All Quotes in AI

Explore 366 quotations from researchers, innovators, writers, and pioneers who have shaped how we understand artificial intelligence.

QUOTE OF THE DAY IN AI
Conceptual illustration for today’s AI quote
AI QUOTE

“We propose complexity-based prompting, a simple and effective example selection scheme for multi-step reasoning.”

— Yao Fu and colleagues

Explore the quote

Browse every AI quote

Organized alphabetically by the first named speaker’s surname.

Select a letter to jump to that section.

A11 quotes

AI QUOTE

Martín Abadi and colleagues

The models should not expose private information in these datasets.
Explore the quote
AI QUOTE

Marah Abdin and colleagues

We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens.
Explore the quote
AI QUOTE

Josh Achiam and colleagues

We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs.
Explore the quote
AI QUOTE

Joshua Ainslie and colleagues

We introduce grouped-query attention, a generalization of multi-query attention.
Explore the quote
AI QUOTE

Jean-Baptiste Alayrac and colleagues

We introduce Flamingo, a family of Visual Language Models with this ability.
Explore the quote
AI QUOTE

Sam Altman

We believe it is essential to develop regulations that incentivize AI safety while ensuring that people are able to access the technology’s many benefits.
Explore the quote
AI QUOTE

Dario Amodei

The single most important thing to understand about AI is how fast it is moving.
Explore the quote
AI QUOTE

Marcin Andrychowicz and colleagues

Dealing with sparse rewards is one of the biggest challenges in Reinforcement Learning.
Explore the quote
AI QUOTE

Rohan Anil and colleagues

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding.
Explore the quote
AI QUOTE

Anish Athalye and colleagues

We identify obfuscated gradients, a kind of gradient masking, as a phenomenon that leads to a false sense of security.
Explore the quote
AI QUOTE

Audrey Azoulay

The world needs stronger ethical rules for artificial intelligence: this is the challenge of our time.
Explore the quote
Back to A–Z ↑

B25 quotes

AI QUOTE

Jimmy Lei Ba and colleagues

Unlike batch normalization, layer normalization performs exactly the same computation at training and test times.
Explore the quote
AI QUOTE

Alexei Baevski and colleagues

This demonstrates the feasibility of speech recognition with limited amounts of labeled data.
Explore the quote
AI QUOTE

Dzmitry Bahdanau and colleagues

We conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture.
Explore the quote
AI QUOTE

Yuntao Bai and colleagues

We would like to enlist their help to supervise other AIs.
Explore the quote
AI QUOTE

Solon Barocas and Andrew D. Selbst

Data mining can discover surprisingly useful regularities that are really just preexisting patterns of exclusion and inequality.
Explore the quote
AI QUOTE

Iz Beltagy and colleagues

We introduce the Longformer with an attention mechanism that scales linearly with sequence length.
Explore the quote
AI QUOTE

Emily M. Bender and colleagues

A language model is a system for haphazardly stitching together sequences of linguistic forms.
Explore the quote
AI QUOTE

Yoshua Bengio

We must move from concentrated power toward real global dialogue on how this technology is developed for the public good.
Explore the quote
AI QUOTE

Ruha Benjamin

Indifference to social reality is a powerful force that is perhaps more dangerous than malicious intent.
Explore the quote
AI QUOTE

David Berthelot and colleagues

We unify the current dominant approaches for semi-supervised learning to produce a new algorithm, MixMatch.
Explore the quote
AI QUOTE

Maciej Besta and colleagues

We introduce Graph of Thoughts: a framework that advances prompting capabilities in large language models.
Explore the quote
AI QUOTE

Abeba Birhane

Algorithmic tools embed and perpetuate societal and historical biases and injustice.
Explore the quote
AI QUOTE

Erdem Bıyık and colleagues

We introduce several approximations to the batch active learning problem, and provide theoretical guarantees for the convergence of our algorithms.
Explore the quote
AI QUOTE

David M. Blei and colleagues

We describe latent Dirichlet allocation, a generative probabilistic model for collections of discrete data such as text corpora.
Explore the quote
AI QUOTE

