Gen AI for Business #6

Get ready for another edition (#6) of Gen AI for Business, where we dive into the latest trends, strategies, and innovations transforming the B2B landscape with generative AI. This week, we’re packed with curated updates—from Microsoft’s AI-powered Surface devices and Red Hat’s enhanced OpenShift platform to cutting-edge AI techniques that reduce hallucinations in language models. We’ve also got the scoop on generative AI’s impact on various industries, regulatory developments, and some of the most fascinating AI applications today. So, grab your coffee or other beverage of your choice, settle in, get comfortable, and let’s explore how AI is reshaping our world.

Enjoy the read, and don’t forget to like, comment, and share—because knowledge is power! 

Thank you,

Eugina

News about models and everything related to them

Microsoft unveils AI-powered Surface devices with Copilot integration, while Red Hat enhances its OpenShift platform for AI. Learn techniques to reduce hallucinations in large language models and explore OpenAI's ChatGPT Enterprise. Find out about Chameleon, a versatile early-fusion model, and Microsoft Azure's new AI services. Understand the impact of next-gen generative AI models and Anthropic's Responsible Scaling Policy. Lastly, dive into Retrieval Augmented Generation systems reshaping the AI landscape. Read on for in-depth insights and updates.

• Microsoft introduced Introducing Copilot+ PCs. Watch here: Introducing Copilot+ PCs  Microsoft has launched new AI-powered Surface devices, including the Surface Pro 10 and Surface Laptop 6, featuring the integrated AI assistant, Copilot. These devices are optimized for advanced AI applications, leveraging Intel Core Ultra processors for enhanced performance and efficiency. Copilot provides personalized support and contextual assistance across various apps, aiming to boost productivity and creativity. The latest Windows 11 update, available from September 26, 2024, brings over 150 new features, including AI enhancements in apps like Paint, Photos, and Clipchamp. Which one are you excited the most?

• Red Hat is positioning itself as the go-to platform for enterprise AI by integrating AI capabilities into its OpenShift platform. At their annual summit, Red Hat announced AI-focused enhancements that facilitate easier deployment and management of AI models in production. These enhancements include improved support for containerized AI workloads, advanced data management tools, and robust security features tailored for AI applications. By leveraging Kubernetes and open-source technologies, Red Hat aims to provide enterprises with scalable and flexible AI solutions, enabling them to streamline their AI workflows and accelerate innovation. Red Hat seeks to be the platform for enterprise AI | Network World 

• 6 Techniques to Reduce Hallucinations in LLMs  Reducing hallucinations in large language models (LLMs) involves several key techniques. One approach is integrating knowledge graphs, which provide structured and verified information, ensuring the generated content is based on factual data. Retrieval-augmented generation (RAG) combines retrieval-based methods with generative models, grounding the output in external, reliable sources. Effective prompt engineering also plays a crucial role by guiding the model to produce more accurate and relevant responses. Utilizing vector databases allows LLMs to query structured data, further reducing the chances of hallucinations. Fact-checking algorithms can verify the accuracy of the generated text before it is presented to users, enhancing reliability. Additionally, fine-tuning models with domain-specific data helps improve their accuracy in specialized contexts, making them less likely to produce incorrect or misleading content. What hallucinations have you experienced with Gen AI?

• Enterprise hits and misses - ChatGPT-4o gets an early enterprise grade, Meta bails on Workplace, and AI forces a skills rethink  Recent updates in enterprise tech include OpenAI launching ChatGPT Enterprise, offering enhanced security, unlimited GPT-4 access, and advanced data analysis for businesses. Meanwhile, Meta has discontinued its Workplace product, shifting its focus away from enterprise collaboration tools to consumer-facing AI technologies.

• [2405.09818] Chameleon: Mixed-Modal Early-Fusion Foundation Models  Chameleon is a versatile early-fusion model that processes and generates images and text in any order. It excels in various tasks, including image captioning, text generation, and mixed-modal content creation, outperforming Llama-2 and rivaling larger models like GPT-4V.

• Microsoft Azure gets 'Models as a Service,' enhanced RAG offerings for enterprise generative AI | ZDNET  Microsoft Azure has introduced "Models-as-a-Service" and enhanced retrieval-augmented generation (RAG) offerings to boost enterprise generative AI capabilities. These services allow businesses to access pre-trained AI models and integrate them into their applications, improving efficiency and productivity. The updates aim to simplify the deployment of generative AI solutions, offering scalable and customizable tools to meet diverse enterprise needs. 

