up:: [[DBA806 - Applied AI Innovation]] tags:: #source/course #on/AI people:: [[Praphul Chandra]] # DBA806 M1 - A Brief History of AI Innovations The lecture delves into the state of AI in 2022-23, highlighting the emergence of Generative AI, [[Foundation Models]], and their unexplainable nature. The text further discusses the impressive capabilities of AI, including surpassing human performance in various tasks and its ability to learn and adapt. The historical context of AI's development is explored, mentioning milestones like IBM's Deep Blue defeating Garry Kasparov in chess and DeepMind's AlphaGo mastering the game of Go. The text also outlines the evolution of AI in different domains, such as *speech recognition, image processing, and driving*, with examples like Siri, Waymo, and Tesla Autopilot. Several AI applications beyond traditional domains are discussed, including AI-generated images, deepfakes, and *AI's role in designing products and even serving as a CEO*. The progress in AI extends to scientific domains, with *applications like drug discovery* ([[MIT Halicin (2019)]]), *protein folding* (DeepMind [[AlphaFold (2020)]]), and *material analysis* ([[GNoME (2023)]]). The text emphasizes AI's increasing role in governance, economics, and the broader societal impact, touching on Salesforce [[AI Economist (2020)]]'s approach to *discovering effective tax policies*. Lastly, the text reflects on the rapid progress of AI in recent years, surpassing human capabilities in various fields. It raises fundamental questions about the implications of AI on humanity, ethical concerns, and the evolving relationship between machines and humans. #### Summary of Slide Deck **State of AI 2022-23:** - Emergence of Generative AI, [[Foundation Models]], and their unexplainable nature. - AI capabilities surpassing 90% human performance in tasks like SAT, GRE, and legal exams. - AI learning faster than humans, with models capable of learning from each other. **Historical AI Milestones:** - IBM's Deep Blue defeating Garry Kasparov in chess (1997). - DeepMind's [[AlphaGo (2016)]] mastering Go. - Evolution of AI applications in speech recognition, image processing, and autonomous driving. **Diverse AI Applications:** - AI-generated images using [[Generative Adversarial Network (GAN)]] and [[Variational Autoencoder (VAE)]]. - Deepfakes as synthetic media using GANs for manipulation. - AI in driving with examples like Tesla Autopilot and Alphabet Waymo One. **AI in Scientific Domains:** - [[MIT Halicin (2019)]] for drug discovery. - DeepMind [[AlphaFold (2020)]] for protein folding. - DeepMind [[GNoME (2023)]] for materials analysis. **AI in Governance and Economics:** - Salesforce [[AI Economist (2020)]] discovering effective tax policies using deep reinforcement learning. **Broader Societal Impact:** - Reflection on AI's rapid progress and its implications on humanity. - Ethical concerns and the evolving relationship between machines and humans. - *"If AI thinks (or approximates thinking), who are we?"* > Humans are creating and proliferating nonhuman forms of logic with reach and acuity that can exceed our own. - Kissinger, Henry A; Schmidt, Eric; Huttenlocher, Daniel. The Age of AI This comprehensive summary covers the key aspects of the provided text, highlighting the course details, AI advancements, historical milestones, diverse applications, scientific contributions, governance impact, and broader societal considerations. #### Key Discussions in the Lecture **Emerging Technologies and AI:** - Highlighting the significant increase in productivity due to AI. - Introduction of [[Foundation Models]], capable of diverse tasks. - Mentioning the power of [[Deep Learning]] in AI development. **Understanding Natural Language Processing:** - Challenges in [[Natural Language Processing (NLP)]], including context and syntax. - Real-world context and challenges in applying AI to [[Image Recognition]]. - The importance of addressing data quality issues in [[Facial Recognition]]. **Deep Learning and Multimodal AI:** - Introduction to [[Deep Learning]], a preferred algorithm in AI. - Challenges in explaining the workings of deep learning models. - Emergence of [[Multimodal AI]], combining text and image processing. **Synthetic Reality and Implications:** - Introduction to the concept of [[Synthetic Reality]]. - Illustration of AI-driven object detection in images. - Mention of the session exploring the potential impact and implications. **AI in Science:** - AI's role in setting tax rates for economic growth and wealth distribution. - Overview of AI-driven drug discovery systems using reinforcement learning. - Utilizing AI in material discovery and its comparison to AlphaFold for proteins. **Generative AI and Language Models:** - Application of generative AI in language models. - The emergence of [[Code Autocomplete]] on steroids for coding with tools like Co-pilot. - AI's ability to generate suggestions and real-time autocomplete for coding. **AI in Product Design and Creation:** - AI's role in creating the first entirely machine-generated deodorant. - Discussion on AI's potential impact on various professions, including economics. **AI's Evolution and Ethical Concerns:** - Reflection on AI's rapid evolution in less than 25 years. - Philosophical questions regarding machines displaying emotions. - Ethical concerns about AI's impact on social media content and [[Deepfake]] technology. **AI's Role in Healthcare:** - Encouragement to explore AI applications in human health. - The suggestion that *AI can track side effects and outcomes of medical interventions*.