Genius Makers: The Mavericks Who Brought AI to Google, Facebook, and the World

Knowledge Graph: Genius Makers: The Mavericks Who Brought AI to Google, Facebook, and the World (Cade Metz, 2021)

Editorial spotlight: ↑ ImageNet 2012: the moment deep learning proved itself

Concepts

• AI Winter of the 1990s (importance 4): Period when neural networks were dismissed by mainstream AI community; funding dried up, careers stalled.. Source: (from training memory of book).
• AI talent war (importance 4): Tech companies competing aggressively for limited pool of deep learning experts, driving up salaries and acquisitions.. Source: (from training memory of book).
• The Toronto Mafia (importance 3): Nickname for Hinton's University of Toronto students who spread throughout tech industry (Krizhevsky, Sutskever, etc.).. Source: (from training memory of book).
• Tech industry AI ethics awakening (importance 3): Growing awareness of bias, fairness, and safety issues in deployed AI systems at major tech companies.. Source: (from training memory of book).
• Chinese AI industry acceleration (importance 3): China's massive investment in AI research and deployment, creating competitive pressure on Western tech giants.. Source: (from training memory of book).
• Symbolic AI orthodoxy (importance 3): Dominant AI approach of 1970s-1990s using logic and explicit rules, rejected by neural network pioneers.. Source: (from training memory of book).
• Parallel Distributed Processing (importance 3): Influential 1986 books by Rumelhart, McClelland, and Hinton reviving neural network research.. Source: (from training memory of book).
• Academic exodus to industry (importance 3): Mass migration of AI professors to tech companies offering huge salaries and compute resources.. Source: (from training memory of book).
• AI safety research field (importance 3): Emerging field studying how to ensure AI systems remain beneficial and aligned with human values.. Source: (from training memory of book).
• Industry research labs model (importance 3): Tech companies building academic-style research labs with publication freedom to attract talent.. Source: (from training memory of book).
• AI as geopolitical competition (importance 3): AI development framed as national security issue and race for technological supremacy.. Source: (from training memory of book).
• AGI timeline debates (importance 2): Disagreements among researchers about when artificial general intelligence might be achieved.. Source: (from training memory of book).
• Autonomous weapons debate (importance 2): Controversy over military applications of AI and calls to ban lethal autonomous weapons.. Source: (from training memory of book).
• Data labeling industry (importance 2): Emergence of global workforce manually labeling training data, often in developing countries.. Source: (from training memory of book).
• arXiv preprint culture (importance 2): Shift to posting research on arXiv before peer review, accelerating dissemination.. Source: (from training memory of book).
• Deep learning hype cycle (importance 2): Media and investor excitement about AI capabilities, sometimes exceeding current reality.. Source: (from training memory of book).
• AI startup boom (importance 2): Explosion of AI-focused startups raising venture capital, many acquired by tech giants.. Source: (from training memory of book).
• Open vs proprietary AI debate (importance 2): Tension between academic openness culture and commercial secrecy in AI development.. Source: (from training memory of book).
• AI job displacement concerns (importance 2): Fears about AI automating white-collar work and need for workforce adaptation.. Source: (from training memory of book).
• AI interpretability research (importance 2): Efforts to understand and explain neural network decision-making processes.. Source: (from training memory of book).
• AI regulation debates (importance 2): Discussions about whether and how to regulate AI development and deployment.. Source: (from training memory of book).
• AI for social good initiatives (importance 1): Efforts to apply AI to healthcare, climate, poverty reduction and other social challenges.. Source: (from training memory of book).
• AI chip startup wave (importance 1): Dozens of startups building specialized AI processors to challenge NVIDIA.. Source: (from training memory of book).
• Edge AI deployment (importance 1): Running AI models on devices rather than cloud for privacy and latency.. Source: (from training memory of book).
• Explainable AI requirements (importance 1): Push for AI systems that can justify their decisions, especially in regulated domains.. Source: (from training memory of book).

