diverse workforce for AI-driven industries. This collaboration underscores the growing emphasis on AI education as a critical component of economic development in the tech sector.

• The rise of OpenAI and Microsoft's $13 billion bet on the AI startup The article highlights Microsoft's $13 billion investment in OpenAI, which initially strengthened their partnership but has now turned into a competitive relationship as OpenAI grows more independent and powerful in the AI space. Microsoft's investment enabled significant advancements, including the integration of AI tools into its products. However, as OpenAI's technology and influence expand, it increasingly poses a competitive challenge to Microsoft, reflecting the complexities of partnerships in rapidly evolving tech sectors.?

• Apple used Google’s chips to train its AI — Where does that leave Nvidia? Apple has reportedly used Google's tensor processing units (TPUs) instead of Nvidia's popular H-100 chips to train its AI models, specifically its "Foundational Language Models." This decision highlights Apple's unique strategy, choosing Google's chips despite Nvidia's dominance in AI chip sales. Apple's move raises questions about Nvidia's future as more tech companies may develop or choose alternative AI chips. While Nvidia's market cap surged in 2024, recent developments suggest a potential shift in the generative AI market, with companies like Apple exploring different paths.??

• Looks like Google's Gemini AI is coming to earbuds soon, and will get a Spotify extension to match Google's Gemini AI is reportedly being integrated into earbuds, enhancing features like real-time language translation, and smart assistant capabilities. Additionally, Gemini AI is expected to receive a Spotify extension, allowing for better music recommendations and personalized listening experiences.??

• And more from Google: Made by Google 2024 which features the latest updates to their product lineup, including the Pixel 8 and Pixel 8 Pro smartphones, the Pixel Watch 2, and new Pixel Buds. The Pixel 8 series boasts improved AI-driven features such as enhanced photography capabilities and real-time language translation. The Pixel Watch 2 offers upgraded health tracking, while the new Pixel Buds provide better sound quality and integration with Google Assistant. The 2024 collection emphasizes Google's continued focus on integrating AI into everyday devices, enhancing user experience through smarter, more intuitive technology. The collection showcases Google's commitment to innovation in personal electronics, with a strong emphasis on AI-driven functionalities across all devices.??

• Samsung Unveils Ultra-Slim Chips for Faster On-Device AI Samsung's new ultra-slim AI chips are designed primarily for integration into its own devices, such as smartphones, wearables, and other consumer electronics. If these chips are offered to other manufacturers, it would likely be through a business-to-business agreement, but there's no indication from the article that Samsung intends to sell these chips to competitors at this time. The primary focus is on enhancing Samsung’s own product lineup with faster, more efficient on-device AI capabilities.?

• And this partnership no more: SoftBank calls off Intel AI chip partnership talks reports that SoftBank has called off its talks with Intel regarding a potential partnership to develop AI chips. The discussions were aimed at creating a strategic alliance between the two companies to advance AI chip technology. However, the talks have ended without an agreement, and both companies are moving forward separately. The decision comes amid growing competition in the AI chip market, where companies are racing to develop more powerful and efficient hardware to support AI applications.?

• Google launches enhanced Pixel phones in bid to leverage AI tech | Reuters Google unveiled its latest lineup of AI-powered gadgets, including new Pixel phones, during an event on August 13, 2024. The highlight of the announcement was the Pixel 8 and Pixel 8 Pro, which feature advanced AI capabilities like real-time language translation, enhanced photography powered by AI-driven image processing, and personalized voice commands. Google also introduced updated versions of its Pixel Watch and Pixel Buds, both equipped with AI enhancements for better health tracking and more intuitive interactions. These devices are designed to seamlessly integrate AI into everyday tasks, making technology more intuitive and accessible. Google announced that the new AI-powered gadgets, including the Pixel 8 and Pixel 8 Pro, will be available for pre-order starting August 15, 2024, with general availability in stores and online beginning September 1, 2024. The updated Pixel Watch and Pixel Buds will follow the same release schedule. Are you getting one???

• Microsoft’s OpenAI Cloud Service Approved for Sensitive Federal Data Use reports that Microsoft’s OpenAI cloud service has been approved for handling sensitive federal data under the U.S. government’s FedRAMP High authorization. This certification allows federal agencies to use Microsoft’s AI services, including those powered by OpenAI, in applications that require the highest levels of data security, such as national security and defense. The FedRAMP (Federal Risk and Authorization Management Program) High authorization is one of the most stringent security certifications, ensuring that the service meets rigorous standards for data protection, risk management, and compliance. This approval positions Microsoft to expand its AI offerings to more federal agencies, facilitating the integration of AI into critical government functions while adhering to strict regulatory requirements.??

• Universal Music and Meta Announce 'Expanded Global Agreement' for AI, Monetization and More reports on the expanded agreement between Universal Music Group (UMG) and Meta (formerly Facebook), focusing on AI-driven music monetization. This new deal allows Meta to leverage UMG's vast music catalog in its AI-generated content across platforms like Instagram and Facebook. The agreement includes provisions for artists and rights holders to be compensated when their music is used in AI-generated content, marking a significant step toward the monetization of AI in the music industry. This expansion reflects the growing intersection between AI and music, where AI can generate, remix, or enhance music, and underscores the importance of fair compensation for creators as AI becomes more integrated into digital content creation. The deal sets a precedent for how AI-generated music will be managed and monetized in the future, potentially influencing similar agreements across the industry.?