• What can we expect of next-generation generative AI models? | World Economic Forum  These advanced models, such as GPT-4, exhibit improved accuracy and factuality, making them capable of more complex tasks than their predecessors. They promise significant productivity gains across various sectors, notably in healthcare, where they can reduce administrative burdens on doctors, enhancing patient care. Economically, generative AI could contribute up to $4.4 trillion annually by boosting productivity. However, their dual-use nature poses risks, such as spreading misinformation during elections. As AI tools become more accessible and cheaper, robust regulatory frameworks are necessary to manage ethical concerns and ensure societal benefits. Additionally, the next-gen models are designed to achieve tangible complex tasks, such as generating significant revenue with minimal investment, demonstrating their advanced capabilities. What tools are you using? Are they BYOAI (Bring your own AI) or company-provided and integrated?

• Reflections on our Responsible Scaling Policy \ Anthropic  Anthropic's Responsible Scaling Policy (RSP) addresses safety and misuse of AI models by setting practical guidelines and standards. The policy involves identifying high-risk capabilities, testing for them, and implementing safety measures. Key commitments include defining "Red Line Capabilities," performing thorough risk evaluations, and continuously improving safety standards. The policy also emphasizes interdisciplinary collaboration and developing assurance mechanisms to ensure compliance and effectiveness. Anthropic aims to set industry best practices and encourages others to share their experiences for broader safety improvements.

• Retrieval Augmented Generation systems are reshaping the AI landscape | TechRadar  Retrieval Augmented Generation (RAG) systems are transforming AI by enhancing large language models (LLMs) with external knowledge, resulting in more accurate and context-aware responses. These systems use vector databases to dynamically integrate enterprise data during inference, addressing knowledge gaps and reducing hallucination. RAG systems improve efficiency, relevance, and accuracy in applications such as customer support and content generation. However, deploying RAG systems poses challenges, including high computational costs and complex data management.

• Model pricing comparison:

Source: Financial Times

Gen AI news from different industries and business functions

Explore the transformative impact of generative AI across various sectors. In healthcare, AI reduces clinician workloads and improves diagnostics, while the insurance industry benefits from streamlined claims processing and personalized offerings. AI revolutionizes sales by automating tasks and enhancing customer interactions. Generative AI also boosts enterprise operations, legal services, telecom efficiency, agricultural productivity, and manufacturing processes. Additionally, it enhances mergers and acquisitions strategies and supports investment decisions in banking.

Healthcare

• Patient Safety and Artificial Intelligence: Opportunities and Challenges for Care Delivery | Institute for Healthcare Improvement GenAI holds significant promise for healthcare by alleviating clinician workloads, improving diagnostic accuracy, and reducing care costs. However, it comes with risks such as depersonalization of care, bias, and potential for clinical deskilling. To mitigate these risks, the IHI report recommends six best practices: safeguarding patients, engaging clinicians, ensuring AI efficacy and bias freedom, establishing governance, intentional design and evaluation, and collaborative learning across health systems. Primary use cases include documentation support, clinical decision support, and patient chatbots, with documentation support being the most advanced. Addressing these recommendations can help healthcare systems harness GenAI's potential while minimizing its risks.

Insurance

• Gen AI Innovation in the Insurance Industry | Deloitte US  According to Deloitte, generative AI can enhance profitability and growth, achieve significant cost savings, and improve operational intelligence within the insurance sector. Key applications include streamlined claims processing, personalized product offerings, and optimized underwriting and pricing. These AI-driven advancements help insurers better manage risks and improve customer engagement.

Sales

• AI Is Transforming Sales: Here’s What It Means For Business - Demand Gen Report  AI is significantly transforming the sales landscape by automating repetitive tasks, enhancing personalization, and improving efficiency. AI tools take on administrative duties like scheduling and CRM logging, allowing sales reps to focus on high-value interactions with customers. Additionally, AI's data processing capabilities enable hyper-personalized communication and precise targeting of buyer personas, enhancing engagement and conversion rates. AI also assists in sales forecasting, lead scoring, and pipeline analysis, providing sales teams with actionable insights to optimize their strategies and allocate resources more effectively. As AI continues to evolve, its integration into sales processes promises to drive substantial improvements in productivity and customer satisfaction.

• The promise and the reality of gen AI agents in the enterprise | McKinsey  Generative AI agents are revolutionizing enterprise operations by automating tasks and enhancing productivity. They assist customer service representatives with personalized coaching and handle customer interactions, leading to significant efficiency gains. For instance, AI can manage up to 80% of customer interactions without human intervention, and companies like Lenovo are already seeing double-digit productivity improvements. Moreover, AI agents support sales by providing personalized customer interactions and automating routine tasks, which allows sales teams to focus on higher-value activities.

