Postdoctoral Fellow in neuroscience at Harvard, exploring the future of human-computer interaction.
We live in the era of the frictionless transaction. With a single click, a biometrically scanned thumbprint, or an automated smart contract on the blockchain, deals worth millions are finalized in seconds. Digital signatures have… Read More »The Value of a Handshake in a World of Digital Contracts
A highly sophisticated malware targets computers running Microsoft IIS servers. It causes billions in damages and exposes corporate vulnerabilities. Part of the 16 Internet Evolution Facts: The 2001 Edition archive. HistoricallyVerified
Dan Libby and Ramanathan V. Guha at Netscape develop RSS (RDF Site Summary). This allows users to “subscribe” to website updates, creating the foundation for news aggregators and future podcasts. Part of the 19 Internet Evolution Facts:… Read More »The Birth of RSS
Named directly after the core Ovidian myth of Galatea, this specific category strips away corporate PR and multi-billion dollar stock valuations to evaluate the underlying code of localized neural networks. It honors the open-source architectures… Read More »The Living Sculpture and Pure Intent in Model Design
“Can machines think?”. Part of the 15 AI Roots Facts: The Early Twentieth Century & Turing Era Edition archive. HistoricallyVerified
In 1837, Charles Babbage conceptualized the Analytical Engine, a mechanical, gear-driven computer designed to feature a processing unit (the “mill”) and memory storage (the “store”). It was the world’s first blueprint for a fully programmable,… Read More »The Analytical Engine Blueprints
Diederik Kingma and Jimmy Ba published a landmark paper presenting the Adam (Adaptive Moment Estimation) optimizer. By combining the principles of Momentum and RMSProp to compute adaptive learning rates for individual network parameters based on… Read More »The Introduction of the Adam Optimization Algorithm
Vinod Nair and Geoffrey Hinton published a landmark paper at the International Conference on Machine Learning (ICML) proving that substituting traditional sigmoid or tanh functions with Rectified Linear Units (ReLU) drastically accelerated the training speed… Read More »The Introduction of the Rectified Linear Unit (ReLU) Activation Function
Within her 1843 notes, Ada Lovelace made a profound conceptual leap, stating that the Analytical Engine could process objects other than numbers, such as music, graphics, and abstract symbols, predicting the exact trajectory of modern… Read More »Lovelace’s Vision of Non-Numerical AI
Long Ouyang and the OpenAI alignment team deployed InstructGPT, proving that using small human-labeled instruction datasets to train reward models via proximal policy optimization forced raw language models to follow explicit user prompts while drastically… Read More »The Formulation of the InstructGPT Alignment Paradigm