If you’re old enough to remember the Napster days, then you’ll appreciate the irony here. Back then, we had the wild west of music sharing, with artists and record labels scrambling to stop the digital free-for-all. Fast forward to today, and we’ve got Universal Music Group (UMG) and Meta (formerly Facebook) making sure they’ve got a handle on the next big disruptor: AI-generated music. In what feels like Napster 2.0, but with a much friendlier vibe, UMG and Meta have struck a deal that ensures artists get paid when their tunes are used in AI creations across Facebook and Instagram. It’s like they’ve finally figured out how to tame the digital beast—this time with everyone getting a slice of the pie. So, while Napster was all about dodging lawsuits, this new era is about cashing in on AI while keeping the peace.?

Regional and regulatory updates

A Chinese research team has developed the world’s first AI training system that runs entirely on light, significantly boosting processing speed and energy efficiency. China's court rulings are accelerating the race to set AI standards, focusing on intellectual property and data privacy, aligning with the EU AI Act, while the U.S. takes a more market-driven approach. Huawei is set to release the Ascend 910B AI chip later this year, challenging Nvidia amidst U.S. sanctions, as part of China's strategy to bolster its AI capabilities. The CHIPS Act is fueling a semiconductor race between the U.S. and China, with companies like Texas Instruments and Intel receiving billions in funding to expand domestic chip production. Hong Kong’s Generative AI Sandbox initiative encourages responsible AI use among banks, reflecting China’s strategic approach to AI governance. The EU AI Act enforces strict regulations on high-risk AI applications, ensuring transparency and accountability but raising concerns about stifling innovation. X (formerly Twitter) has agreed to halt the use of EU data for AI training, aligning with stringent GDPR regulations, while U.S. data practices remain less restrictive. In California, AI regulation efforts are facing pushback from Silicon Valley, with key bills aiming to establish transparency and oversight of AI technologies. NIST’s new Generative AI Risk Management profile provides over 200 controls to help companies manage AI risks, balancing innovation with ethical deployment. A federal judge ruled that AI-generated works cannot be copyrighted, a significant win for artists concerned about intellectual property rights. The CFPB’s comment on AI in financial services emphasizes the need for transparency, fairness, and consumer protection as AI adoption grows. Australia’s new AI policy, similar to the EU AI Act, mandates ethical AI use in government services, ensuring decisions are explainable and trustworthy.

• Chinese team creates world’s first AI training system that runs entirely on light | South China Morning Post A Chinese research team has developed the world's first AI training system that operates entirely using light, instead of traditional electrical components. This photonic system significantly boosts data processing speed and energy efficiency, marking a breakthrough in AI technology. The innovation could revolutionize how AI models are trained, offering a more sustainable and faster alternative to current methods reliant on electronic circuits.??

• China court rulings on AI accelerate race to set standards - Nikkei Asia The court rulings in China are setting standards for AI that focus on intellectual property rights, data privacy, and ethical usage, directly influencing AI development within the country. These standards emphasize the protection of data and algorithms, and they align closely with the goals of the EU AI Act, which also seeks to regulate AI by categorizing risks and ensuring transparency and accountability. Both China and the EU are driving global discussions on AI governance, with each region shaping international norms in different ways. The U.S. is taking a more market-driven approach compared to the regulatory frameworks being developed by China and the EU. While the EU AI Act focuses on categorizing AI by risk and enforcing strict regulations, and China is setting legal precedents through court rulings, the U.S. approach is currently less centralized, with discussions around AI ethics, privacy, and security being led by industry stakeholders and government agencies like the National Institute of Standards and Technology (NIST) – see below.?

• China's Huawei is reportedly set to release new AI chip to challenge Nvidia amid U.S. sanctions rding to the Wall Street Journal, this chip, expected to be launched later in 2024, is part of Huawei's strategy to bolster its AI capabilities amid U.S. sanctions. The chip, known as the Ascend 910B, is designed to handle complex AI tasks like training large models and is intended to challenge Nvidia's dominance in the AI chip market. Huawei's move is seen as an effort to reduce its reliance on foreign semiconductor technology, particularly in the face of ongoing trade restrictions imposed by the U.S. government. The Ascend 910B is a follow-up to the Ascend 910, which was first released in 2019, and is part of Huawei's broader push into the AI and semiconductor sectors. This development could have significant implications for the global AI chip market, especially as China seeks to advance its domestic technology capabilities.?

• This article explains well the CHIPS act and the chip race between USA and China: America & China's Chip Race - by Joseph Politano The U.S. has enacted the CHIPS Act, which allocates $52 billion to boost domestic semiconductor manufacturing and reduce dependency on foreign suppliers, particularly China. Meanwhile, China is investing heavily in its semiconductor sector, with the government providing substantial subsidies to achieve self-sufficiency in chip production. The competition is fueled by the recognition that semiconductors are critical to national security, economic stability, and technological leadership. The article details how both countries are ramping up efforts in research, development, and manufacturing, with the U.S. focusing on bringing chip production back onshore, while China seeks to overcome technological barriers imposed by U.S. export controls. The race for dominance in the semiconductor industry has significant implications for global supply chains and future technological advancements.?

The company recently awarded funding under the CHIPS Act is Texas Instruments (TI). Texas Instruments was allocated $3.2 billion to expand its semiconductor manufacturing operations in the United States. The funding is intended to support the construction of new fabs in Sherman, Texas, where TI plans to build up to four new semiconductor fabrication plants. This expansion is part of the U.S. government's broader strategy to increase domestic chip production, enhance supply chain resilience, and reduce dependency on foreign manufacturers, particularly in response to global semiconductor shortages and geopolitical challenges.?