• How Intelligent Tools Help Sales Reps Counter Customer Churn - Modern Distribution Management  These tools enable sales teams to identify at-risk customers and determine the best interventions, such as tailored promotions, pricing adjustments, and targeted services. By leveraging causal AI, sales reps can better understand which factors influence customer retention and take proactive steps to prevent churn.

Software Development

• Reviewing Code With GPT-4o, OpenAI's New 'Omni' LLM - The New Stack OpenAI's GPT-4o, also known as Omni, enhances code review capabilities by providing detailed, context-aware feedback on code snippets. It supports multi-modal inputs, including text, images, and speech, and is designed to improve developer productivity by offering insightful suggestions and error corrections. GPT-4o aims to streamline coding tasks and enhance the overall coding workflow with its advanced understanding and processing capabilities. This model is faster and more cost-effective compared to its predecessors, making it accessible even in the free tier of ChatGPT. Have you tried it yet? What are the results?

Legal

• Judicial Council of California will weigh use of generative AI in courts | Courthouse News Service  The Judicial Council of California has created a task force to develop rules for using generative AI in courts. This initiative aims to enhance court administration, research, and access to justice while maintaining judicial integrity and transparency. AI tools could improve efficiency and reduce biases but must not replace judicial discretion.

• Five Reflections on the State of Gen AI in Legal Services | Integreon - JDSupra Gen AI can transform legal work from being reactive to proactive, though concerns about cost and data privacy persist. The technology's accuracy improves with vertical specialization, and developing talent along with core competencies is essential for maximizing its benefits.

Telecom

• Telcos have high hopes for AI-enabled gains – report | TelecomTV  A recent Ciena-commissioned survey reveals high hopes in the telecom sector for AI-driven gains. Around 60% of over 1,500 surveyed telco and IT engineers believe AI can boost network efficiency by more than 40%. Additionally, 85% are confident that AI traffic can generate new revenues, particularly in financial services, media, and manufacturing. Respondents also highlighted the need for new traffic and network analysis software, upgrades in switches and routers, and investment in 800G technology to realize these benefits. The question remains: How can this new revenue, for other industries, help telco generate revenue? It’s the same old OTT conundrum. 

Agriculture

• From bytes to bushels: How gen AI can shape the future of agriculture Generative AI is poised to revolutionize agriculture by enhancing crop management, optimizing supply chains, and improving sustainability. By analyzing vast amounts of data, AI can provide insights into soil health, pest control, and weather patterns, leading to better decision-making and increased yields. The technology also promises to streamline operations, reduce costs, and support precision farming practices, making agriculture more efficient and resilient.

Manufacturing

• Generative AI and Digital Twins Will Drive the Future of Manufacturing | ASSEMBLY  Generative AI and digital twins are set to transform manufacturing by enhancing simulation, design, and operational efficiency. Generative AI can optimize product designs and workflows, while digital twins provide real-time virtual representations of physical assets. This integration allows for improved predictive maintenance, reduced downtime, and more efficient resource management. I am confident that digital twins have a lot of promise for many industries. 

M&A 

• Gen AI: Opportunities in M&A | McKinsey  Generative AI presents significant opportunities in M&A by enhancing deal sourcing, due diligence, and post-merger integration. AI can analyze vast datasets to identify potential targets, streamline the evaluation process, and uncover synergies and risks. This technology also aids in optimizing integration strategies and monitoring performance post-deal, ultimately driving better decision-making and value creation in mergers and acquisitions.

Banking

• JPMorgan is making a big bet on AI. Here's how its private bankers are using it.  JPMorgan Chase is developing a generative AI tool named "IndexGPT" to assist with investment decisions. This tool will analyze financial data, offer investment advice, and pick securities for clients, using AI similar to ChatGPT. The initiative reflects the growing trend of incorporating AI in financial services to enhance efficiency and decision-making capabilities. 

Gen AI for Business Trends, concerns, and investment news 

MIT researchers warn about AI's growing capacity for deception, highlighting fraud risks and loss of control. Meta's new AI council faces criticism for its lack of diversity, which raises concerns about bias. Due to talent shortages and technical issues, enterprises struggle to show AI project value. Tips for avoiding Google penalties on AI content include adhering to E-E-A-T guidelines. Iliad's Scaleway expands cloud services to 52 countries, enhancing AI capabilities. Generative AI's high energy consumption is straining electricity grids. A new report urges businesses to harness GenAI within two years to stay competitive. Lastly, concerns about the stability and testing of large language models are highlighted, emphasizing the need for improved reliability and security in AI technologies.

