-
Logical model of a binary “neuron” that inaugurated the connectionist approach to AI.
-
Proposes an operational test for machine intelligence and raises foundational questions.
-
Newell and Simon present one of the first AI programs, capable of proving logic theorems.
-
Workshop that defined the initial research agenda and coined “Artificial Intelligence.”
-
First supervised learning neural classifier; inspired decades of work on neural networks.
-
McCarthy creates LISP, a programming language central to AI research for decades.
-
Description: Weizenbaum demonstrates a dialogue system mimicking a therapist; showcased both potential and limits of early NLP.
-
Developed at SRI, it integrated perception, planning, and action in an autonomous robot.
-
Natural language system that “understood” and acted within a microworld; far ahead of its time.
-
Diagnostic system for infections; popularized rule-based “if–then” logic and fueled the expert system boom.
-
Large-scale industrial expert system for computer configuration; proved commercial value of symbolic AI.
-
Rumelhart, Hinton, and Williams systematized training multilayer networks via backpropagation, reviving neural networks.
-
Unmet expectations and funding cuts slowed progress; later revived by statistical and machine learning approaches.
-
Milestone in search-based AI; first victory against the reigning world chess champion.
-
Hinton and colleagues showed effective layer-wise training, relaunching deep learning.
-
Massive labeled image dataset that catalyzed benchmark-driven progress in vision.
-
howcased open-domain question answering through integrated IR and statistical methods.
-
Deep CNN drastically reduced error rates, sparking the modern deep learning wave in vision.
-
Goodfellow introduced adversarial training, foundational to modern generative AI.
-
Combined deep learning and Monte Carlo tree search; landmark in reinforcement learning.
-
New architecture based solely on attention; ushered in the era of foundation models.
-
Bidirectional pretraining revolutionized NLP by enabling fine-tuning across tasks.
-
Generative language model popularized zero/one/few-shot learning across diverse tasks.
-
Massive scaling (175B parameters) showcased strong in-context learning without fine-tuning.
-
Learned to plan without knowing environment rules; advanced reinforcement learning.
-
Predicted protein structures with near-atomic accuracy and released millions publicly.
-
Open-source latent diffusion model accelerated creativity and derivative ecosystems.
-
Conversational interface that democratized access to LLMs and triggered global adoption.
-
Text-to-image generation with unprecedented realism and compositionality.
-
Multimodal model with near-human performance across exams and reasoning tasks.
-
Meta released Llama 3, consolidating the wave of powerful open-source LLMs.
-
Unified model for text, vision, and audio in real time; reduced costs and latency.
-
OpenAI showcased high-fidelity video generation of up to ~1 minute, pushing multimodality forward.
-
OpenAI announced GPT-5, marking the next leap in capability and safety.