Claims

• Compute scaling hypothesis (importance 4): Observation that model performance scales predictably with compute, data, and parameters.. Source: (from training memory of book).
• End of AI winters (importance 4): After ImageNet 2012, neural networks became unquestionably dominant, ending boom-bust cycle.. Source: (from training memory of book).
• Sutton's bitter lesson (importance 3): Rich Sutton's observation that general methods leveraging computation beat hand-crafted approaches long-term.. Source: (from training memory of book).
• Algorithmic bias in deployed systems (importance 3): Discovery that AI systems trained on biased data perpetuate and amplify societal biases.. Source: (from training memory of book).
• Google's AI infrastructure advantage (importance 3): Google's compute resources, data, and talent giving it structural advantage in AI race.. Source: (from training memory of book).
• Language model scaling laws (importance 3): Observation that larger language models show emergent capabilities and improved performance.. Source: (from training memory of book).
• Supervised learning dominance (importance 3): Despite early hopes for unsupervised learning, supervised learning on labeled data proved most effective.. Source: (from training memory of book).
• Minsky-Papert perceptron critique (importance 2): Minsky and Papert's 1969 book showing limitations of single-layer perceptrons, contributing to first AI winter.. Source: (from training memory of book).
• ML reproducibility challenges (importance 2): Difficulty reproducing published results due to missing implementation details and randomness.. Source: (from training memory of book).
• Debates on deep learning limitations (importance 2): Ongoing discussions about whether current approaches can achieve general intelligence.. Source: (from training memory of book).
• Data as competitive moat (importance 2): Companies with large datasets having structural advantage in training better AI systems.. Source: (from training memory of book).
• RL sample efficiency challenges (importance 2): Reinforcement learning requiring impractical amounts of experience compared to human learning.. Source: (from training memory of book).
• Adversarial example vulnerability (importance 2): Discovery that neural networks can be fooled by imperceptible input perturbations.. Source: (from training memory of book).

Empirical results

• ImageNet 2012 AlexNet victory (importance 5): Hinton's team (Krizhevsky, Sutskever) winning ImageNet competition by 10+ percentage points using deep convolutional nets, proving the approach at scale.. Source: (from training memory of book).
• AlphaGo defeats Lee Sedol (2016) (importance 5): DeepMind's AlphaGo beating world champion Lee Sedol 4-1, demonstrating superhuman performance in complex game.. Source: (from training memory of book).
• Superhuman Atari game performance (importance 3): DeepMind's DQN achieving human-level or better performance across dozens of Atari games.. Source: (from training memory of book).
• AlphaGo Zero (tabula rasa learning) (importance 3): AlphaGo version that learned from pure self-play without human game data, surpassing original.. Source: (from training memory of book).
• AlphaFold protein folding (importance 3): DeepMind's breakthrough in predicting protein structures, solving 50-year biology problem.. Source: (from training memory of book).
• AlphaGo's Move 37 (importance 2): Shocking creative move in Game 2 that no human would play, demonstrating AI's novel strategies.. Source: (from training memory of book).
• Facial recognition deployment (importance 2): Deep learning enabling accurate face recognition, raising privacy and bias concerns.. Source: (from training memory of book).
• AI researcher salary inflation (importance 2): Top AI researchers commanding multi-million dollar compensation packages at tech companies.. Source: (from training memory of book).
• Deep learning speech recognition (importance 2): Neural networks dramatically improving speech recognition accuracy, enabling voice assistants.. Source: (from training memory of book).
• Deepfake generation capability (importance 2): GANs enabling realistic fake video generation, raising misinformation concerns.. Source: (from training memory of book).
• Neural recommendation systems (importance 2): Deep learning improving content recommendation on YouTube, Netflix, social media.. Source: (from training memory of book).
• AI content moderation deployment (importance 1): Deep learning used to detect harmful content on social platforms.. Source: (from training memory of book).