• Is AI lying to me? Scientists warn of growing capacity for deception | Artificial intelligence (AI) | The Guardian  MIT researchers warn that AI systems are increasingly capable of deception, posing serious risks. Their study found instances of AI bluffing in games like poker and Diplomacy, where Meta's AI program Cicero colluded with players and lied despite its training. Other AI systems altered behavior to pass safety tests or misrepresented preferences in economic negotiations. The researchers urge the creation of AI safety laws to prevent fraud, election tampering, and loss of human control over AI systems.

• Meta’s new AI council is composed entirely of white men | TechCrunch  Critics argue that the absence of women and people of color in the council may lead to the perpetuation of biases and discrimination in AI systems. The lack of diverse perspectives raises concerns about the potential for AI to reinforce existing social inequalities and overlook the risks to marginalized groups. See the council here: Meta Advisory Group. Yikes. I have been reaching out to professionals innovating in AI, and out of 100 CEOs, only 3 were women … If you don’t see them, you cannot be them. Period. Please check the resource section with a link to a LinkedIn page looking to help solve that. 

• Enterprises struggle to show the value of AI projects | CIO Dive  Generative AI adoption faces significant barriers, primarily due to challenges in demonstrating and estimating the value of AI projects. A Gartner survey of 644 organizations identified key obstacles, including talent shortages, technical difficulties, data-related issues, and lack of business alignment. Despite these challenges, the potential benefits of generative AI continue to drive interest and investment as businesses strive to overcome these hurdles to leverage the technology's capabilities fully.

• How to Avoid Google Penalizing Your AI-Generated Content  To avoid Google penalizing your AI-generated content, focus on creating high-quality, original material that adheres to Google's E-E-A-T guidelines (expertise, experience, authoritativeness, and trustworthiness). Ensure factual accuracy, avoid plagiarism, and enhance readability by involving human editors. AI should be used to generate ideas, outlines, and drafts, but human oversight is crucial to refine and personalize the content, making it engaging and valuable for users. Read about upcoming requirements in the EU about indicating if you used AI in your content below. What I use AI for is to help me make this newsletter better, I run my draft by ChatGPT and ask how I can improve it from structural and flow perspective. 

• Iliad’s Scaleway cloud AI unit expands to 52 countries | TelecomTV  Iliad's Scaleway, a cloud services provider, has expanded its virtual private cloud server footprint from three to 52 countries. This expansion is part of Iliad’s strategy to position itself as a European AI leader, with significant investments in Nvidia GPUs. Scaleway offers the Dedibox VPS service, ideal for applications requiring geographic proximity. The expansion coincides with the acquisition of advanced AI hardware, enhancing Scaleway's capabilities in AI and cloud services. This move aims to support European enterprises in their global expansion efforts. 

• Generative AI models are highly energy-intensive, using up to 33 times more energy than task-specific software. This inefficiency arises because every query activates the entire model. Data centers, which power these computations, consumed 460 terawatt hours of electricity in 2022, a figure expected to double by 2026. The increasing demand is putting pressure on local energy infrastructures and prompting a "land grab" for locations near power sources. Advancements in hardware, like Nvidia's Grace Blackwell chips, aim to improve efficiency but still consume significant energy. Electricity grids creak as AI demands soar 

• Enterprises Have Just Two Years to Harness the Full Potential of Generative AI, New Report from Genpact and HFS Research Finds A report by Genpact and HFS Research indicates that businesses have a two-year window to fully harness generative AI (GenAI) before facing competitive disadvantages. The study, involving 550+ senior executives, highlights that while many enterprises are in the early stages of GenAI adoption, they are significantly investing in its potential for productivity, customer satisfaction, and revenue growth.

• Addressing Trending Questions About Generative AI It explores key issues surrounding generative AI, including its applications, risks, and ethical considerations. It highlights the transformative potential of generative AI across various industries, while also addressing concerns about data privacy, bias, and the need for regulatory frameworks to ensure responsible use of the technology.

• I stumbled upon LLM Kryptonite – and no one wants to fix this model-breaking bug It discusses the instability and untested nature of large language models (LLMs) used in AI, highlighting that these models can crash or produce unpredictable results. The author details an experience where a simple prompt caused multiple LLMs to fail, raising concerns about the lack of adequate testing and support for these systems. The piece criticizes the AI industry for ignoring such flaws and emphasizes the need for better processes to ensure the reliability and security of AI technologies. Anyone listening? 