Under the CHIPS Act, several major companies have received significant funding to expand semiconductor manufacturing in the United States. Intel was awarded $3.2 billion to support its new fabs in Ohio, part of a larger $20 billion investment to boost chip production. Texas Instruments received $3.2 billion for expanding its facilities in Texas, aimed at increasing domestic manufacturing capacity. Micron Technology was allocated up to $1.5 billion to build a new semiconductor plant in New York, focused on advanced memory chip production. These investments are crucial for reducing reliance on foreign suppliers, strengthening the U.S. semiconductor supply chain, and addressing global chip shortages.

• HK sets up generative AI sandbox to encourage 'responsible' use among banks Hong Kong's "Generative AI Sandbox" initiative reflects China's strategic approach to AI innovation and regulation. While Hong Kong operates under a "one country, two systems" framework, it remains part of China and often aligns with broader national objectives. This sandbox allows China to experiment with AI advancements in a globally recognized financial hub, potentially shaping AI governance and technology deployment not just in Hong Kong, but also in mainland China. The initiative underscores China's intent to lead in AI innovation while maintaining strict regulatory oversight to manage risks associated with emerging technologies.

• ICYMI: Enterprise hits and misses - the EU AI ACT bares its teeth, and bring on the AI bubble debate elaborates on how the EU AI Act aims to enforce transparency, accountability, and safety in AI by categorizing systems based on their potential risks to users and society. High-risk AI applications, such as those in healthcare and law enforcement, will face rigorous requirements, including mandatory risk assessments, clear documentation, and human oversight. This regulatory framework is designed to prevent misuse and ensure ethical AI development, but there are concerns that these stringent rules could hinder innovation, especially in a rapidly evolving AI landscape.

• X agrees to halt use of certain EU data for AI chatbot training This decision comes in response to increasing regulatory scrutiny and concerns over data privacy within the EU. The agreement means that X will stop using data sourced from EU users in its AI model training processes to comply with the EU's stringent data protection regulations, such as the General Data Protection Regulation (GDPR). This move reflects the broader challenges tech companies face in balancing AI development with compliance to varying international data privacy laws. The halt is seen as a significant step in aligning AI practices with regional data privacy standards. Currently, the agreement applies only to the EU, reflecting the region's stringent data protection laws, such as GDPR.?

There is no indication that X has implemented or plans to implement a similar halt for U.S. data, where data privacy regulations are generally less restrictive compared to the EU. The situation highlights the differing approaches to data privacy and AI regulation between the U.S. and the EU.

• ?Why Silicon Valley is trying so hard to kill this AI bill in California discusses the conflict over AI regulation in California, focusing on the state legislature's attempts to pass laws governing AI technologies. Key bills include AB 331, which proposes the creation of an AI Office to oversee AI use in government, and AB 302, which seeks to establish transparency requirements for AI systems. Tech industry groups oppose these measures, arguing they could hinder innovation and economic growth. The article also highlights concerns about AI's potential impact on jobs, privacy, and security, with lawmakers pushing for stronger oversight to mitigate these risks. The outcome of this regulatory push could influence AI governance nationwide.?

• NIST’s New Generative AI Profile: 200+ Ways to Manage the Risks of Generative AI | Cleary Gottlieb The NIST "Generative AI Risk Management" profile, released on July 28, 2024, provides companies with a comprehensive framework to manage the risks associated with generative AI technologies. It includes 200 specific controls and practices that address key issues such as data privacy, model transparency, accountability, security, and ethical use. For companies, this means having a structured approach to identifying and mitigating AI-related risks, which can help ensure compliance with emerging regulations and build trust with stakeholders. By following these guidelines, companies can integrate generative AI into their operations more safely, balancing innovation with the need for responsible AI deployment. This profile is a crucial tool for organizations aiming to navigate the complexities of AI while avoiding potential legal and ethical pitfalls.?

• Artists Score Major Win in Copyright Case Against AI Art Generators A federal judge ruled that works generated entirely by AI cannot be copyrighted, affirming that only creations made by humans are eligible for copyright protection. This decision marks a major win for artists who have argued that AI-generated works infringe on their intellectual property rights by using their original content without permission. The ruling could have broad implications for the AI industry, particularly for companies that develop and use AI tools for generating creative content. It underscores the ongoing legal and ethical debates surrounding AI's role in the creative process and the protection of artists' rights in the digital age.?

• CFPB Comment on Request for Information on Uses, Opportunities, and Risks of Artificial Intelligence in the Financial Services Sector The Consumer Financial Protection Bureau (CFPB) has issued a comment in response to a Request for Information (RFI) concerning the use, opportunities, and risks of artificial intelligence (AI) in the financial services sector. The CFPB acknowledges the potential benefits of AI, such as improving access to credit and enhancing customer service, but also highlights significant risks, including potential biases, privacy concerns, and the need for transparency in AI decision-making processes. The agency emphasizes the importance of ensuring that AI systems in financial services operate fairly and do not perpetuate discrimination. The CFPB is particularly focused on the responsible use of AI, urging industry stakeholders to prioritize consumer protection and to develop AI technologies that enhance, rather than undermine, fairness and accountability in financial services. This comment reflects the CFPB's proactive stance in addressing the evolving challenges and opportunities presented by AI in the financial sector, aiming to guide the development of regulations and practices that safeguard consumers while fostering innovation. Make sure you submit yours before the deadline!

