Jeff Hirsh Read: February 2022
Inspiration: What are the latest trends in AI? What is the future outlook?
Summary
Written with the help of ChatGPT, below is a brief summary to understand what is covered in the book.
“AI 2041: Ten Visions for Our Future”, published in 2021 by author and AI expert Kai-Fu Lee, discusses the potential future developments in artificial intelligence and their potential impacts on society. The book is divided into ten chapters, each of which presents a vision for how AI could shape the world in the coming decades. Some of the potential developments discussed in the book include the creation of superintelligent AI, the use of AI for augmenting human intelligence, the potential for AI to solve some of the world’s most pressing problems, and the possible emergence of new forms of consciousness and intelligence. The book also addresses the ethical and societal implications of these developments, including issues related to employment, privacy, and the distribution of wealth. Overall, the book offers a thought-provoking and wide-ranging exploration of the future of AI and its potential impacts on humanity.
Unedited Notes
Direct from my original book log, below are my unedited notes (abbreviations and misspellings included) to show how I take notes as I read.
Deep learning is under AI umbrella—prominence in 2016 (works without human nudges, just give examples and it learns), deep learning maximizes objective function (e.g. probability of correct recognition, biases leak into AI via data so need AI audits and hippocratic oath from engineers, computers likely to required antideepfake software, computer vision (CV) gets computers to see/understand as humans do (process images, object detection/recognition/tracking), CV used to detect explicit content in videos, CV unlocks iphones with face, convolutional neural networks (CNNs) work like brain to help CV—process in layers (lower simple and higher gets intricate to build full picture), GAN (generative adversarial network)—make deepfakes, has forager network which tries to make something looks real and then detective network compares to real—forager iterates until equilibrium, NLP natural language processing—subbranch of AI, understand human communication, supervised NLP requires human input of correct answer, goal is self superviser general NLP, GPT-3=generative pretrained transformer, 2020 released to analyze language via 500k lifetimes of data, self supervised general NLP key to AI tutors, 2020 DeepMind’s AlphaFold 2 is greatest AI in science—determine protein folding, simulate 3D structure of inseen proteins—once know structure can develop drugs/repurpose, X Reality (XR) is umbrella for vr/ar/mr, with autonomous vehicles—question is who is accountable in accident (car manufacturer, AI algo provider, engineer who wrote code, human backup driver), quantum computing use qubits (superpostion and entanglement key new characteristics—multiple states at once and connected to one another), QC disrupted easily—quantum decoherence, deterrence theory works poorly for autonomous weapons b/c harder to trace source, AI not really designed to create jobs—replace human routine tasks (exacerbate inequality), UBI could be conditioned on enroll in training/education, AI cannot easily replace jobs with creativity, dexterity, empathy, AI improving happiness requires trusting entity with our data—unlike GDPR, entity to trust could be open source commune, blockchain, smaller country monarchy (need to have incentives aligned unlike public co), plenitude=with new tech, most needs/energy will be widely available and free (minimal scarcity), $218bn of food discsrded while cost to rid hunger $25bn, 5x empty houses to homeless people in ’21, lack of scarcity upends all econ models—corps will want to resist (requires monumental change to embrace plenitude)