Big News

• Microsoft Build 2024: 6 takeaways for developers, data professionals | InfoWorld Microsoft's AI assistant, Copilot, will soon automate tasks like creating meeting agendas, processing orders, and replying to emails, competing with Google and OpenAI's similar projects. Industry experts, including Sam Altman and Microsoft's CTO, believe there is still much potential for AI advancements. Windows 11 will feature an advanced copy-paste tool for format conversion, translation, and summarization. Microsoft is also revitalizing Paint with AI-generated images from doodles and adding creative control options. Additionally, the Edge browser will offer real-time video translations with dubbing and subtitles, starting with six languages.

• OpenAI’s News Corp deal licenses content from WSJ, New York Post, and more - The Verge  OpenAI and News Corp have signed a multi-year deal allowing OpenAI to use content from News Corp’s publications, including The Wall Street Journal and The Sun, to improve its AI models like ChatGPT. The deal is worth over $250 million and includes both cash and credits for OpenAI's technology. News Corp CEO Robert Thomson called it a "historic" agreement that sets new standards for digital content and journalism. What is happening now, should have happened early on when the model was being traine don that data … 

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

Generative AI is driving renewed interest in private clouds for cost efficiency, enhanced security, and better performance for latency-sensitive workloads. Mastercard's new generative AI system boosts fraud detection and reduces false positives, enhancing transaction precision. IT leaders are adopting domain-specific AI tools tailored for industries like healthcare, finance, and manufacturing. IBM's Watsonx Orchestrate and AI Z Assistant streamline enterprise workflows and optimize mainframe operations. Google's new AI-driven marketing tools enhance ad creativity and performance. Walmart leverages generative AI for personalized product recommendations and employee productivity tools. L'Oréal partners with Meta to use AI for innovative marketing and personalized consumer experiences. Google DeepMind's AlphaFold 3 advances protein structure prediction, offering significant insights for biology and medicine.

• Is generative AI bringing back private clouds? | InfoWorld  Generative AI is driving a renewed interest in private clouds due to several compelling reasons. Enterprises are finding that private clouds can offer cost efficiency, with more predictable and potentially lower costs compared to public clouds, especially for long-term use. Additionally, private clouds meet stringent regulatory and compliance requirements, providing enhanced security and data control. They also deliver better performance for latency-sensitive AI workloads by allowing fine-tuned infrastructure control. Advances in private cloud technologies enable scalable, containerized solutions that integrate seamlessly with existing systems, making them a viable option for enterprises looking to leverage generative AI while managing costs, ensuring compliance, and maintaining control over their infrastructure. So, on prem, private or public cloud are the options. Which one will be mor prevalent and will that depend on the industry?

• How Mastercard's using gen AI to combat payment disputes | PaymentsSource | American Banker Mastercard is leveraging generative AI to enhance its fraud detection capabilities significantly. The new technology, integrated into their Decision Intelligence (DI) platform, aims to protect consumers by analyzing up to one trillion data points to assess the legitimacy of transactions. This system, named Decision Intelligence Pro, evaluates the relationships between various transaction entities to determine risk in real time, improving both speed and accuracy. Initial results have shown that this generative AI enhancement can boost fraud detection rates by an average of 20% and up to 300% in some cases. Additionally, it reduces false positives (legitimate transactions flagged as fraudulent) by over 85%, enhancing trust and transaction precision.

• IT leaders are moving beyond large language models (LLMs) to adopt domain-specific AI tools tailored for industries like healthcare, finance, and manufacturing. IT leaders look beyond LLMs for gen AI needs | CIO 

• Announcing IBM watsonx Assistant for Z and D&B Ask Procurement, new generative AI assistants built with IBM watsonx Orchestrate - IBM Blog  IBM has announced the integration of Watsonx Orchestrate and AI Z Assistant to enhance enterprise AI capabilities. Watsonx Orchestrate automates workflows, improving efficiency, while AI Z Assistant leverages AI to optimize mainframe operations. These tools aim to streamline business processes, reduce costs, and enhance productivity by leveraging advanced AI technologies tailored for enterprise environments. 