• Responsible choices: a new policy for using AI in the Australian Government | Digital Transformation Agency The article from the Digital Transformation Agency (DTA) of Australia outlines a new policy for the responsible use of AI within the Australian government. The policy emphasizes ethical considerations, transparency, and accountability in AI deployment across government services. It includes guidelines to ensure AI is used in ways that respect human rights, prevent bias, and maintain public trust. The policy mandates that all AI applications must be explainable, ensuring that decisions made by AI can be understood and justified. Additionally, the policy calls for rigorous testing and monitoring of AI systems to prevent unintended consequences. This initiative reflects the Australian government's commitment to harnessing AI's benefits while safeguarding against potential risks, setting a standard for responsible AI use in the public sector.?

?The Australian government's new AI policy aligns with similar efforts in the EU and the U.S. to regulate and guide the ethical use of AI. Like the EU's AI Act, which focuses on transparency, accountability, and preventing bias, Australia's policy emphasizes ethical AI use, requiring explainability and rigorous oversight. In the U.S., while there's no comprehensive federal AI regulation yet, policies and guidelines increasingly stress the importance of ethical AI practices and transparency, similar to Australia's approach. All three regions are moving towards frameworks that aim to balance innovation with public trust and ethical considerations in AI deployment.

Gen AI for Business Trends, Concerns, and Predictions:?

The merging of AI and blockchain is expected to revolutionize industries by combining advanced data processing with enhanced security and transparency, though scalability and regulatory challenges remain. A report shows that while 90% of women view generative AI as crucial for career growth, only 35% feel equipped to use it, highlighting the need for targeted AI training programs. Companies are using AI to fight increasingly sophisticated online scams, emphasizing the ongoing battle between AI-driven scams and AI-powered defenses. There's a significant gap between AI expectations and outcomes in the workplace, with only 25% of companies seeing significant productivity improvements despite high expectations. The progress of large language models (LLMs) is slowing, prompting a shift toward optimizing existing technologies rather than creating larger models. Microsoft's report on generative AI in workplaces shows varying productivity gains across job functions, with tools like Copilot enhancing productivity but also adding cognitive load. Research introduces Bias-Aware Low-Rank Adaptation (BA-LoRA) to mitigate biases in LLMs during fine-tuning. Chief Legal Officers are increasingly responsible for managing AI-related compliance, ESG reporting, and cybersecurity, reflecting the growing complexity of their roles. A report indicates that more than 60% of companies lack policies on generative AI usage, stressing the need for guidelines to manage risks. The top 10 enterprise dilemmas around generative AI include managing ethical implications, data privacy, legal complexities, and justifying ROI. PwC's 2024 survey reveals that while AI usage is widespread, responsible AI practices are lagging, with only 35% of organizations fully implementing them. AI burnout is becoming an issue as workers struggle to keep up with AI-driven workloads, raising concerns about balancing productivity with employee well-being. Google DeepMind highlights the risks of generative AI misuse, calling for better safeguards and collaboration to prevent harmful applications like deepfakes and misinformation.

• The merging of AI and blockchain was inevitable – but what will it mean? - AI News The merging of AI and blockchain technologies is considered inevitable due to their complementary strengths. AI offers advanced data processing and decision-making capabilities, while blockchain ensures transparency, security, and decentralization. Together, they could revolutionize industries by enabling more secure, efficient, and trustworthy applications, particularly in finance, supply chain management, and data integrity. This convergence could also lead to innovations like decentralized AI marketplaces and enhanced data privacy. However, challenges like scalability and regulatory concerns remain.

• ?90% Women View Gen AI As Crucial For Career Growth, Only 35% Equipped To Use It: Report A report highlights that while 90% of women see generative AI as crucial for career growth, only 35% feel equipped to use it, potentially impacting their job opportunities and widening the gender wealth gap. The growing reliance on AI in the workplace means women without these skills may miss out on high-paying, future-oriented roles. To close this gap and ensure equitable wealth distribution, there is a pressing need for targeted AI training and education programs for women.

• ?Putting ‘Scam Dens’ Out of Business Means Using AI to Fight AI discusses the use of AI to combat online scams, highlighting how scammers increasingly use AI to create more sophisticated schemes. To counter this, companies are deploying AI tools to detect and prevent fraudulent activities more effectively. The focus is on the importance of staying ahead of scammers by leveraging AI-driven solutions that can quickly identify and mitigate risks, ultimately aiming to shut down these "scam dens" and protect consumers.The battle between AI-driven scammers and AI-powered defenses is ongoing, with both sides continuously evolving their tactics. While companies are making strides in using AI to detect and prevent fraud, scammers are also becoming more sophisticated by leveraging AI to create more convincing schemes. The outcome of this battle hinges on who can innovate faster—scammers with their evolving tactics, or companies with their defensive technologies.

• The gap between AI expectations and outcomes in the workplace are wide The data referenced in the article is from a survey of business leaders, where 92% expected AI to positively impact their companies, but only 25% saw significant productivity improvements. The survey also highlighted that 60% of companies face challenges integrating AI, particularly due to issues like data quality and lack of skilled personnel. These figures underline the gap between AI expectations and outcomes in the workplace.?

• LLM progress is slowing — what will it mean for AI? | VentureBeat discusses the deceleration in advancements of large language models (LLMs), showing diminishing performance improvements as models scale. It points out that while earlier models saw significant gains with increased parameters, recent models are experiencing reduced returns on investment in size and training. This slowdown may shift focus toward optimizing and applying existing AI technologies rather than creating new, larger models, and could lead to innovation in alternative AI areas, addressing issues like efficiency and real-world applicability.?

