- TechSolopreneur
- Posts
- ChatGPT’s First Anniversary
Image credit: Reddit user amanj203
Visit my company page 👆
☕️ Greetings, and welcome to my daily dive into the Generative AI landscape.
This newsletter is evolving with a goal to increase its value to the readers. You will continue to see structural changes in the coming days and your feedback is welcome.
In today’s issue:
The generative AI economy: Worth up to $7.9T.
Rogue superintelligence.
Making an image with generative AI uses as much energy as charging your phone.
Harnessing the Power of Generative AI in B2B Sales .
And more.
🔔 Please forward this newsletter to your friends and team members and invite them to join. This will help me grow my reach. Thanks, Qamar.
♨️ TOP STORY
In its first year, ChatGPT has remarkably transformed into a ubiquitous AI chatbot, propelled by its mobile app launches on iOS and Android, resulting in over 110 million downloads and nearly $30 million in revenue.
The Report Card:
ChatGPT's mobile app launches in May (iOS) and July (Android) 2023 significantly expanded its user base.
The apps have achieved 110 million combined installs and nearly $30 million in consumer spending.
Revenue is mainly from ChatGPT Plus subscriptions, priced at $19.99/month, offering enhanced features and access.
Despite its higher subscription cost compared to other services, consumer interest in ChatGPT Plus has grown.
Global consumer spending on ChatGPT's mobile apps is approximately $28.6 million.
ChatGPT faces competition from other AI chatbot apps, like Poe and Ask AI.
Ask AI, with its varied subscription options, leads in mobile revenue over ChatGPT.
ChatGPT's download figures are unprecedented, with rapid adoption on Android and consistent weekly downloads.
India and the U.S. are leading in ChatGPT downloads, closely competing for the top spot.
While ChatGPT is popular, it ranks third in usage behind Character AI and Chai, indicating diverse user preferences in AI apps.
Projections suggest ChatGPT will continue to grow in installs and revenue by the end of 2023.
📈 TRENDS
Generative AI is revolutionizing industries and consumer experiences, stimulating significant economic growth. Its broad applications range from problem-solving to creating digital content, impacting learning, work, and leisure. Economic forecasts vary, with estimates reaching trillions of dollars. This growth is driven by its vast potential in infrastructure services, digital advertising, and specialized software. Generative AI is expected to contribute immensely to global GDP, enhancing productivity and personal experiences beyond financial measures. Its influence spans both business and consumer sectors, indicating a profound and expanding economic footprint.
Reference: The generative AI economy: Worth up to $7.9T
🗞️ SUSTAINABILITY
A study by Hugging Face and Carnegie Mellon University reveals the significant carbon footprint of AI use. Generating an AI image consumes as much energy as a full smartphone charge, while text generation is less energy-intensive. The research, awaiting peer review, highlights the environmental impact of AI's operational phase, exceeding even its training phase in carbon emissions. The study compared the energy usage of various AI tasks, finding image generation notably carbon-heavy. It urges the adoption of specialized, energy-efficient models over large, general-purpose ones to reduce AI's environmental toll.
🗣️ CULTURE
Targeted online ads, based on personal data like names or interests, are increasingly common and raise privacy concerns. These ads, a result of surveillance capitalism, commodify personal information. With generative AI's entry into advertising, concerns about manipulation, prediction inaccuracies, and biased ads intensify. Such practices, while not new, are evolving rapidly with AI's integration. This raises significant consumer protection issues, necessitating legislative oversight and expert intervention. Existing consumer protection frameworks are insufficient to address these AI-driven challenges, highlighting the need for a new approach to safeguard individual privacy and prevent discriminatory practices in advertising.
🔦 INDUSTRY SPOTLIGHT
At AWS re:Invent 2023, a panel discussed Canadian companies' use of cloud and generative AI. Companies from various sectors shared their projects and rationale for choosing generative AI over traditional AI methods like machine learning (ML). Telus is enhancing customer service and smart home solutions with AI, focusing on ease of use and personalization. WAHI Realty uses AI for real estate, offering personalized home searches. Experience.Monks employs generative AI for efficient, customized advertising and creating realistic avatars. AlayaCare uses AI for healthcare scheduling and data management, improving efficiency and patient care. Wealthsimple leverages cloud-based generative AI for quicker, more responsive solutions. The panel acknowledged the advantages and challenges of generative AI, noting its speed, cost-effectiveness, and unique capabilities, especially in handling visual and unstructured data, highlighting its potential as a transformative tool in business.
⚙️ FUNCTION FOCUS
Generative AI is revolutionizing B2B sales, shifting focus from manual tasks to building high-value relationships. McKinsey reports 90% of commercial leaders plan to frequently use AI, enhancing sales ROI by up to 20%. AI streamlines processes like scoring and follow-ups, allowing sales teams to prioritize effectively. It offers automated lead scoring, predictive analytics, improved personalization, and efficient content creation. By automating routine tasks, generative AI boosts efficiency, enabling sales teams to focus on impactful interactions and relationships.