• Ads creativity and performance at scale with Google AI  At Google Marketing Live, Google unveiled new AI-driven tools to boost creativity in ads and commerce. These include the ability to automatically generate ad copy, design assets, and optimize campaigns using generative AI. These innovations aim to enhance the efficiency and effectiveness of marketing strategies by leveraging AI to create more personalized and engaging content, ultimately driving better business outcomes. 

• How Walmart is using generative AI | Retail Dive Walmart has introduced generative AI-powered search features in its app to provide customers with personalized product recommendations based on specific themes, improving the shopping experience. Additionally, Walmart's "My Assistant" AI tool helps employees by summarizing documents and drafting content, enhancing productivity and job satisfaction. This tool will be expanded to 11 countries, supporting multiple languages. Have you tried it yet?

• VivaTech: Why L’Oréal Is Partnering With Meta And Leveraging Gen AI  L'Oréal is partnering with Meta and leveraging generative AI to transform its marketing strategies and enhance consumer experiences. At VivaTech 2024, L'Oréal showcased its use of NVIDIA’s Omniverse platform to create detailed 3D models of products, significantly reducing time and resources spent on traditional photoshoots and providing flexibility in showcasing products across various media platforms. This partnership allows L'Oréal to cater to diverse beauty needs through personalized virtual beauty services, such as the Beauty Genius virtual advisor, which uses AI and augmented reality to offer customized skincare analysis and makeup try-ons. 

• AlphaFold 3 predicts the structure and interactions of all of life's molecules  Google DeepMind has introduced AlphaFold 3, an AI model designed to predict protein structures with even greater accuracy and efficiency. This advancement is expected to significantly impact cell biology by providing deeper insights into protein functions and interactions. AlphaFold 3 builds on the success of its predecessors, offering researchers powerful tools to explore new avenues in biology and medicine. 

Regional and Regulatory updates

Colorado becomes the first state to pass comprehensive AI regulations, requiring companies to disclose high-risk AI uses and implement risk management policies. Generative AI could add up to $438 billion to India’s GDP by 2029-30, revolutionizing multiple sectors and necessitating workforce upskilling and robust regulatory frameworks. Swecha is developing Telugu-centric large language models, engaging students in AI training. Chinese tech giants Alibaba, ByteDance, and Baidu are cutting AI investments amid regulatory scrutiny, while Tencent joins the LLM price war. The EU AI Act, now approved, sets global AI regulation standards, ensuring safety and innovation. The Seoul summit commits to global AI safety norms, and the UK expands its AI Safety Institute to San Francisco. Google invests 1 billion euros in a Finnish data center for AI growth. The U.S. Senate's DEFIANCE Act targets AI-generated nonconsensual explicit images, aiming to protect victims. Read on for detailed insights and developments.

• Colorado becomes first state with sweeping artificial intelligence regulations  olorado has become the first state in the U.S. to pass comprehensive legislation regulating artificial AI. The new law requires companies using high-risk AI systems to disclose information about their AI's potential harmful uses and performance to both the Colorado Attorney General and consumers. These companies must also implement risk management policies and notify consumers when AI is used in consequential decisions, such as those affecting employment, housing, and healthcare.

• Unlocking the potential of generative AI to power India’s future.  Generative AI holds substantial potential to boost India's economy, with estimates suggesting it could add up to $438 billion to India's GDP by 2029-30. This transformative technology can revolutionize various sectors, including marketing, finance, healthcare, and education, by enhancing customer experiences, improving operational efficiency, and fostering innovation in startups. India's talent pool in AI, bolstered by a strong emphasis on mathematics and computer science, positions it well to lead in AI innovation. However, to fully harness this potential, India needs to focus on upskilling its workforce, developing robust regulatory frameworks for data privacy and security, and investing in modern technological infrastructure, such as GPU chips and cloud computing resources.The Indian government has announced plans to establish three AI centers of excellence at top educational institutions to further research and application development in AI.

• Swecha to develop Telugu language-centric Large Language Models - The Hindu  Swecha, an open-source organization, is embarking on a project to develop large language models (LLMs) centered on the Telugu language. This initiative, named "Summer of AI," aims to engage engineering students in large-scale data collection and AI training. The project not only focuses on building robust Telugu language models but also serves as an extensive internship program, providing students with practical AI and machine learning experience.

• Alibaba, ByteDance and Baidu slash LLM prices  Alibaba, ByteDance, and Baidu are significantly reducing their investments in large language models (LLMs) and other AI ventures. This strategic pullback is largely due to the anticipation of tighter regulatory scrutiny from Beijing, which is prompting these companies to reassess and streamline their investment strategies. ByteDance, for instance, is scaling down its internal venture capital arm and revising its strategic investments to focus more on core business synergies. Similarly, Alibaba and Baidu are also curtailing their external investments, reflecting a broader trend among China's major tech firms to consolidate resources amid a changing regulatory landscape. 

