Gen AI for Business #11

Welcome to this week's edition of Gen AI for Business! We've packed this newsletter with the latest trends, tools, and innovations in generative AI. From Apple and Meta's potential AI partnership to McKinsey’s bold AI-driven strategies, and groundbreaking research on energy-efficient AI models, there's a lot to explore. Discover new models, practical how-to guides, and insightful investment tips. Join us on this AI adventure, stay informed on regulatory updates, and don’t forget to share with your colleagues and friends!

Here’s what you need to know about Gen AI from last week.

Enjoy your reading! And if you found it useful, please like and share with your friends and colleagues.

Eugina

News about models and everything related to them

Meta has split Reality Labs into "Wearables" and "Metaverse" units, focusing on VR/AR hardware and virtual environments. Despite layoffs, Meta invests in AI with new models like Chameleon and JASCO. The Meta Large Language Model Compiler optimizes code using pre-trained models. Amazon’s "Metis" chatbot aims to compete with ChatGPT. OpenAI delays its advanced Voice Mode. Diffusion models, like SEDD, offer NLP advancements. Hugging Face’s Open LLM Leaderboard v2 improves model evaluations. Alibaba's Qwen leads AI benchmarks. Researchers eliminate matrix multiplication in LLMs to cut costs and impact. Claude advances in coding.

• Meta restructures Reality Labs amid AI push and losses Meta has restructured its Reality Labs division, splitting it into "Wearables" and "Metaverse" units. The Wearables unit will focus on hardware such as VR headsets and AR glasses, while the Metaverse unit will concentrate on software and virtual environments. Despite workforce reductions, Meta is heavily investing in AI, unveiling five new AI models, including Chameleon for mixed-modal understanding and JASCO for text-to-music generation. This restructuring and AI focus underscore Meta’s commitment to advancing its metaverse and AI initiatives despite financial losses.

• Meta Large Language Model Compiler: Foundation Models of Compiler Optimization | Research - AI at Meta? The Meta Large Language Model Compiler (LLM Compiler) is a suite of pre-trained models designed for code optimization tasks, built on Code Llama. It understands compiler intermediate representations (IRs), assembly language, and optimization techniques. Trained on 546 billion tokens of LLVM-IR and assembly code, it includes models with 7 billion and 13 billion parameters. Fine-tuned versions improve code size optimization and disassembly from x86_64 and ARM assembly back to LLVM-IR, achieving significant optimization potential. Released under a commercial license, it aims to support further research and development in compiler optimization.?

• Report: Amazon developing AI chatbot that would compete with ChatGPT and others – GeekWire Amazon is reportedly developing "Metis," an AI chatbot designed to rival ChatGPT and other AI assistants. Leveraging a new foundational model, Metis aims to provide advanced conversational capabilities through web browsers. This initiative is part of Amazon's broader AI strategy, which includes the launch of "Remarkable Alexa," a revamped version of Alexa potentially offering a paid tier. These advancements highlight Amazon's commitment to staying competitive in the AI landscape and enhancing user experience through innovative technology. The name "Metis" has significant connotations. In Greek mythology, Metis was a Titaness associated with wisdom, craft, and skill. She was known for her deep knowledge and cunning intelligence. Naming the AI chatbot "Metis" likely signifies Amazon's intention to imbue it with advanced intelligence, wisdom, and problem-solving capabilities, aiming to make it a sophisticated competitor in the AI chatbot market. This choice reflects the aspiration to create a tool that is not only smart but also wise and resourceful.?

• OpenAI is delaying roll out of the advanced Voice Mode OpenAI on X: "We're sharing an update on the advanced Voice Mode we demoed during our Spring Update, which we remain very excited about: We had planned to start rolling this out in alpha to a small group of ChatGPT Plus users in late June, but need one more month to reach our bar to launch." / X?

• The Rise of Diffusion-Based Language Models: Comparing SEDD and GPT-2 - MarkTechPost? The article explores the limitations of traditional autoregressive models, such as slow processing and exposure bias. Diffusion models are introduced as a potential solution. While SEDD performs well in many areas, there's still space for improvement in diversity and conditional generation. The emergence of diffusion-based language models is an exciting development in the field of natural language processing. Their potential to overcome the limitations of autoregressive models is significant. However, as the study suggests, there's still work to be done, particularly in areas like diversity and conditional generation. It will be interesting to see how diffusion-based models evolve and how they can be leveraged to create even more powerful and versatile language models.?