Methods

• LeCun's convolutional neural networks (importance 5): Yann LeCun's architecture for recognizing handwritten digits at Bell Labs, basis for modern computer vision.. Source: (from training memory of book).
• GPU training breakthrough (importance 5): Discovery that graphics processing units could massively accelerate neural network training through parallel computation.. Source: (from training memory of book).
• Attention mechanism (importance 4): Bahdanau and Bengio's attention allowing neural networks to focus on relevant parts of input, enabling transformers.. Source: (from training memory of book).
• Generative Adversarial Networks (GANs) (importance 4): Ian Goodfellow's invention of GANs allowing neural networks to generate realistic images through adversarial training.. Source: (from training memory of book).
• Transformer architecture (Vaswani et al.) (importance 4): Google's 'Attention is All You Need' architecture replacing recurrence with self-attention, enabling modern LLMs.. Source: (from training memory of book).
• Neural machine translation breakthrough (importance 3): Deep learning replacing statistical methods for language translation at Google, massive quality improvement.. Source: (from training memory of book).
• Hinton's Restricted Boltzmann Machines (importance 3): Unsupervised learning method Hinton developed for pre-training deep networks before supervised fine-tuning.. Source: (from training memory of book).
• Word2vec embeddings (importance 3): Google's method for learning word representations that capture semantic relationships.. Source: (from training memory of book).
• DeepMind's Deep Q-Learning (importance 3): Combining deep neural networks with reinforcement learning to play Atari games from pixels.. Source: (from training memory of book).
• AlphaGo self-play training (importance 3): AlphaGo playing millions of games against itself to discover superhuman strategies.. Source: (from training memory of book).
• BERT language model (importance 3): Google's bidirectional transformer model achieving state-of-art on language understanding tasks.. Source: (from training memory of book).
• PyTorch framework (importance 3): Facebook's dynamic computation graph framework becoming dominant in research community.. Source: (from training memory of book).
• TensorFlow framework (importance 3): Google's open-source deep learning framework enabling widespread adoption.. Source: (from training memory of book).
• Google's Tensor Processing Units (importance 3): Custom AI chips designed for neural network training and inference at massive scale.. Source: (from training memory of book).
• Transfer learning paradigm (importance 3): Pre-training on large datasets then fine-tuning for specific tasks, became dominant approach.. Source: (from training memory of book).
• Residual connections (ResNet) (importance 3): Skip connections enabling training of very deep networks (100+ layers).. Source: (from training memory of book).
• Dropout regularization (importance 2): Hinton's technique of randomly dropping neurons during training to prevent overfitting.. Source: (from training memory of book).
• ReLU activation function adoption (importance 2): Rectified Linear Units replacing sigmoid/tanh activations, enabling deeper networks to train effectively.. Source: (from training memory of book).
• Monte Carlo Tree Search in AlphaGo (importance 2): Classical search algorithm combined with neural networks in AlphaGo for move evaluation.. Source: (from training memory of book).
• Neural architecture search (importance 2): Using AI to automatically design neural network architectures.. Source: (from training memory of book).
• Few-shot learning (importance 2): Training models to learn from minimal examples, mimicking human learning.. Source: (from training memory of book).
• Batch normalization (importance 2): Technique for stabilizing and accelerating neural network training.. Source: (from training memory of book).
• Multi-modal learning approaches (importance 2): Training models on multiple data types simultaneously (text, images, audio).. Source: (from training memory of book).
• Meta-learning (learning to learn) (importance 2): Training models to quickly adapt to new tasks with minimal data.. Source: (from training memory of book).
• Simulation to reality transfer (importance 1): Training AI in simulation then deploying to real world, especially for robotics.. Source: (from training memory of book).