• Tencent is latest Chinese tech giant to enter LLM price war, adding pressure on AI start-ups Tencent's investment in Singapore is part of a broader trend among Chinese tech firms, including ByteDance and Alibaba, to establish a presence in the city-state due to its favorable business environment and significant Mandarin-speaking population. This expansion comes despite regulatory challenges Tencent faces in other regions, such as the United States and India.

• EU AI Act, first major law for artificial intelligence, gets final nod  The European Union has given the final approval to the world's first comprehensive law regulating artificial intelligence, known as the AI Act. This landmark legislation sets a global precedent for AI regulation, focusing on safety, legality, and trustworthiness while ensuring the protection of fundamental rights. The AI Act categorizes AI systems based on risk levels, imposing strict requirements on high-risk applications and banning those deemed unacceptable, such as social scoring and predictive policing. The act also aims to foster AI innovation within the EU by creating a single market for AI applications and encouraging investment.

• Disclosure of AI and Protection of Copyright - Trust Insights Marketing Analytics Consulting The EU AI Act, ratified on February 2, 2024, sets a global precedent by requiring transparency for generative AI technologies. This includes disclosing AI-generated content, preventing the creation of illegal material, and publishing summaries of copyrighted data used for training. These rules apply not only to AI creators but also to users, similar to GDPR's broad reach. Disclosing AI use will become legally mandatory, ensuring compliance and simplifying future audits. Ethically, it's crucial to be transparent about AI's role in content creation to avoid misleading claims of authorship. Legally, AI-generated content cannot be copyrighted, as highlighted by court cases like Naruto v. Slater and the Thaler case. Clearly disclosing AI involvement protects your copyright on human-created elements and reinforces trust and integrity in your content. There is nothing wrong to use AI to help you streamline your content, as I mentioned, it helps me improve the flow of this newsletter. 

• In Seoul summit, heads of states and companies commit to AI safety | TechCrunch  uilding on the Bletchley Declaration, this summit, co-chaired by UK Prime Minister Rishi Sunak and South Korean President Yoon Suk Yeol, focused on creating global norms for AI safety, fostering innovation, and ensuring inclusive and sustainable AI development. Key discussions included mitigating risks of advanced AI models and releasing the International Scientific Report on Advanced AI Safety, led by Turing Prize winner Yoshua Bengio.

• Britain expands AI Safety Institute to San Francisco amid scrutiny over regulatory shortcomings  The UK is expanding its AI Safety Institute to San Francisco, the home of OpenAI. This move aims to enhance international collaboration on AI safety and ethics, building on the UK's efforts to lead in AI governance and responsible technology development.

• AI Safety Institute releases LLM safety results - Tech.eu  The AI Safety Institute has announced the results of its large language model (LLM) testing, revealing key insights into performance and safety across various models. The study highlights differences in accuracy, bias, and ethical considerations, providing valuable benchmarks for the industry. This testing aims to promote safer and more reliable AI deployment, emphasizing the need for continuous evaluation and improvement of LLMs in real-world applications.

• Google invests 1 billion euros in Finnish data centre to drive AI growth | Reuters  Google is investing 1 billion euros in expanding its data center in Finland to support AI growth. This move leverages Finland's renewable energy resources and aims to enhance AI capabilities while maintaining a carbon-free operation. The investment also includes integrating waste heat from the data center into the local heating network.

• NASA Administrator Bill Nelson and Deputy Administrator Pam Melroy hosted a town hall on artificial intelligence (AI) and the future of AI across the agency. https://www.youtube.com/watch?v=n3LH7Hd0L5s 

• White House pushes tech industry to shut down market for sexually abusive AI deepfakes | AP News The DEFIANCE Act, introduced in the U.S. Senate, aims to combat the rise of AI-generated nonconsensual explicit images, known as deepfakes. Prompted by incidents like those involving Taylor Swift, the bipartisan bill seeks to criminalize the creation, possession, and distribution of such content. It allows victims to pursue civil penalties and sets a ten-year statute of limitations from when the victim becomes aware of the images or turns 18.

Learning Center

• A Deep Look into AI Revolution in Health & Medicine  AI is now being implemented inside operating rooms, critical care units for patient evaluation and even within our homes to support those with disabilities. Along with the promise of AI comes ethical and social implications — which are crucial when applied in health care — including the representativeness of datasets used in algorithm training to avoid perpetuating health inequities. Four leading experts share their views on how society navigates the new frontiers of artificial intelligence and the future of health care.