• Hugging Face Releases Open LLM Leaderboard 2: A Major Upgrade Featuring Tougher Benchmarks, Fairer Scoring, and Enhanced Community Collaboration for Evaluating Language Models - MarkTechPost The Hugging Face Open LLM Leaderboard v2 marks a significant advancement in evaluating language models, addressing key challenges from its predecessor. Initially popular but increasingly limited by benchmark saturation, the v2 introduces six new benchmarks like MMLU-Pro and GPQA to test advanced reasoning and knowledge capabilities. A major improvement is the adoption of normalized scoring, ensuring fairer comparisons across benchmarks of varying difficulty. Collaborative efforts with EleutherAI enhance reproducibility, while a new interface by Gradio promises faster user interactions. The leaderboard now includes community-voted models, prioritizing relevance and fostering innovation in language model development.

• Chinese AI models storm Hugging Face's LLM chatbot benchmark leaderboard — Alibaba runs the board as major US competitors have worsened | Tom's Hardware Big win for China in the world of AI! Alibaba's Qwen large language models (LLMs) dominated the latest leaderboard from Hugging Face, a popular AI benchmark platform. These LLMs excelled in tasks that require reasoning, knowledge, and even following instructions – skills crucial for real-world applications. Interestingly, some major US competitors actually saw their LLM performance dip. This suggests a potential overspecialization for past benchmarks. The takeaway? Well-rounded training is key for building truly powerful LLMs. In the section on regional updates, learn what other steps Alibaba is taking to dominate the market.??

• Researchers upend AI status quo by eliminating matrix multiplication in LLMs | Ars Technica Researchers from the University of California Santa Cruz, UC Davis, LuxiTech, and Soochow University have developed a novel method to run AI language models more efficiently by eliminating matrix multiplication, a core operation in neural networks. This new approach has the potential to significantly reduce both the environmental impact and operational costs of AI systems. By creating a custom 2.7 billion parameter model that performs similarly to conventional large language models without using matrix multiplication, they challenge the prevailing belief that such operations are indispensable. Additionally, they demonstrated a 1.3 billion parameter model running on a GPU accelerated by a custom-programmed FPGA chip consuming only 13 watts of power. This breakthrough could make large language models more accessible, efficient, and sustainable, especially for deployment on resource-constrained hardware like smartphones.??

• In a recent tweet, Alex Albert highlighted the impressive advancements of Claude, a leading language model, in coding and autonomously fixing pull requests. These developments suggest that within a year, a significant portion of software code could be generated by large language models (LLMs). This shift indicates a transformative change in the software development industry, where AI tools will increasingly take over coding tasks, enhancing efficiency and potentially redefining the roles of human developers. https://x.com/alexalbert__/status/1803804677701869748?

• An LLM Approach For Large-Scale SoC Security Verification And Policy Generation (U. of Florida) The University of Florida has developed an innovative approach using large language models (LLMs) for large-scale System-on-Chip (SoC) security verification and policy generation. This method leverages LLMs' advanced capabilities in understanding and processing complex data to enhance security protocols and verification processes in SoC designs. The approach aims to improve the efficiency and accuracy of security measures, addressing the growing complexities and challenges in modern SoC development. Download here: SoCureLLM: An LLM-driven Approach for Large-Scale System-on-Chip Security Verification and Policy Generation??

Gen AI news from different industries

generative AI is revolutionizing various industries: over 200,000 Amazon sellers are using AI for product listings; the U.S. Army is piloting AI for logistics; fintech is enhancing fraud detection and customer service; accounting is automating tasks to allow more advisory roles; finance is using AI for risk assessment and operations; pharma is accelerating drug discovery; healthcare is improving diagnostics and patient care with AI; and manufacturing is integrating AI for better efficiency and sustainability. Despite challenges, AI continues to drive innovation and efficiency across these sectors.

eCommerce and Retail

• Over 200,000 Amazon Sellers Have Used Gen-AI Listing Tools - EcommerceBytes Over 200,000 Amazon sellers are now using generative AI tools to create and enhance product listings. These tools streamline the process by generating high-quality descriptions, titles, and other details from minimal input, such as a brief description, an image, or a URL. This technology significantly reduces effort and improves product discoverability, particularly benefiting small businesses. Many sellers accept the AI-generated content with minimal edits, enhancing their efficiency and the customer shopping experience.

Defense?

• Army to Launch GenAI Pilot Amid Security Concerns – MeriTalk The U.S. Army is set to launch a generative AI pilot this summer to explore its applications within its acquisition and logistics workforce. This initiative aims to improve efficiency in tasks like contract writing and data analysis. The pilot will use a large language model trained on Army-specific data and include measures to address security concerns, such as ensuring a "human in the loop." This effort is part of a broader strategy to integrate AI tools while mitigating associated risks and improving operational use.