Doreen Bogdan-Martin

We need guardrails to ensure AI is safe, responsible and inclusive.
Explore the quote
AI QUOTE

Rishi Bommasani and colleagues

We call these models foundation models to underscore their critically central yet incomplete character.
Explore the quote
AI QUOTE

Nick Bostrom

Machine intelligence is the last invention that humanity will ever need to make.
Explore the quote
AI QUOTE

Cynthia Breazeal

The goal is to build a socially intelligent machine that learns things as we learn them, through social interactions.
Explore the quote
AI QUOTE

Leo Breiman

There are two cultures in the use of statistical modeling to reach conclusions from data.
Explore the quote
AI QUOTE

Tom B. Brown and colleagues

Scaling up language models greatly improves task-agnostic, few-shot performance, sometimes reaching competitiveness with prior fine-tuning approaches.
Explore the quote
AI QUOTE

Joanna Bryson

Both AI and ethics are artefacts, so there is no necessary position for AI artefacts in society.
Explore the quote
AI QUOTE

Collin Burns and colleagues

We propose directly finding latent knowledge inside the internal activations of a language model in a purely unsupervised way.
Explore the quote
Back to A–Z ↑

C21 quotes

AI QUOTE

Nicolas Carion and colleagues

We present a new method that views object detection as a direct set prediction problem.
Explore the quote
AI QUOTE

Nicholas Carlini and colleagues

This makes it difficult to apply neural networks in security-critical areas.
Explore the quote
AI QUOTE

Mathilde Caron and colleagues

We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels.
Explore the quote
AI QUOTE

Mark Chen and colleagues

We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities.
Explore the quote
AI QUOTE

Ricky T. Q. Chen and colleagues

We introduce a new family of deep neural network models.
Explore the quote
AI QUOTE

Tianqi Chen and Carlos Guestrin

We describe a scalable end-to-end tree boosting system called XGBoost.
Explore the quote
AI QUOTE

Ting Chen and colleagues

This paper presents SimCLR: a simple framework for contrastive learning of visual representations.
Explore the quote
AI QUOTE

Wei-Lin Chiang and colleagues

We introduce Chatbot Arena, an open platform for evaluating LLMs based on human preferences.
Explore the quote
AI QUOTE

Kyunghyun Cho and colleagues

In this paper, we propose a novel neural network model called RNN Encoder-Decoder that consists of two recurrent neural networks.
Explore the quote
AI QUOTE

François Chollet

The intelligence of a system is a measure of its skill-acquisition efficiency over a scope of tasks.
Explore the quote
AI QUOTE

Noam Chomsky, Ian Roberts, and Jeffrey Watumull

The human mind is not, like ChatGPT and its ilk, a lumbering statistical engine for pattern matching.
Explore the quote
AI QUOTE

Krzysztof Choromanski and colleagues

We introduce Performers, Transformer architectures which can estimate regular full-rank-attention Transformers with provable accuracy.
Explore the quote
AI QUOTE

Aakanksha Chowdhery and colleagues

Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning.
Explore the quote
AI QUOTE

Paul Christiano and colleagues

We explore a new direction for scaling up reinforcement learning from human feedback.
Explore the quote
AI QUOTE

Hyung Won Chung and colleagues

Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization.
Explore the quote
AI QUOTE

Kevin Clark and colleagues

As an alternative, we propose a more sample-efficient pre-training task called replaced token detection.
Explore the quote
AI QUOTE

Karl Cobbe and colleagues

To increase performance, we propose training verifiers to judge the correctness of model completions.
Explore the quote
AI QUOTE

Corinna Cortes and Vladimir Vapnik

The support-vector network is a new learning machine for two-group classification problems.
Explore the quote
AI QUOTE

Kate Crawford

Artificial intelligence is neither artificial nor intelligent.
Explore the quote
Back to A–Z ↑