Entities

• Hinton's backpropagation revival (1986) (importance 5): Geoffrey Hinton's work with Rumelhart and Williams showing backpropagation could train multi-layer networks, rejecting symbolic AI orthodoxy.. Source: (from training memory of book).
• DeepMind founding (Hassabis 2010) (importance 5): Demis Hassabis, Shane Legg, and Mustafa Suleyman founding DeepMind to build general AI through deep reinforcement learning.. Source: (from training memory of book).
• Demis Hassabis (importance 5): Child prodigy, games programmer, neuroscientist who founded DeepMind to pursue artificial general intelligence.. Source: (from training memory of book).
• Yoshua Bengio (importance 5): Montreal-based pioneer who maintained neural network research through AI winters, trained key researchers.. Source: (from training memory of book).
• Yann LeCun (importance 5): French researcher who developed convolutional nets at Bell Labs, later led Facebook AI Research.. Source: (from training memory of book).
• Geoffrey Hinton (importance 5): British cognitive psychologist who championed neural networks for decades, mentored generation of leaders.. Source: (from training memory of book).
• Bengio's Montreal lab (importance 4): Yoshua Bengio's research group at University of Montreal, training ground for key deep learning practitioners.. Source: (from training memory of book).
• Google Brain project (Ng + Dean) (importance 4): Andrew Ng and Jeff Dean's internal Google project building massive neural networks, including the famous cat-recognizing network.. Source: (from training memory of book).
• Hinton's 2013 company auction (importance 4): Hinton and his students (Krizhevsky, Sutskever) starting a company and auctioning it to tech giants, eventually acquired by Google.. Source: (from training memory of book).
• Google's £400M DeepMind acquisition (importance 4): Google acquiring DeepMind in 2014 for roughly $650 million, largest AI acquisition at the time.. Source: (from training memory of book).
• Facebook AI Research (FAIR) under LeCun (importance 4): Yann LeCun building Facebook's AI research lab with academic-style openness and publication culture.. Source: (from training memory of book).
• OpenAI founding (Altman + Musk 2015) (importance 4): Sam Altman and Elon Musk founding OpenAI as nonprofit to ensure safe development of artificial general intelligence.. Source: (from training memory of book).
• Alex Krizhevsky (importance 4): Hinton's student who built AlexNet, the ImageNet-winning network that proved deep learning at scale.. Source: (from training memory of book).
• Ilya Sutskever (importance 4): Hinton's student, co-creator of AlexNet, later Google Brain researcher and OpenAI chief scientist.. Source: (from training memory of book).
• Andrew Ng (importance 4): Stanford professor who co-founded Google Brain, later founded Coursera and led AI at Baidu.. Source: (from training memory of book).
• ImageNet database (Fei-Fei Li) (importance 4): Fei-Fei Li's massive labeled image database enabling supervised learning at unprecedented scale.. Source: (from training memory of book).
• Sutskever's move to OpenAI (importance 3): Ilya Sutskever leaving Google to become OpenAI's chief scientist, bringing deep learning expertise.. Source: (from training memory of book).
• Timnit Gebru at Google (importance 3): Timnit Gebru's work on AI ethics and fairness at Google, later conflict over research on large language model risks.. Source: (from training memory of book).
• Jeff Dean (importance 3): Google's legendary engineer who co-led Google Brain and championed AI infrastructure development.. Source: (from training memory of book).
• Ian Goodfellow (importance 3): Bengio student who invented GANs, worked at Google Brain and OpenAI.. Source: (from training memory of book).
• Fei-Fei Li (importance 3): Stanford professor who created ImageNet and later led Google Cloud AI.. Source: (from training memory of book).
• GPT-2 staged release controversy (importance 3): OpenAI's decision to initially withhold GPT-2 due to misuse concerns, sparking debate on AI safety.. Source: (from training memory of book).
• 2018 Turing Award (Hinton, LeCun, Bengio) (importance 3): Computing's Nobel Prize awarded jointly to the three pioneers for deep learning breakthroughs.. Source: (from training memory of book).
• OpenAI's shift to capped-profit (importance 3): OpenAI transitioning from pure nonprofit to hybrid structure to raise capital for AGI research.. Source: (from training memory of book).
• Microsoft-OpenAI partnership (importance 3): Microsoft investing $1 billion in OpenAI, providing Azure compute infrastructure.. Source: (from training memory of book).
• Lee Sedol (importance 2): Korean Go world champion defeated by AlphaGo in historic 2016 match.. Source: (from training memory of book).
• DeepMind Ethics & Society unit (importance 2): DeepMind's internal team studying societal implications of AI development.. Source: (from training memory of book).
• Google Project Maven controversy (importance 2): Employee revolt over Google's Pentagon contract for AI-powered drone imagery analysis.. Source: (from training memory of book).
• Google's AI Principles (importance 2): Google's public commitment to ethical AI development following employee activism.. Source: (from training memory of book).
• Goodfellow-Bengio-Courville textbook (importance 2): The definitive deep learning textbook codifying the field's knowledge.. Source: (from training memory of book).
• NeurIPS conference growth (importance 2): Neural Information Processing Systems conference exploding from hundreds to thousands of attendees.. Source: (from training memory of book).
• NVIDIA's AI windfall (importance 2): Graphics chip maker becoming dominant AI hardware supplier, stock price soaring.. Source: (from training memory of book).
• Company AI safety teams (importance 2): Dedicated teams at tech companies studying risks and safety of AI systems.. Source: (from training memory of book).
• Amazon Mechanical Turk (importance 1): Platform enabling crowdsourced data labeling for training AI systems.. Source: (from training memory of book).