• Generative AI in Real-World Workplaces report by Microsoft discusses the real-world impact of generative AI, specifically through Microsoft's AI and Productivity research. It highlights that tools like Microsoft Copilot are already enhancing productivity across various job functions, though the benefits vary significantly by role. For instance, customer service and sales professionals reported more substantial productivity gains and higher job satisfaction compared to those in legal roles. The research also explored the cognitive load associated with using AI tools, finding that while tasks often feel less stressful and demanding with AI assistance, the cognitive impact can differ. A large-scale study involving over 6,000 employees across 60 organizations demonstrated notable productivity improvements when using Copilot in everyday work settings. These findings underscore the importance of understanding how AI tools can be best utilized across different job functions to maximize their effectiveness in the workplace.

• https://arxiv.org/abs/2408.04556v1 The research paper titled "Bias-Aware Low-Rank Adaptation: Mitigating Catastrophic Inheritance of Large Language Models" introduces a new method called Bias-Aware Low-Rank Adaptation (BA-LoRA). This method aims to mitigate bias in large language models (LLMs) during the fine-tuning process, which is typically computationally intensive. BA-LoRA incorporates three regularization terms—consistency, diversity, and singular vector decomposition—to enhance the model's performance while reducing inherited biases from pre-training data. The study shows BA-LoRA's effectiveness across various natural language tasks.?

• Gen AI is reporting for work duty but who’s in charge of it? | Legal Dive covers the growing responsibilities of Chief Legal Officers (CLOs) in managing the risks associated with generative AI, ESG (Environmental, Social, and Governance) reporting, and cybersecurity. It highlights how CLOs are increasingly expected to oversee AI-related compliance, ensure accurate ESG disclosures, and mitigate cybersecurity threats. The article notes that CLOs must balance these duties while maintaining legal and ethical standards, as regulatory scrutiny intensifies. The integration of generative AI into business practices adds complexity to their role, making it crucial for CLOs to stay informed and proactive in these areas. Do you know what your AI is doing today??

• More Than 60% of Companies Worldwide Don’t Have Policies On Gen AI Usage: Report - News18 discusses the need for companies to develop and implement policies governing the use of generative AI. It emphasizes the importance of these policies in addressing concerns related to data privacy, ethical use, intellectual property, and compliance with legal standards. The article highlights that as generative AI becomes more widely adopted in various sectors, businesses are increasingly recognizing the importance of having clear guidelines to manage the risks associated with AI-generated content. Companies are also focusing on ensuring transparency and accountability in how AI tools are used, both internally by employees and externally in customer-facing applications. The move towards formalizing generative AI usage policies is seen as crucial for maintaining trust and safeguarding against potential misuse or legal challenges. Does your company have an AI policy??

• ?Top 10 Enterprise Gen AI Dilemmas Confronting the C-suite | HackerNoon These challenges include managing the ethical implications of AI-generated content, ensuring data privacy and security, and navigating the legal complexities surrounding intellectual property rights. Executives are also grappling with the integration of AI into existing workflows without disrupting operations, as well as the need to upskill employees to work alongside AI tools. Another significant concern is the potential for AI to produce biased or inaccurate outputs, which could harm the company's reputation or lead to regulatory scrutiny. Additionally, there is the challenge of balancing innovation with compliance, particularly in industries with strict regulatory environments. The article also highlights the difficulties in justifying the ROI on AI investments, managing the cultural shift within organizations as AI becomes more prevalent, and ensuring transparency in AI decision-making processes. These dilemmas reflect the broader tension between leveraging AI's transformative potential and mitigating the risks associated with its adoption in the enterprise environment.??

• PwC's 2024 US Responsible AI Survey reveals that while over 75% of organizations are using AI, only 35% have fully implemented responsible AI practices, which include managing issues like bias, transparency, and accountability. The survey highlights that 60% of executives are concerned about reputational risks associated with AI, yet many companies still lack the necessary governance frameworks. The report indicates that as AI adoption grows, there will be increasing pressure on organizations to establish robust, responsible AI practices to meet regulatory and stakeholder expectations.??

• Yes, AI burnout is already happening at work. Here's how to prevent it highlights the growing issue of AI burnout among workers, as companies increasingly implement AI tools to boost productivity. While AI can streamline tasks and improve efficiency, the pressure to keep up with AI-driven workloads is leading to significant stress and fatigue among employees. Many workers feel overwhelmed by the rapid pace of AI integration and the constant demand to adapt to new technologies. This raises an important question: As we embrace the benefits of AI in the workplace, how do we ensure that our teams aren’t burning out in the process? Are we doing enough to balance productivity gains with the well-being of our employees??

• Mapping the misuse of generative AI - Google DeepMind explores the potential risks and misuses of generative AI technologies. It discusses how generative AI can be exploited in harmful ways, such as creating deepfakes, spreading misinformation, or generating malicious content. The article emphasizes the importance of understanding these risks to develop better safeguards and policies that can prevent misuse. DeepMind also highlights the need for collaboration between researchers, policymakers, and industry leaders to create a framework that ensures generative AI is used ethically and responsibly. The discussion includes potential strategies for monitoring and mitigating the risks associated with the widespread deployment of these technologies.?

