This Post Has 8 Comments

1. Superb history of AI/LLMs , practical tools, and future forecasts
   UCSD professor, Salk Institute for Biological Studies Chair, winner of Gruber Neuroscience and Brain Prizes, Dr. Sejnowski notes the debates around large language models’ (LLM) language comprehension, thinking and intelligence; gives a history of AI LLMs; discusses the intersection of AI and human intelligence; and discusses the potential of AI in current issues (energy efficiency, protein models and medical, appliances, science). Unique: Sejnowski’s using ChatGPT as ‘coauthor’, showing prompts and results for resources, edits, brainstorming.Personal takeaways: An insightful history and analysis of current/potential AI/LLM integration with human development, with actionable LLM prompt examples. Thought-provoking for companies and board directors to assess AI’s potential, an excellent read.Key takeaways, by Parts and Chapters:I: Living with LLMs1) Describes LLMs and how they’re used in various fields, identifies jobs challenged and created by AI/internet, discusses the impact of training data and limitations on LLM understanding, autonomy and human interaction.2) How chatbots change lives: help professionals be more productive, efficient and focused (generate text, translate language, answer questions, summarize text, write creative content), in healthcare, finance, law, education, programing, architecture, movies, music.3) Example interviews and conclusions with LaMDA, Chat 3.2, GPT-3, GPT 3.3, Bing GPT-4, GPT 3.5.4) Priming pre-trained LLMs with prompts influences subsequent responses. Examples: prompt pictures, persona, engineering. See p. 70 for how to improve results.5) Intelligence – LLMs’ math understanding can be a good starting point for a new theory of intelligence. Thinking – LLMs can manage convergent, divergent, critical and creative thinking. Consciousness – LLMs demonstrate qualia, self-awareness, introspection, states, artificial consciousness.Part II: Transformers6) Origins (waves) of deep learning: 1st 1960, 2nd 1985, 3rd 2010; LLM technical terms (transformer, token, etc) and guidance on how to train transformers, set context length, represent symbols and fine-tune.7) LLMs built on math foundations, extending to high-dimensional spaces8) LLMs used in customer service, content creation, data analysis, business intelligence, internal communication, sales/lead generation, training, process automation, R&D.9) Can AI be superintelligent? “Move forward optimistically, expect surprises and prepare for unintended consequences.”10) LLMs touch on privacy, copyright, liability, employment, discrimination laws/regulation.III Back to the Future11) AI evolved from rule-based symbol processing to data-driven machine learningNext gen will:12) integrate human-like hippocampus, biochemical activity and neuro-modulatory systems.13) learn from brains, language, sensorimotor systems.

   Reply

2. Understand LLMs Better: Insights from a Pioneer
   I recommend this book to general readers looking to further educate themselves about LLMs such as chatGPT, especially those skeptical about them. The author, who has worked for decades in the neural network research that preceded the development of LLMs, uses chatGPT to analyze and summarize his writing in the book. You can assess the depth of analysis for yourself and, in the process, begin to learn how to use LLMs effectively.You won’t fully understand how transformers work by reading this book, but you will increase your LLM knowledge. Comprehensive footnotes.

   Reply

3. Insightful
   I found chapters 1-5 most interesting, with the remainder of the book being too technical for me. I like to know how to drive a car and where to go (Part 1), rather than how the car’s systems were developed and how they work together (Parts 2 and 3). Worth a read, though.

   Reply

4. Great book about a large language model from a great neuroscientist
   Good lay explanantion of ChatGPT. The author is very talented.

   Reply

5. Excellent
   Excellent history, current status, applications and in general very enjoyable. Been watching this field for a couple decades and am just amazed.

   Reply

6. That it cover the subject thoroughly
   It was very informative and well written

   Reply

7. Fantastic and Heavy information
   We are in the new age of technology. We need to coexist with LLMs to unleash our full potentially. Beside that, LLMs is a technology which can be used by Angel and Evil. We need to set in mind that we use it to support for a more peaceful world.

   Reply

8. Perspectives on what to expect from AI in years to come
   This is a uniquely challenging book to discuss because Terrence J. Sejnowski makes brilliant use of a highly unorthodox structure to provide information, insights, and counsel about the “Deep Language Revolution” from highly unorthodox sources: ChatGPT and four Large Langauge Models (LLMs).As Sejnowski explains, “This book has three parts: part I gives you an overview of where LLMs are today; Part II introduces you to the mysterious transformer at the heart of all LLMs; and part III is about what to expect in the future. The focus is on how AI is evolving and its impact on society. A theme throughout this book is the convergence of AI and neuroscience, accelerating advances in AI and our understanding of brains. Along the way, there will be diversions to put LLMs in historical perspective.”These are among the subjects and issues of greatest interest and value to me, also listed to indicate the nature and scope of Sejnowsli’s coverage:o  Future of AI (Pages xii-xiii)o Using Large Language Models (17-19)o Computer Programming (40-42)o Prompting a Picture (64-66)o Prompting to Teach (67-68o The essence of intelligence (78-81o Reflection and analysis used as a process (81-83)o Causality (84)o Age of Information and deep learning (213-214)o The “California Gold Rush” (135-137)o Powering AI (142-143)o AI “at the Edge” (145-146)o Sam Altman’s significance (163-164o Why LLMs need a long-term memory (183-187o How Brains Evolved (189-191)o Reverse engineering brains (191-194)o How language evolved (194-197)o Deep learning and biology (206-209 and 213-214)o Learning < > Programmingo What can be learned from nature (201-202)I commend Terrence Sejnowski for the abundance of invaluable material that he provides in a book that will be embraced by technologists because he explains so much in ways and to an extent that all senior-level executives will be able to understand. (Personal note: My undergraduate and graduate studies were in the field of comparative literature. For me and probably for many others, much of the nomenclature associated with the Deep Language  Revolution is indeed “foreign.”) I view this book as a must-read for all senior-level executives. It is a brilliant achievement. Bravo!* * *Here are two suggestions while you are reading ChatGPT and the Future of AI: First, highlight key passages Also,  perhaps in a notebook kept near-at-hand (e.g. Apica Premium C.D. Notebook A5), record your comments, questions, and action steps (preferably with deadlines).These two simple tactics — highlighting and documenting — will expedite frequent reviews of key material later.

   Reply