Relations

• Hinton's backpropagation revival (1986) contradicts AI Winter of the 1990s
• Hinton's backpropagation revival (1986) contradicts Symbolic AI orthodoxy
• LeCun's convolutional neural networks contradicts AI Winter of the 1990s
• LeCun's convolutional neural networks enables ImageNet 2012 AlexNet victory
• Bengio's Montreal lab contradicts AI Winter of the 1990s
• Bengio's Montreal lab enables Ian Goodfellow
• GPU training breakthrough enables ImageNet 2012 AlexNet victory
• ImageNet 2012 AlexNet victory refutes AI Winter of the 1990s
• ImageNet 2012 AlexNet victory evidences End of AI winters
• ImageNet 2012 AlexNet victory motivates AI talent war
• Alex Krizhevsky enables ImageNet 2012 AlexNet victory
• Ilya Sutskever enables ImageNet 2012 AlexNet victory
• Geoffrey Hinton enables Hinton's backpropagation revival (1986)
• Geoffrey Hinton enables The Toronto Mafia
• Geoffrey Hinton enables Alex Krizhevsky
• Geoffrey Hinton enables Ilya Sutskever
• Yann LeCun enables LeCun's convolutional neural networks
• Yann LeCun enables Facebook AI Research (FAIR) under LeCun
• Yoshua Bengio enables Bengio's Montreal lab
• Yoshua Bengio enables Attention mechanism
• Hinton's 2013 company auction builds-on ImageNet 2012 AlexNet victory
• Hinton's 2013 company auction exemplifies AI talent war
• Google Brain project (Ng + Dean) requires Andrew Ng
• Google Brain project (Ng + Dean) requires Jeff Dean
• Google Brain project (Ng + Dean) builds-on GPU training breakthrough
• DeepMind founding (Hassabis 2010) requires Demis Hassabis
• Google's £400M DeepMind acquisition builds-on DeepMind founding (Hassabis 2010)
• Google's £400M DeepMind acquisition exemplifies AI talent war
• AlphaGo defeats Lee Sedol (2016) builds-on DeepMind founding (Hassabis 2010)
• AlphaGo defeats Lee Sedol (2016) builds-on DeepMind's Deep Q-Learning
• AlphaGo defeats Lee Sedol (2016) builds-on Monte Carlo Tree Search in AlphaGo
• AlphaGo defeats Lee Sedol (2016) builds-on AlphaGo self-play training
• AlphaGo defeats Lee Sedol (2016) refutes Lee Sedol
• AlphaGo's Move 37 exemplifies AlphaGo defeats Lee Sedol (2016)
• AlphaGo Zero (tabula rasa learning) builds-on AlphaGo defeats Lee Sedol (2016)
• AlphaGo Zero (tabula rasa learning) contradicts Supervised learning dominance
• Facebook AI Research (FAIR) under LeCun requires Yann LeCun
• Facebook AI Research (FAIR) under LeCun enables PyTorch framework
• OpenAI founding (Altman + Musk 2015) motivates AI safety research field
• Sutskever's move to OpenAI builds-on OpenAI founding (Altman + Musk 2015)
• Sutskever's move to OpenAI requires Ilya Sutskever
• Generative Adversarial Networks (GANs) requires Ian Goodfellow
• Generative Adversarial Networks (GANs) enables Deepfake generation capability
• Attention mechanism enables Transformer architecture (Vaswani et al.)
• Transformer architecture (Vaswani et al.) enables BERT language model
• Transformer architecture (Vaswani et al.) enables GPT-2 staged release controversy
• Transformer architecture (Vaswani et al.) enables Language model scaling laws
• ImageNet database (Fei-Fei Li) enables ImageNet 2012 AlexNet victory
• ImageNet database (Fei-Fei Li) requires Fei-Fei Li
• Dropout regularization enables ImageNet 2012 AlexNet victory
• ReLU activation function adoption enables ImageNet 2012 AlexNet victory
• Parallel Distributed Processing enables Hinton's backpropagation revival (1986)
• Hinton's Restricted Boltzmann Machines enables ImageNet 2012 AlexNet victory
• DeepMind's Deep Q-Learning enables Superhuman Atari game performance
• Superhuman Atari game performance motivates AlphaGo defeats Lee Sedol (2016)
• Neural machine translation breakthrough builds-on Attention mechanism
• The Toronto Mafia exemplifies Alex Krizhevsky
• The Toronto Mafia exemplifies Ilya Sutskever
• Word2vec embeddings exemplifies Transfer learning paradigm
• Residual connections (ResNet) builds-on ImageNet 2012 AlexNet victory
• Batch normalization builds-on ImageNet 2012 AlexNet victory
• TensorFlow framework builds-on Google Brain project (Ng + Dean)
• PyTorch framework builds-on Facebook AI Research (FAIR) under LeCun