Fintech

• The Future of Gen AI in FinTech, Finserv and Payments? Generative AI is revolutionizing the fintech industry by enhancing efficiency, agility, and personalization. It is used for tasks like fraud detection, customer service, and payment processing. Financial institutions are increasingly integrating Gen AI to improve operations and customer experiences. Innovations include personalized payment methods, advanced data analysis, and the use of AI for cybersecurity. The rapid adoption of Gen AI in fintech is driven by its ability to innovate and streamline various processes, despite challenges like regulatory compliance.

Accounting

• Death, Taxes, and AI: How Generative AI Will Change Accounting | Andreessen Horowitz Generative AI is set to revolutionize accounting by automating repetitive tasks, enhancing data collection, and improving client services. AI can streamline processes like data reconciliation, research, report generation, and client advice, addressing the anticipated shortage of accounting professionals. While generative AI excels in natural language tasks, it still requires human judgment for complex analyses. The technology promises to increase efficiency and allow accountants to focus on higher-value advisory roles, transforming client relationships from transactional to ongoing engagements.?

Finance

• Artificial Intelligence in Finance: Applications and Benefits – Business – automating tasks like risk assessment, fraud detection, and trading strategies using advanced algorithms such as machine learning and natural language processing. From enhancing personal finance management through smart budgeting tools to optimizing back-end operations with ERP integrations and chatbot-driven customer service, AI promises increased efficiency and profitability. Despite challenges like legacy systems and data quality issues, embracing AI is crucial for financial institutions aiming to meet the digital demands of today's consumers and stay competitive in a rapidly evolving market landscape.?

• [2406.11903] A Survey of Large Language Models for Financial Applications: Progress, Prospects and Challenges The paper "A Survey of Large Language Models for Financial Applications: Progress, Prospects and Challenges" explores the impact of LLMs in the financial sector. It highlights their capabilities in understanding context, processing vast data, and generating content. The survey categorizes applications into areas like linguistic tasks, sentiment analysis, and financial reasoning, discussing methodologies such as textual and knowledge-based analysis. It also provides resources for researchers and outlines future research challenges and opportunities to enhance the adoption and development of LLMs in finance.

Pharma

• How generative AI is shaping drug discovery Generative AI is poised to revolutionize drug discovery by significantly enhancing efficiency and innovation. NVIDIA's BioNeMo platform and Merck's use of Variational AI's Enki platform exemplify how this technology can accelerate the development of new drugs. BioNeMo offers tools for tasks like protein structure prediction and molecular optimization, reducing the need for extensive experimentation, while Enki generates novel small molecules that meet specific target profiles, streamlining the discovery process. These advancements promise faster, cost-effective drug development, although regulatory and security challenges remain. The full potential of generative AI in healthcare continues to unfold, signaling a transformative era for the pharmaceutical industry.?

• Capturing the genAI boom for drug development - Pharmaceutical Technology discusses the surge of interest and investment in generative AI (genAI) within the drug development sector. GenAI is enhancing efficiency in drug discovery, clinical trials, and overall pharma operations. Companies like Bristol Myers Squibb, Sanofi, and Genentech are making significant deals to integrate AI technologies. However, challenges regarding transparency, regulatory compliance, and ethical use remain. Despite these hurdles, investment in AI-centric biotech firms continues to grow, driven by the technology's potential to revolutionize the industry.??

Healthcare

• Exploring Gen AI: What Could Ultrasound Look Like in a ChatGPT World? | GE HealthCare (Singapore) Generative AI is poised to transform ultrasound imaging by enhancing diagnostic precision, automating routine tasks, and supporting clinical decision-making. It can guide real-time image acquisition, improve image quality by reducing noise, and generate detailed reports. Additionally, generative AI can help analyze extensive patient datasets, provide training simulations, and optimize operational processes, ultimately improving patient care and reducing clinician workload. This integration necessitates careful ethical considerations and collaboration among stakeholders to harness AI's full potential in healthcare.??

• AI Helps Predict Dementia Using Speech Patterns | MedPage Today A new study shows promise in using AI to detect Alzheimer's disease early. Researchers analyzed speech patterns from neuropsychological exams and found that AI could identify patients with mild cognitive impairment who were likely to develop Alzheimer's within six years. This could lead to a non-invasive screening tool for early detection, but more studies are needed.??

Manufacturing

• Convergence of AI, Sustainability in the Manufacturing Sector iscusses the integration of AI and sustainability in the manufacturing sector. It highlights how AI-driven solutions are being utilized to improve energy efficiency, reduce waste, and enhance overall sustainability. Key applications include predictive maintenance, optimized resource management, and the development of smart factories. These advancements not only contribute to environmental goals but also enhance operational efficiency and cost savings for manufacturers.?