News and updates around? finance, Cost and Investments

Developing an AI chat app like Character AI can cost between $40,000 to $300,000, depending on factors such as complexity, features, and location of the development team. Investors in the consumer goods sector are seeing ROI in generative AI through enhanced product development, marketing, and customer engagement, leading to improved efficiency, cost savings, and revenue growth. Major companies like Microsoft, Google, and Amazon are heavily investing in generative AI, with projections showing the global cloud AI market could reach $52 billion by 2025 and the overall generative AI market growing to $110.8 billion by 2030. Google Cloud's generative AI services are driving significant ROI for businesses, with improvements ranging from 20% to 40% in areas like customer service and content creation. AI startup deal-making is gaining momentum, with over $40 billion in AI startup deals announced in the first half of 2024, as tech giants like Microsoft, Google, and Amazon aggressively acquire startups to bolster their AI capabilities and secure a competitive edge in the rapidly growing AI market.

• Cost to Build an App Like Character AI: AI Chat App Development Developing an AI chat app like Character AI can cost between $40,000 to $300,000, depending on complexity, features, and team location. Key factors influencing cost include the AI model's complexity, data gathering, personalization, UI design, and compliance with security standards. The development process involves market research, defining features, choosing the tech stack, and ongoing maintenance. Integrating advanced technologies such as NLP, voice recognition, and emotion analysis is essential for creating a dynamic and engaging user experience.?

• Where Generative AI Investors are Finding ROI | Consumer Goods Technology discusses how investors in the consumer goods sector are finding returns on investment (ROI) in generative AI technologies. It highlights that AI is being leveraged to enhance product development, marketing, and customer engagement. Key areas of focus include personalized recommendations, automated content creation, and supply chain optimization. The article also notes that companies are seeing ROI through improved efficiency, cost savings, and increased revenue as AI-driven strategies become more integrated into business operations. Investors are particularly interested in AI's ability to drive innovation and create competitive advantages in a rapidly evolving market.??

• Gen AI: Who is spending what and where will revenues come from By Investing.com details the substantial investments in generative AI (Gen AI) by major technology companies and the anticipated revenue streams. Microsoft has invested over $13 billion in its partnership with OpenAI, aiming to integrate AI across its product suite, including Azure and Office 365. Google has allocated around $20 billion for AI research and infrastructure, focusing on integrating AI into its cloud services and search algorithms. Amazon Web Services (AWS) is also heavily investing in AI, with an estimated $10 billion directed towards AI and machine learning tools.

Revenue projections indicate that AI-driven cloud services will be a major growth area, with the global cloud AI market expected to reach $52 billion by 2025. Additionally, AI-enhanced productivity tools and consumer applications are projected to generate significant income, with the overall generative AI market anticipated to grow to $110.8 billion by 2030. Companies are banking on these investments to secure a leading position in the rapidly expanding AI market, with revenues expected to surge across cloud computing, software subscriptions, and AI-powered consumer products.

• ?Google Cloud: Gen AI Driving Significant ROI for Businesses | Technology Magazine reports that Google Cloud's generative AI services are driving significant return on investment (ROI) for businesses across various industries. According to the report, companies using Google Cloud's generative AI tools have seen ROI improvements ranging from 20% to 40% in areas such as customer service, content creation, and product development. The article highlights case studies where businesses have leveraged AI to automate tasks, enhance customer engagement, and accelerate innovation, leading to substantial cost savings and revenue growth. Google Cloud’s AI platform, which includes tools like Vertex AI and custom AI models, is noted for its scalability and ease of integration, making it accessible for both large enterprises and smaller companies. The article underscores the growing adoption of generative AI as a critical component of digital transformation strategies, with Google Cloud positioned as a key player in this space.??

• AI Startup Deal-Making Gaining Momentum — Is That Good? reports on the surge in deal-making activity among AI startups, with major tech companies like Microsoft, Google, and Amazon leading the charge. According to the article, AI-related mergers and acquisitions (M&A) have seen a significant uptick, with over $40 billion in AI startup deals announced in the first half of 2024 alone. Microsoft, Google, and Amazon are actively acquiring AI startups to bolster their AI capabilities, focusing on areas like generative AI, machine learning, and cloud-based AI services. The article highlights that these tech giants are not only acquiring startups but also forming strategic partnerships and investing heavily in AI research and development. This aggressive investment strategy is driven by the growing demand for AI technologies across various industries, as companies seek to integrate AI into their products and services to gain a competitive edge. The M&A activity is expected to continue as the race to dominate the AI sector intensifies among the world's largest technology companies.??

What/where/how Gen AI solutions are being implemented today?

JPMorgan Chase has introduced an AI assistant powered by OpenAI's technology to help employees efficiently access and interpret data, enhancing productivity and decision-making within the bank. In pediatric mental health care, Dr. David Idel highlights how AI is transforming diagnosis and treatment by identifying early signs of issues like anxiety and depression, emphasizing the importance of integrating AI with human expertise. Businesses can leverage AI as a partner by automating customer service, improving content creation, and assisting in strategic decision-making through data analysis. A report from PCWorld shows that AI chatbots are primarily used for customer support and sales, though they struggle with complex interactions. Walmart executives revealed that generative AI is 100 times more productive than humans at updating product pages, allowing for rapid adjustments and improved customer engagement. IBM and the USTA enhanced the 2024 US Open experience with AI features that deliver personalized match insights and predictions, showcasing AI's impact on live sports. At the Paris 2024 Olympics, AI played a transformative role, with contributions from Google, IBM, Alibaba, and Intel, making the games more interactive and personalized for viewers through advanced analytics, AI-generated content, and immersive experiences.