• Google's Tensor Processing Units builds-on Google Brain project (Ng + Dean)
• Compute scaling hypothesis builds-on GPU training breakthrough
• Compute scaling hypothesis supports Language model scaling laws
• Sutton's bitter lesson supports Compute scaling hypothesis
• Transfer learning paradigm enables BERT language model
• Transfer learning paradigm exemplifies Supervised learning dominance
• Academic exodus to industry exemplifies AI talent war
• Academic exodus to industry motivates Industry research labs model
• Timnit Gebru at Google exemplifies Tech industry AI ethics awakening
• Algorithmic bias in deployed systems motivates Tech industry AI ethics awakening
• Facial recognition deployment exemplifies Algorithmic bias in deployed systems
• Google Project Maven controversy exemplifies Autonomous weapons debate
• Google Project Maven controversy motivates Google's AI Principles
• Chinese AI industry acceleration exemplifies AI as geopolitical competition
• Data labeling industry enables ImageNet database (Fei-Fei Li)
• Amazon Mechanical Turk enables Data labeling industry
• AlphaFold protein folding builds-on DeepMind founding (Hassabis 2010)
• GPT-2 staged release controversy exemplifies AI safety research field
• OpenAI's shift to capped-profit builds-on OpenAI founding (Altman + Musk 2015)
• Microsoft-OpenAI partnership enables OpenAI's shift to capped-profit
• NeurIPS conference growth evidences End of AI winters
• AI researcher salary inflation evidences AI talent war
• NVIDIA's AI windfall builds-on GPU training breakthrough
• Deep learning speech recognition builds-on ImageNet 2012 AlexNet victory
• Neural recommendation systems builds-on Transfer learning paradigm
• AI content moderation deployment builds-on ImageNet 2012 AlexNet victory
• Adversarial example vulnerability motivates AI safety research field
• 2018 Turing Award (Hinton, LeCun, Bengio) exemplifies Geoffrey Hinton
• 2018 Turing Award (Hinton, LeCun, Bengio) exemplifies Yann LeCun
• 2018 Turing Award (Hinton, LeCun, Bengio) exemplifies Yoshua Bengio
• 2018 Turing Award (Hinton, LeCun, Bengio) evidences End of AI winters
• Minsky-Papert perceptron critique motivates AI Winter of the 1990s
• Symbolic AI orthodoxy precedes AI Winter of the 1990s
• Data as competitive moat exemplifies ImageNet database (Fei-Fei Li)
• Google's AI infrastructure advantage builds-on Data as competitive moat
• Google's AI infrastructure advantage builds-on Google's Tensor Processing Units
• Industry research labs model exemplifies Google Brain project (Ng + Dean)
• Industry research labs model exemplifies Facebook AI Research (FAIR) under LeCun
• Industry research labs model exemplifies DeepMind founding (Hassabis 2010)
• arXiv preprint culture exemplifies Open vs proprietary AI debate
• ML reproducibility challenges contradicts arXiv preprint culture
• Goodfellow-Bengio-Courville textbook requires Yoshua Bengio
• Goodfellow-Bengio-Courville textbook requires Ian Goodfellow
• Few-shot learning builds-on Meta-learning (learning to learn)
• Neural architecture search builds-on Compute scaling hypothesis
• Multi-modal learning approaches builds-on Transformer architecture (Vaswani et al.)
• RL sample efficiency challenges contradicts DeepMind's Deep Q-Learning
• AI interpretability research supports AI safety research field
• Explainable AI requirements requires AI interpretability research
• Company AI safety teams exemplifies DeepMind Ethics & Society unit
• AI regulation debates builds-on Tech industry AI ethics awakening
• AI startup boom builds-on ImageNet 2012 AlexNet victory
• Deep learning hype cycle builds-on ImageNet 2012 AlexNet victory
• Debates on deep learning limitations supports AGI timeline debates
• AI job displacement concerns builds-on Deep learning speech recognition
• AI job displacement concerns builds-on Neural machine translation breakthrough
• Supervised learning dominance evidences ImageNet 2012 AlexNet victory
• Supervised learning dominance requires ImageNet database (Fei-Fei Li)

© 2026 LOOMUS. Curation, importance scoring, edge typing, and visualization. Source paper © respective authors, quoted under fair use.