Regional and regulatory updates

The "Responsible AI Playbook for Investors" by the World Economic Forum emphasizes the role of investors in promoting responsible AI practices, encouraging engagement with corporate boards to implement strong governance frameworks and standards. A UiPath study reveals that 3 in 5 workers in Singapore use generative AI, saving time but facing security and compliance concerns. OpenAI plans to block tool access in China from July. The U.S. is moving to restrict investments in China's AI sector to safeguard national security. A GAO report highlights generative AI's applications and risks, stressing the need for oversight. China leads with over 4,500 AI companies, while Alibaba opens its AI model platform internationally. Apple delays AI features in Europe due to regulatory concerns. Tech Mahindra launches Project Indus to support Indic languages with AI. OpenAI and Microsoft face a lawsuit from the Center for Investigative Reporting over copyright infringement.

• Responsible AI Playbook for Investors 2024 | World Economic Forum The "Responsible AI Playbook for Investors" by the World Economic Forum highlights the critical role of investors in promoting responsible AI (RAI) practices. It underscores the importance of engaging with corporate boards and stakeholders to implement strong governance frameworks and RAI standards. The playbook provides guidance on balancing AI's potential with mitigating risks, enhancing customer trust, and complying with regulations, ultimately driving growth and ensuring ethical AI development.??

• 3 in 5 Singapore workers using generative AI | HRD Asia A study by UiPath found that 3 in 5 workers in Singapore use generative AI in their jobs. The technology helps 62% of users save time, with 42% saving 10 or more hours per week. However, 65% do not combine generative AI with business automation. Concerns include security risks (38%), inaccurate outputs (34%), and compliance issues (31%). The report suggests that leveraging both AI and automation can boost productivity and work-life balance.?

• VIDEO: https://www.youtube.com/watch?v=ZyPjGAeUYbQ OpenAI has warned developers in China it will begin blocking their access to its tools and software from July, local media reported, suggesting the ChatGPT creator is taking a more active stance to bar users from nations where it doesn’t offer services.

• U.S. is closer to curbing investments in China's AI, tech sector? The U.S. is advancing towards implementing regulations that would restrict certain investments in China's AI and technology sectors to safeguard national security. Draft rules issued by the U.S. Treasury Department propose banning or requiring notification of investments that could aid China's development of advanced technologies, including AI, semiconductors, and quantum computing. These measures aim to prevent U.S. resources from enhancing China's technological capabilities, particularly those that could be leveraged for military purposes. The regulations are expected to be finalized by the end of the year, with public comments open until August 4. This move highlights the increasing scrutiny and due diligence required from U.S. investors engaging with Chinese tech sectors, ensuring that national security concerns are addressed comprehensively.?

• Artificial Intelligence: Generative AI Technologies and Their Commercial Applications | U.S. GAO The U.S. Government Accountability Office (GAO) released a report on the potential applications and risks of generative AI technologies. These AI systems, like ChatGPT, can create diverse content and have broad uses in healthcare, education, and business. However, they also pose risks such as spreading disinformation and national security threats. The report emphasizes the need for robust oversight and ethical guidelines to manage these technologies effectively. The GAO plans further studies to explore best practices and the societal impacts of generative AI.?

• China has more than 4,500 companies in field of artificial intelligence | TV BRICS, 23.06.24 China now boasts over 4,500 companies in the field of artificial intelligence, significantly contributing to the nation's technological advancements. This data, presented by Vice Minister Shan Zhongde at the World Intelligent Expo 2024 in Tianjin, highlights the rapid integration of AI across various sectors, including smart manufacturing, healthcare, and transportation. The rise of AI companies is reshaping global competition, with China leading in national and provincial-level digital enterprises, emphasizing the transformative impact of AI technologies on everyday life. In comparison, in 2024, the United States hosts around 16,800 AI companies, reflecting the country's leading position in AI talent, infrastructure, and research. This growth is part of a broader trend, with AI companies doubling since 2017. Meanwhile, Europe has a significant AI presence with around 10,000 AI companies, with key hubs in the UK, Germany, and France. Both regions are investing heavily in AI technology, contributing to advancements across various sectors, including healthcare, finance, and manufacturing.?

• Alibaba Opens AI Model Platform to International Users? In a move to expand its reach, Alibaba has opened its AI model platform, ModelScope, to international users. ModelScope is a collection of generative AI models, primarily trained on Chinese language data. The platform includes models developed by Alibaba itself, alongside offerings from other Chinese companies. By making ModelScope available to a wider audience, Alibaba hopes to attract more developers to its cloud platform.?

• Alibaba Cloud's 'AI programmer' gets mixed reactions from real programmers Alibaba Cloud has introduced an upgraded "AI programmer" powered by its proprietary large language model. This tool aims to assist developers by automating code generation and improving efficiency. While the AI programmer is designed to handle various programming tasks, its reception among real programmers is mixed. Some appreciate the potential for increased productivity, while others express concerns about the quality and reliability of AI-generated code. This innovation is part of Alibaba's broader strategy to monetize its cloud services and enhance its technological offerings. Have you tried this??