• JPMorgan Chase is giving its employees an AI assistant powered by ChatGPT maker OpenAI JPMorgan Chase has launched an artificial intelligence (AI) assistant, utilizing technology similar to OpenAI's ChatGPT, to help employees access and interpret research, data, and client information more efficiently. This move reflects the bank's ongoing efforts to integrate AI into its operations to enhance productivity and decision-making. The AI assistant aims to streamline internal processes by providing quick, accurate responses to employee inquiries, leveraging JPMorgan's extensive data resources. This development underscores the growing trend of financial institutions adopting AI to optimize their services.?

• https://www.ama-assn.org/practice-management/digital/advancements-pediatrics-uses-artificial-intelligence-mental-health Dr. David Idel, a pediatrician, discusses how AI is transforming pediatric mental health care by improving diagnosis and treatment. AI tools can identify early signs of mental health issues like anxiety and depression, enabling more personalized care. He highlights the importance of integrating AI with human expertise to address the mental health crisis among children. The interview also covers the ethical considerations and the potential for AI to revolutionize mental health care in pediatrics by providing more timely and accurate assessments.

• 3 ways you can use AI as your business partner The article provides specific ways AI can be integrated into business operations. First, it highlights how AI can automate customer service by handling queries via chatbots, freeing up human resources for more complex issues. Second, AI can improve content creation by generating tailored marketing copy and social media posts. Finally, AI tools can assist in strategic decision-making by analyzing large datasets to identify trends and opportunities, helping businesses stay competitive in dynamic markets.?

• New report shows the truth of how people actually use AI chatbots | PCWorld A new report reveals that chatbots are most commonly used for customer support, with 47% of companies deploying them for tasks like answering questions and resolving issues. Additionally, 40% of businesses use chatbots for sales and marketing, helping customers with product recommendations and purchases. The report also notes that while chatbots handle simple tasks effectively, they often struggle with more complex interactions, leading to mixed satisfaction levels among users.?

• Walmart executives say generative AI is 100 times more productive at updating product pages than people - Modern Retail? highlights statements made by Walmart executives regarding the productivity gains achieved through the use of generative AI in updating product pages. According to Walmart, generative AI is 100 times more productive than human workers when it comes to creating and updating content on product pages. This dramatic increase in efficiency is attributed to AI's ability to quickly generate, test, and optimize product descriptions, images, and other content elements at scale. The technology is being used to ensure that product information is up-to-date, accurate, and tailored to customer preferences, leading to improved customer engagement and potentially higher sales. Walmart executives also noted that AI allows for rapid adjustments in response to market trends and inventory changes, which is crucial in the fast-paced retail environment. This shift toward AI-driven content management reflects a broader trend in the retail industry, where companies are increasingly relying on AI to automate and enhance various operational processes.?

• IBM, USTA Serve Up Enhanced Generative AI Features at 2024 US Open IBM and the USTA have teamed up to bring enhanced generative AI features to the 2024 US Open, taking the fan experience to a whole new level. This year, AI is not just keeping score—it’s generating real-time match insights, personalized player stats, and even custom highlights, all tailored to each fan's preferences. IBM's AI is powering interactive features that allow viewers to dive deep into match analytics, receive predictions, and get updates like never before. It’s like having a personal sports analyst right in your pocket, making the US Open more engaging and immersive for fans worldwide. This collaboration showcases how AI is transforming live sports, making every serve, volley, and point more thrilling and personalized.

At the Paris 2024 Olympics, generative AI stole the show played a transformative role, with contributions from multiple major tech companies. Google's Gemini AI system provided real-time insights and personalized content, enhancing fan engagement by offering detailed athlete performance analysis and predictive analytics tailored to individual viewer preferences. IBM Watson continued its legacy in sports AI, delivering AI-generated commentary and in-depth match analysis, while Alibaba used its AI-powered cloud services to create instant, personalized highlight reels, making it easier for fans to relive key moments. Intel also played a crucial role with its AI-driven athlete tracking and 3D simulations, which provided deeper insights into athletic performance and event predictions. This multi-faceted integration of AI from different tech giants at the Paris Olympics set a new standard for how AI can enhance the viewing experience, making the games more interactive, personalized, and engaging for audiences worldwide.

Women Leading in AI?

Tune in for a conversation on “AI's Impact on Healthcare and Demystifying AI Terms with Neha Goel.” She shares her personal career journey, including battling gender bias in the tech industry. Neha demystifies common AI terms and introduces the concept of RAG and its use in generating relevant information using large language models and external knowledge. AI's Impact on Healthcare and Demystifying AI Terms with Neha Goel by Women And AI??

I am so honored to be this week’s Featured AI Leader: ??Women And AI’s Featured Leader - Eugina Jordan ??Your author of this amazing AI newsletter! She is always featuring everyone else so the Women And AI community wanted to feature he for a change!?

SO GRATEFUL for this – thank you.?

Learning Center and How To’s

• CodexGraph: An Artificial Intelligence AI System that Integrates LLM Agents with Graph Database Interfaces Extracted from Code Repositories - MarkTechPost introduces CodexGraph, an AI system that integrates Large Language Model (LLM) agents with graph database interfaces derived from code repositories. This integration enables more efficient and intelligent querying of software code, enhancing tasks like code search, analysis, and recommendation. By combining the capabilities of LLMs and graph databases, CodexGraph aims to improve the management and understanding of complex software projects.??