🧰 TIPS
This blog post, part two of a series, details accelerating generative AI models using PyTorch's new performance features. It builds on part one, which demonstrated an over 8x speedup in Segment Anything using PyTorch. This segment focuses on optimizing Large Language Models (LLMs). Key highlights include leveraging native PyTorch optimizations like Torch.compile, GPU quantization, Speculative Decoding, and Tensor Parallelism. These techniques notably boosted a from-scratch LLM's speed by nearly 10x without accuracy loss. The post also explores CPU overhead reduction, memory bandwidth bottlenecks, and the use of int8 and int4 weight-only quantization, concluding with the integration of tensor parallelism for further performance enhancement. The techniques are demonstrated with less than 1000 lines of PyTorch code, emphasizing PyTorch's simplicity, usability, and now, enhanced performance.
📦 USE CASES
Generative AI use cases in e-commerce:
Personalized Product Visualization: Utilizes generative AI to create custom visual representations of products based on customer preferences, such as style, color, and size. This enhances the shopping experience by allowing more tailored product interactions.
Virtual Try-Ons: Employs generative AI to provide customers with realistic digital fittings of clothes or accessories, considering factors like body shape, skin tone, and personal style, enhancing confidence in online purchases.
Human-Like Chatbots: Generative AI chatbots deliver more nuanced and empathetic customer interactions than traditional chatbots, providing better customer support and after-sales service with natural language processing capabilities.
Enhanced Product Discovery and Search Personalization: Improves product recommendations by analyzing user behavior and preferences, streamlining the shopping process with intuitive, conversational search interfaces and more accurate tagging.
Content Generation: AI-powered tools generate product descriptions, listings, and promotional materials, streamlining content creation for e-commerce platforms and reducing manual effort.
Market Research and Data Analysis: Assists in gathering and interpreting vast amounts of market data, identifying new market segments, and providing insights for strategy formulation.
Promotions and Marketing Campaigns: Analyzes customer and transaction data to create personalized customer loyalty programs, incentives, and effective marketing strategies, enhancing customer engagement and sales.
Optimizing Retail Media Networks (RMNs): Uses AI to analyze customer data, helping retailers attract suitable advertisers to their networks, and assists advertisers in optimizing ad spend and campaign configurations.
Supply Chain and Inventory Management: Enhances supply chain efficiency by predicting demand, optimizing stock levels, and improving logistics through analysis of sales data and market trends.
Gen AI-Driven Pricing Strategies: Employs AI models for price optimization based on market scenarios, demand curves, and competitive analysis, aiding in dynamic pricing decisions.
Fraud Detection and Prevention: Leverages generative AI to identify and adapt to new fraud patterns, enhancing the security of ecommerce transactions by analyzing customer behavior and historical fraud data.
👩💻👨💻 SKILLS & CAREERS
Udemy, a leader in online learning, observed significant growth in generative AI skills, particularly ChatGPT, since its launch a year ago. They offer over 1,000 ChatGPT-related courses in 25 languages, with 2.8 million enrollments, highlighting a global surge in AI education. Industries like professional services, technology, and education are keenly adopting these skills. Udemy's insights show AI's potential to automate 30% of work by 2030 and contribute $15 trillion globally. Their focus is on empowering workforce transformation and advancing a skills-based economy through AI education.
📚LEARNING
The evolution of Large Language Models (LLMs) like ChatGPT and LLaMA underscores a transformative era in AI. These models, initially pre-trained on colossal text datasets, undergo meticulous fine-tuning and reinforcement learning. Their training, demanding immense computational power and memory, often faces GPU limitations and synchronization challenges. While pre-training is resource-intensive, fine-tuning offers a cost-effective approach to specialize these models for specific tasks. The continuous refinement, including Reinforcement Learning from Human Feedback (RLHF), elevates their ability to align with human preferences. With deployment, challenges in model governance, security, and ethical considerations emerge. These aspects underscore the need for responsible AI development and usage, guiding LLMs towards safe and impactful applications.
Reference: LLM Explained: The LLM Training Landscape
🔬 RESEARCH
Multimodal language models integrate various data types like images, text, and audio, overcoming the limitations of text-only models. This paper defines multimodal concepts, traces their evolution, introduces products from major tech firms, provides a practical guide on technicalities, lists algorithms and datasets, explores applications, and discusses development challenges.
Reference: Multimodal Large Language Models: A Survey
💡 IDEAS
Ilya Sutskever, co-founder and chief scientist of OpenAI, is shifting his focus from developing generative models like GPT and DALL-E to preventing artificial superintelligence from going rogue. He sees a future where AI's power is widely recognized and humans may merge with machines. Sutskever, known for his deep thinking and modest lifestyle, believes in the potential of AI to revolutionize various aspects of life but is also cautious about its risks. His vision includes managing AI's immense capabilities responsibly, acknowledging the transformative impact of OpenAI's technologies like ChatGPT, and envisaging a future where AGI (Artificial General Intelligence) is a reality.
Your Feedback
I want this newsletter to be valuable to you so if there's anything on your mind—praises, critiques, or just a hello—please drop me a note. You can hit reply or shoot me a message directly at my email address: [email protected].
Join my community by subscribing to my newsletter below:
Join my LinkedIn group communities below:
Reply