• Chinese AI firms woo OpenAI users as US company plans API restrictions | Reuters The reaction to OpenAI's decision to cut access to its tools for developers in China and other regions has been mixed. Local developers and tech communities are concerned about the impact on their AI projects and the potential slowdown in innovation due to restricted access to crucial AI resources. Some view it as a setback in technological collaboration, while others are looking for alternative solutions and tools to continue their work.??

• Apple says it won't roll out AI features in Europe due to regulatory concerns Apple has decided to delay the rollout of its new AI features in Europe due to regulatory uncertainties related to the EU's Digital Markets Act (DMA). The DMA, designed to curb the power of big tech companies, has raised concerns for Apple regarding user privacy and data security. Apple announced the postponement after introducing "Apple Intelligence" at WWDC 2024, which includes advanced AI capabilities such as email summaries and custom emojis. The company is working with the European Commission to find a solution that allows these features to be launched without compromising user safety. Apple's delay in rolling out its new AI features in Europe, due to regulatory uncertainties related to the EU's Digital Markets Act, presents both positive and negative implications. On the positive side, it highlights Apple's commitment to user privacy and data security, showing their willingness to comply with stringent regulations. However, it also means that European users will miss out on these new features for now, potentially affecting Apple's competitiveness and innovation speed in the region. Overall, while the delay underscores Apple's regulatory compliance efforts, it also poses challenges in maintaining market leadership and user satisfaction.??

• Tech Mahindra launches Project Indus Large Language Model (LLM) Tech Mahindra has launched Project Indus, a large language model (LLM) designed to support multiple Indic languages and dialects, starting with Hindi. In collaboration with Dell Technologies and Intel, the project aims to simplify the deployment of advanced AI models through a 'GenAI in a box' framework. This initiative will leverage high-performance computing and AI technologies to enhance various industries, including healthcare, education, and finance. Project Indus marks a significant step in making AI more accessible and scalable.??

• OpenAI, Microsoft sued by Center for Investigative Reporting as news industry bolsters attack on AI The Center for Investigative Reporting (CIR), which publishes Mother Jones and Reveal, has filed a lawsuit against OpenAI and its primary backer, Microsoft. The lawsuit, filed in the U.S. District Court for the Southern District of New York, accuses the companies of copyright infringement for using CIR’s content to train their AI models without permission or compensation. CIR's CEO, Monika Bauerlein, emphasized that this "free rider behavior" is both unfair and a violation of copyright laws, underscoring the value of journalistic work and the need for proper licensing agreements. This lawsuit follows similar legal actions taken by other media organizations, including The New York Times and various other publishers, who argue that their content has been improperly used to develop AI technologies like ChatGPT. These cases highlight the ongoing tension between AI developers and content creators over the use of copyrighted material for training AI models.

?

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

Major record companies, led by the RIAA, are suing AI music generators Suno and Udio for copyright infringement, claiming unauthorized use of music to train AI models. TikTok has pulled a new AI tool that generated inappropriate content, raising concerns about its ability to moderate disinformation. OpenAI has developed a comprehensive content detection system to identify harmful content. Multiple AI companies are bypassing web standards to scrape publisher content, sparking disputes. Oxford researchers aim to fix AI hallucinations caused by ambiguous words. Apple's latest AI announcement benefits TSM and Microsoft but impacts Alphabet negatively. PwC's report highlights the transformative impact of AI in the workplace, improving efficiency and job satisfaction.

• Major record companies sue AI music generators Suno, Udio for 'mass infringement' of copyright? Major record companies, coordinated by the RIAA, are suing AI music generators Suno and Udio for massive copyright infringement. The lawsuits, filed in Massachusetts and New York, allege that these AI companies used copyrighted music to train their models without authorization. The RIAA argues that using copyrighted materials to train AI does not qualify as "fair use," as the AI-generated outputs directly compete with original works. This legal action aims to enforce ethical AI development and protect artists' rights.

• TikTok pulls new AI tool that spouted Hitler on command, horrified experts | CNN Business As it turns out, it is not the first time. A recent investigation has highlighted significant concerns about TikTok's ability to moderate disinformation, particularly ahead of the EU elections, as all 16 disinformation ads submitted for testing were approved. Despite efforts such as in-app "election centers" to provide trusted information, these measures have proven insufficient. This issue is part of a broader problem with AI-generated disinformation, including deepfakes and fake news, which exploits existing flaws in our information ecosystem and rapidly impacts public perception and behavior.