• Free generative API for learning The OpenAI community post discusses a free generative API designed for educational purposes, allowing users to explore and learn about AI capabilities. This API is intended for those interested in experimenting with generative models, offering an accessible entry point into AI learning and development. The focus is on providing a resource for learning rather than commercial use, emphasizing its role in education and skill-building.

• ?Crab Framework Released: An AI Framework for Building LLM Agent Benchmark Environments in a Python-Centric Way - MarkTechPost The CRAB framework is a newly released AI tool designed for building benchmark environments specifically for Large Language Model (LLM) agents using Python. It allows developers to create standardized testing environments for evaluating LLMs, focusing on ease of use within the Python ecosystem. This framework supports the development and benchmarking of AI agents, aiming to streamline the process for researchers and developers working with LLMs.?

• Quantization-Aware Training for Large Language Models with PyTorch introduces Quantization-Aware Training (QAT), a technique that allows deep learning models to be quantized during training, rather than after, to maintain higher accuracy in low-precision models. QAT simulates the effects of quantization (converting weights and activations from floating-point to lower precision like 8-bit integers) during the training process, enabling the model to learn how to compensate for the reduced precision. This approach leads to more efficient models that are smaller and faster while maintaining performance close to that of full-precision models. QAT is particularly beneficial for deploying AI models on resource-constrained devices, where computational power and memory are limited. The article also details how PyTorch supports QAT, providing tools and APIs to integrate it into existing workflows seamlessly.?

• How to Prune and Distill Llama-3.1 8B to an NVIDIA Llama-3.1-Minitron 4B Model | NVIDIA Technical Blog explains the process of pruning and distilling the LLaMA 3.1 8B model to create a more efficient and smaller version called the NVIDIA LLaMA 3.1 MiniTron 4B model. Pruning involves removing less important neurons and connections in the neural network to reduce the model's size without significantly impacting its performance. Distillation refers to transferring knowledge from the larger model to a smaller one, allowing the smaller model to retain much of the larger model's capabilities. The resulting MiniTron 4B model is optimized for faster inference and lower resource usage while maintaining high accuracy in tasks like natural language processing. This streamlined model is particularly useful for deployment in environments with limited computational resources, making it a practical choice for real-time AI applications.??

Prompt of the week

Prompt Design at Character.AI To design effective prompts for AI on Character.AI, start by clearly defining the goal of the interaction. Provide specific context and set expectations for the AI's behavior. Focus on being precise in your wording to guide the AI effectively. Test different versions of your prompt, analyze the responses, and iterate until the desired outcome is achieved. This process involves refining prompts to align with the AI's strengths and the intended results.??

Tools and Resources

• 22 Generative AI Workplace Tools And How To Use Them | Bernard Marr The article outlines 22 generative AI tools enhancing workplace productivity. These include Jasper for marketing copy, GitHub Copilot for coding, and Synthesia for video creation. ChatGPT and Copy.ai assist with text generation, while Lumen5 converts text into videos. Other tools are Notion AI for task management, Grammarly for writing, Murf.ai for voiceovers, Runway for video editing, and Descript for audio editing. The list also features tools like Frase for SEO content, Writesonic for copywriting, and Pictory for video creation from scripts. Each tool automates tasks, allowing focus on more strategic work. Which ones have you used?

?

• AI Risk Repository The MIT AI Risk Management (AIRisk) website provides information about the AI Risk Management framework developed by the Massachusetts Institute of Technology (MIT). The framework is designed to help organizations identify, assess, and mitigate risks associated with the deployment of artificial intelligence technologies. It covers various aspects of AI risk, including data privacy, security, bias, and the potential for unintended consequences. The website offers resources, tools, and guidelines to assist businesses and policymakers in implementing effective AI risk management practices. The goal is to ensure that AI systems are developed and used responsibly, with a focus on minimizing harm and maximizing societal benefits.?

• Mapify https://mapify.so/ – an AI-powered app designed to transform any content into clear and concise mind maps, making it easy to capture and organize knowledge on the go. Developed by Xmind, Mapify offers enhanced AI features for summarizing documents, articles, videos, and web pages into visual maps. Fun!?

• The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery – Sakana, an advanced AI system designed to assist in scientific research and discovery. This AI platform is capable of autonomously generating hypotheses, designing experiments, and analyzing data, effectively simulating the scientific method. The AI Scientist aims to accelerate research by identifying patterns and insights that might be missed by human researchers, allowing for faster advancements in fields such as biology, chemistry, and materials science. The platform is designed to work alongside human scientists, providing them with powerful tools to enhance their research capabilities and drive innovation in various scientific disciplines.?

• GitHub - facebookresearch/unibench: Python Library to evaluate VLM models' robustness across diverse benchmarks The GitHub repository for UniBench by Facebook Research provides a benchmark suite designed to evaluate the performance and robustness of large-scale language models across diverse tasks. UniBench includes a collection of tasks that measure model capabilities in areas such as natural language understanding, reasoning, and generation. The benchmark is part of Facebook's ongoing efforts to standardize the evaluation of AI models, ensuring that they are tested against a wide range of challenges to better understand their strengths and weaknesses. UniBench aims to provide a comprehensive and rigorous assessment framework that can be used by researchers and developers to improve the development of more reliable and capable AI systems.?

If you enjoyed this newsletter, please comment and share. If you would like to discuss a partnership, or invite me to speak at your company or event, please DM me.