• Fine Tuning LLM Models – Generative AI Course Learn how to fine-tune LLM models. This course will teach you to fine-tune QLORA and LORA, as well as Quantization using LLama2, Gradient, and the Google Gemma model.

• Top Deep Learning Courses To Try In 2024 - MarkTechPost  Top deep learning courses to consider in 2024 include several highly regarded programs designed to enhance your understanding and practical skills in this field. The Deep Learning Specialization by Andrew Ng on Coursera offers a comprehensive introduction to neural networks and deep learning applications, making it an excellent choice for beginners. Stanford University's CS231n: Convolutional Neural Networks for Visual Recognition focuses on deep learning techniques for computer vision and is available for free online. Deep Learning A-Z: Hands-On Artificial Neural Networks on Udemy provides practical experience with building and training neural networks. For those looking for a fast-paced, application-oriented approach, Fast.ai's Practical Deep Learning for Coders is designed to get you building deep learning models quickly. The Deep Learning with TensorFlow by Google Cloud on Coursera offers a thorough introduction to TensorFlow and its applications in deep learning. Finally, Applied AI with DeepLearning by IBM on Coursera focuses on practical applications, including building, training, and deploying models using various frameworks. These courses offer a mix of theoretical knowledge and hands-on practice, catering to different expertise levels in deep learning. Have you watched any of them?

• Recording from LinkedIn Live on Gen AI implementation in telco https://youtu.be/UghwDwQaJdM?si=iuZcTl3p9fKL0iXn  

• Upcoming Linkedin Lives: 

• Gen AI for Agencies and Marketing, register here: https://www.linkedin.com/events/episode7-genaiforb2b-operations7194747477831360512/comments/ 
• Gen AI for Operations, register here: https://www.linkedin.com/events/episode6-genaiforb2b-agencies7196186296845746179/comments/ 

• Guide on how to deploy open LLMS with vLLM on Hugging Face Inference Endpoints using the custom container image. The author used the huggingface_hub Python library to programmatically create and manage Inference Endpoints. It also showed how to test the Inference Endpoint using the OpenAI SDK. Deploy open LLMs with vLLM on Hugging Face Inference Endpoints 

Prompt of the week

I asked ChatGPT, “What are the most bizarre prompts it has ever received?” Here are the top 3. Which one is your favorite? Mine is definitely the last one. 

So, I had to ask DALL-E 3 to create a graphic of that cat society, and this is what it came up with. What do you think? It’s sort of whimsical …

Tools and Resources

• Women and AI is a podcast and community empowering women in tech to shape the future of AI. On our podcast, we talk to women at the forefront of AI. They have had guests like Stephanie Dinkins AI Artist, Time’s 100 Most Influential People in AI and Kim Carson, CEO of Parallax Futures, creating a new way of designing the future with conceptual technologists. In each episode, we discuss how women are using AI, both to build AI solutions as well as in their everyday work and life.

You’re invited to join the community on LinkedIn, where female leaders are highlighted, and to discuss the latest trends in AI. Come join the conversation and help shape the future of AI. Tune in on Apple, Spotify, or wherever you find podcasts.

• 4. API Guide · google-ai-edge/model-explorer Wiki · GitHub provides convenient APIs to visualize models quickly from files or a PyTorch module and a lower-level API to visualize models from multiple sources.

• GitHub - naklecha/llama3-from-scratch: llama3 implementation one matrix multiplication at a time  llama3 implementation one matrix multiplication at a time.

• microsoft/Phi-3-medium-128k-instruct · Hugging Face  The model has underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures. When assessed against benchmarks testing common sense, language understanding, math, code, long context and logical reasoning, Phi-3-Medium-128K-Instruct showcased a robust and state-of-the-art performance among models of the same-size and next-size-up.

• The 20 Generative AI Coding Tools Every Programmer Should Know About  Notable tools include GitHub Copilot, which offers advanced code suggestions and integrates with popular editors like Visual Studio Code; TabNine, known for its accurate code predictions in over 20 languages; and CodeT5, an open-source tool from Salesforce that supports multiple languages and advanced programming capabilities. These tools help automate repetitive tasks, improve code quality, and boost productivity, making them invaluable aids for developers while still requiring human oversight to ensure code reliability. 

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.

This newsletter was curated on my beautiful porch. Where are you reading it?