• A Holistic Approach to Undesired Content Detection in the Real World | OpenAI OpenAI's new content detection system offers a holistic methodology for identifying harmful content, enhancing moderation workflows. It uses comprehensive taxonomies, precise labeling, and an active learning pipeline to capture rare events and adapt to diverse needs. This robust approach ensures platforms can effectively filter undesired content like sexual, hateful, violent, self-harm, and harassment material, improving user safety and experience. Download paper here https://arxiv.org/pdf/2208.03274??

• Exclusive: Multiple AI companies bypassing web standard to scrape publisher sites, licensing firm says | Reuters? Multiple AI companies are bypassing the "robots.txt" web standard to scrape publisher content without permission, as revealed by content licensing startup TollBit. This practice, involving companies like Perplexity, has sparked disputes with publishers such as Forbes. The News Media Alliance expressed concerns that ignoring "do not crawl" signals undermines content monetization and journalist funding. TollBit offers a solution by facilitating licensing deals between publishers and AI companies, ensuring fair compensation for content use.?

• Researchers hope to quash AI hallucination bugs that stem from words with more than one meaning | Tom's Hardware Researchers at Oxford have developed a method to address AI hallucinations caused by words with multiple meanings. Using the concept of semantic entropy, they can detect when AI is likely hallucinating by identifying ambiguous word usage. This approach doesn't require additional human supervision, making it efficient for various tasks. This advancement aims to improve the reliability of AI-generated content, allowing users to better trust AI outputs while being aware of potential errors.??

• 2 Winners and 2 Losers From Apple's Latest Artificial Intelligence (AI) Announcement Winners include Taiwan Semiconductor Manufacturing (TSM), benefiting from increased demand for iPhone models incorporating Apple Intelligence, and Microsoft (MSFT), due to its partnership with OpenAI, which provides the AI technology for Apple. Losers include Alphabet (GOOG, GOOGL), as Apple's decision to use ChatGPT from OpenAI over Alphabet's Gemini model disrupts their traditional collaboration and potentially impacts revenue from search. Apple devices. What do you think??

• PwC: How Technology and Gen AI is Reshaping the Workplace PwC's 2024 Global Workforce Hopes and Fears report highlights the transformative impact of AI and generative AI on the workplace. Employees are increasingly adopting AI to manage workloads and advance their careers, with 82% of daily Gen AI users expecting increased efficiency and 76% anticipating higher salaries. Despite rising workloads and rapid changes, job satisfaction is improving. The report underscores the necessity for continuous skill development, urging employers to provide better upskilling opportunities and support to help workers leverage AI effectively.?

Business News and updates around partnerships, cost and Investments

Apple and Meta are reportedly discussing integrating Meta's generative AI into Apple's AI system, potentially reducing Apple's reliance on OpenAI. Apple is also negotiating with Anthropic and Perplexity for AI collaborations. This partnership aims to enhance AI features while maintaining user privacy. McKinsey is focusing on generative AI, with 40% of its work already AI-related, integrating AI across business processes. Investors are advised to consider ethical and regulatory challenges when investing in AI. Researchers are addressing AI's energy demands by developing efficient hardware and sustainable practices.

• Apple and Meta in Focus After Reportedly Discussing Generative AI Partnership Apple and Meta are reportedly discussing a partnership to integrate Meta’s generative AI model into Apple’s new AI system, Apple Intelligence. Apple is also in talks with startups Anthropic and Perplexity for potential AI collaborations. These partnerships would allow companies to sell premium subscriptions through Apple's distribution network. The deal could reduce Apple’s reliance on OpenAI, but it is still being negotiated. More details: Apple, Meta held discussions to add Llama Gen AI to iOS 18, likely to be added soon – Firstpost This integration is part of Apple's broader strategy to enhance its AI offerings, which include advanced features like real-time email summaries, custom emojis, and improved Siri functionalities. These features aim to maintain user privacy by processing data on-device rather than in the cloud, aligning with Apple's commitment to privacy and security.???

• McKinsey says it needs to reinvent itself and that AI is the answer: 'It's going to be most of what we do in the future' McKinsey & Company is focusing on generative AI as a central part of its future strategy, with AI and analytics already comprising about 40% of its work. The firm's AI division, QuantumBlack, employs around 2,000 data scientists and has completed roughly 400 generative AI projects in the last six months. This move is part of a broader effort to integrate AI into various business processes, emphasizing responsible AI, change management, and workflow integration. McKinsey is also partnering with AI startups like Cohere to enhance its capabilities and build client trust, expecting AI to become integral to most of its operations in the future.??

• How investors should approach 'gen AI' Investing in companies leading in AI infrastructure, applications, and AI-enabled industries can be beneficial. However, investors must also be mindful of the ethical and regulatory challenges posed by AI advancements.??

• https://www.goldmansachs.com/intelligence/pages/gs-research/gen-ai-too-much-spend-too-little-benefit/report.pdf? Tech giants and other companies are set to spend over $1tn on AI capex in coming years, with so far little to show for it. This spending is on data centers, chips, other AI infrastructure, and the power grid.?

• Taking a closer look at AI’s supposed energy apocalypse | Ars Technica Generative AI's growing computational demands are significantly impacting electricity consumption and posing challenges for power grids. As AI applications, particularly large language models, become more widespread, the associated data centers are expanding, resulting in increased power usage. Training a large AI model can consume as much electricity as a small town over a year, which raises concerns about grid stability. To address this, researchers and companies are developing energy-efficient hardware, such as specialized AI chips, and optimizing AI training and inference processes. Additionally, integrating AI operations with renewable energy sources and improving data center cooling technologies are key strategies to manage and mitigate the environmental footprint of AI. These combined efforts aim to ensure that the benefits of generative AI can be realized without overburdening power infrastructure.?
• Researchers run high-performing large language model on the energy needed to power a lightbulb Researchers at UC Santa Cruz have developed a novel method to run large language models (LLMs) efficiently by eliminating matrix multiplication. This innovation significantly reduces environmental impact and operational costs. The study demonstrates that a 2.7 billion parameter model performs comparably to traditional models and showcases a 1.3 billion parameter model running on a GPU accelerated by a custom-programmed FPGA chip, consuming only 13 watts. This advancement could make LLMs more accessible and sustainable.??

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

• The State of Generative AI in 2024 The state of generative AI in 2024 shows a significant increase in adoption across various sectors, with 72% of firms now using AI. The highest growth is seen in professional services, marketing, sales, and product development. Generative AI is employed primarily in marketing, sales, and IT functions. The Asia-Pacific and Greater China regions have the highest AI adoption rates. Larger industries are investing more in analytical AI, but overall AI investment is expected to grow significantly over the next three years.?

• How the CIA is using generative AI — now and into the future | FedScoop The CIA is leveraging generative AI to enhance its operations and intelligence analysis. This technology helps automate data processing, generate insights from vast amounts of information, and improve decision-making processes. By integrating generative AI, the agency aims to streamline workflows and augment human capabilities, ensuring more efficient and effective intelligence activities.???

Women Leading in AI?

New Podcast:? Tune in to Navigating the Digital Poly Crisis with ChaVon (CJ) Clarke-Joell. CJ shares insights from her new book, 'The Digital Polycrisis,' which provides strategies for navigating the complexities of the digital world. She emphasizes the importance of planting seeds of interest in AI from a young age and highlights her experience with Mia’s Global AI Leadership Program. She envisions a future where humans and AI work together, and encourages continuous learning and adaptation to stay relevant in the age of AI.

Featured AI Leader: ??Women And AI’s Featured Leader - Fleur Prince ??

Congratulations to Fleur Prince for being named a Featured AI Leader. Learn how Fleur is using AI and what her recommended AI resources are.?

Learning Center and How To’s

Interested in integrating AI into your workflow? Check out our how-to guides and let us know how it goes!

• Recording from LinkedIn Live on supply chain: https://youtu.be/evQ_9E2k3u0?si=dpMNy3sA_kEwvQyP with Laurie Paciocco-Wood and Mariana Saddakni

• Have questions or feedback? Contact us or join our next LinkedIn Live session! LinkedIn Live “How to AI for CSPs” with Jillian Kaplan . Register here: https://www.dhirubhai.net/events/7212492285660250112/comments/?

• Introduction to Generative AI | Coursera by Google??

• An introduction to embedding an LLM into your application ? The Register –? a detailed guide for integrating large language models (LLMs) into applications using Mistral.rs, a Rust-based LLM engine. It covers hardware and software requirements, including dependency installations, and offers step-by-step instructions for setting up and running Mistral.rs. The guide includes examples of using Mistral.rs with CPU and CUDA, testing with quantized models, and integrating the engine via HTTP server or directly using Rust or Python APIs. The full tutorial is aimed at intermediate developers.

• New AI algorithm flags deepfakes with 98% accuracy — better than any other tool out there right now The new AI algorithm, MISLnet, developed by Drexel University, detects deepfakes by identifying sub-pixel markers unique to AI-generated content. This tool can be integrated into existing verification systems or used as standalone software for organizations needing to authenticate video content. For practical application, you would typically install or integrate the algorithm with your media analysis tools, then use it to scan and analyze videos for authenticity, flagging potential deepfakes with high accuracy.??

• Building a personalized code assistant with open-source LLMs using RAG Fine-tuning By integrating real-time code snippets from repositories during training and inference phases, the fine-tuned Mistral 7B Instruct v0.2 model demonstrates substantial enhancements in accuracy, outperforming non-fine-tuned models and competing LLMs like Claude 3 Opus and GPT-4o across multiple AI codebases. This approach not only mitigates issues like hallucination and outdated knowledge but also significantly improves speed and reduces costs, making it a promising advancement for developing practical and efficient AI code assistants.

• How To Solve LLM Hallucinations - by Dr. Ian Cutress This innovative approach utilizes a Mixture of Memory Experts (MoME) framework, enabling near-perfect recall of specific information without compromising the model's general reasoning ability. Currently targeted at Fortune 500 companies, Lamini's technology represents a significant advancement in fine-tuning LLMs for enterprise-level applications, suggesting potential future impacts on AI development and hardware requirements in the industry.?

• GenAI with Python: RAG with LLM (Complete Tutorial) | by Mauro Di Pietro | Jun, 2024 | Towards Data Science? The tutorial on Towards Data Science explains how to use Retrieval-Augmented Generation (RAG) with Python and large language models (LLMs). RAG combines LLMs with a retriever to enhance the model's ability to access and utilize external knowledge, improving accuracy and relevance in responses. The tutorial covers setting up the environment, using the Hugging Face Transformers library, and integrating retrievers with LLMs for various applications.

Prompt(s) of the week

• GitHub - teknium1/Prompt-Engineering-Toolkit The Prompt Engineering Toolkit on GitHub by Teknium1 is a comprehensive resource for prompt engineers. It provides a collection of tools, techniques, and examples for crafting effective prompts for AI models. The toolkit includes various prompt formats, fine-tuning methods, and best practices to enhance AI performance and output quality. It's designed to help users create precise and efficient prompts, making AI interactions more productive and reliable.??

• Claude Artifacts is the greatest innovation in AI this year — 5 prompts to try it now | Tom's Guide highlights five prompts to try with Claude Artifacts. These prompts include asking Claude to write a short story based on a given theme, generating creative ideas for a project, providing detailed answers to complex questions, crafting personalized messages or emails, and offering productivity tips tailored to individual workflows. These examples showcase Claude's versatility in creative and practical applications, making it a valuable tool for various tasks. Have you tried them yet??

Tools and Resources

• The Open-Source Libraries to Check Out for LLM Building | HackerNoon? outlines several essential open-source libraries crucial for developing Large Language Models (LLMs). These libraries serve various purposes, from optimizing model training to enhancing natural language processing capabilities. One notable library mentioned is Hugging Face's Transformers, which provides pre-trained models and tools for fine-tuning and deploying LLMs efficiently. Another key resource highlighted is PyTorch, favored for its flexibility and powerful capabilities in deep learning tasks, including LLM development. The article also mentions TensorFlow, renowned for its scalability and extensive community support, making it a robust choice for constructing and training LLMs. Additionally, libraries such as Fairseq and OpenAI's GPT-3 API are recommended for their contributions to advanced NLP tasks and access to cutting-edge models like GPT-3. These libraries collectively form a comprehensive toolkit for developers aiming to leverage open-source solutions in building and deploying effective Large Language Models.

• Uncensor any LLM with abliteration The Hugging Face blog post introduces "abliteration," a technique to uncensor large language models (LLMs) without retraining. This method identifies and removes the model's refusal mechanism, enhancing flexibility and responsiveness. By analyzing harmful and harmless prompts, researchers can alter the model’s weights to prevent it from refusing any request. It provides a detailed guide on how to implement it, showcasing how users can apply this method to uncensor large language models without the need for retraining. The article includes instructions and examples, particularly using the Daredevil-8B model, to help users understand and utilize this technique effectively.

• Character.ai is a platform that enables users to create and interact with artificial intelligence-driven characters. These characters can engage in realistic and dynamic conversations, simulating human-like responses. The platform utilizes advanced AI technologies to bring these characters to life, allowing for a wide range of applications, including entertainment, education, and personal assistance. Users can customize characters, define their personalities, and share them with others, creating a unique and engaging interactive experience. Introducing Character Calls? – a new feature that allows users to engage in real-time voice conversations with AI characters. This innovative offering aims to enhance the interactive experience by providing a more immersive and personal way to communicate with AI. The feature is designed to simulate natural phone calls, enabling dynamic and context-aware interactions with the AI characters. This development marks a significant step in making AI interactions more lifelike and accessible to users.

• AI’s Hidden Opportunities: Shawn "swyx" Wang on New Use Cases and Careers | Heavybit discusses several AI tools that are beneficial for software developers. These include GitHub Copilot, which assists with code completion by suggesting snippets and automating repetitive tasks, and DeepCode, an AI-powered code review tool that identifies potential issues and suggests improvements. It also mentions TabNine and Codota for AI code completion across multiple programming languages and MLflow for managing machine learning workflows. These tools help boost productivity, improve code quality, and streamline development processes.?

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.