D16 quotes

AI QUOTE

Wenliang Dai and colleagues

We conduct a systematic and comprehensive study on vision-language instruction tuning based on the pretrained BLIP-2 models.
Explore the quote
AI QUOTE

Zihang Dai and colleagues

We propose a novel neural architecture Transformer-XL that enables learning dependency beyond a fixed length without disrupting temporal coherence.
Explore the quote
AI QUOTE

Tri Dao and colleagues

We argue that a missing principle is making attention algorithms IO-aware.
Explore the quote
AI QUOTE

Yann N. Dauphin and colleagues

We propose a novel simplified gating mechanism and investigate the impact of key architectural decisions.
Explore the quote
AI QUOTE

Jeffrey Dean and colleagues

We have developed a software framework called DistBelief that can utilize computing clusters with thousands of machines to train large models.
Explore the quote
AI QUOTE

Jiankang Deng and colleagues

We first introduce an Additive Angular Margin Loss, which not only has a clear geometric interpretation but also significantly enhances discriminative power.
Explore the quote
AI QUOTE

Emily Denton and colleagues

We propose that every dataset be accompanied with a datasheet that documents its motivation, composition, collection process, and recommended uses.
Explore the quote
AI QUOTE

Tim Dettmers and colleagues

QLoRA backpropagates gradients through a frozen, 4-bit quantized pretrained language model into Low Rank Adapters.
Explore the quote
AI QUOTE

Jacob Devlin and colleagues

We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers.
Explore the quote
AI QUOTE

Prafulla Dhariwal and Alex Nichol

We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models.
Explore the quote
AI QUOTE

Laurent Dinh and colleagues

We extend the space of such models using real-valued non-volume preserving transformations.
Explore the quote
AI QUOTE

Carl Doersch and colleagues

This work explores the use of spatial context as a source of free and plentiful supervisory signal for training a rich visual representation.
Explore the quote
AI QUOTE

Finale Doshi-Velez and Been Kim

Interpretability is the ability to explain or to present in understandable terms to a human.
Explore the quote
AI QUOTE

Alexey Dosovitskiy and colleagues

A pure transformer applied directly to sequences of image patches can perform very well on image classification tasks.
Explore the quote
AI QUOTE

Danny Driess and colleagues

We propose embodied language models to directly incorporate real-world continuous sensor modalities into language models.
Explore the quote
AI QUOTE

Cynthia Dwork and colleagues

The main conceptual contribution of this paper is a framework for fair classification.
Explore the quote
Back to A–Z ↑

E5 quotes

AI QUOTE

Nelson Elhage and colleagues

This paper provides a toy model where polysemanticity can be fully understood, arising as a result of models storing additional sparse features in superposition.
Explore the quote
AI QUOTE

Ahmed Elnaggar and colleagues

Taken together, the results implied that protein language models learned some of the grammar of the language of life.
Explore the quote
AI QUOTE

Jesse Engel and colleagues

We introduce the Differentiable Digital Signal Processing library, which enables direct integration of classic signal processing elements with deep learning methods.
Explore the quote
AI QUOTE

Andre Esteva and colleagues

Using a single convolutional neural network, we match the performance of at least 21 dermatologists.
Explore the quote
AI QUOTE

Virginia Eubanks

We’ll continue to create systems that not only automate inequality but amplify inequality.
Explore the quote
Back to A–Z ↑

F12 quotes

AI QUOTE

William Fedus and colleagues

Mixture of Experts defies this and instead selects different parameters for each incoming example.
Explore the quote
AI QUOTE

Chelsea Finn and colleagues

In effect, our method trains the model to be easy to fine-tune.
Explore the quote
AI QUOTE

Luciano Floridi

AI does not lead to any fanciful realization of science fiction scenarios.
Explore the quote
AI QUOTE

Jakob N. Foerster and colleagues

In these environments, agents must learn communication protocols in order to share information that is needed to solve the tasks.
Explore the quote
AI QUOTE

Pope Francis

No machine should ever choose to take the life of a human being.
Explore the quote
AI QUOTE

Jonathan Frankle and Michael Carbin

Dense, randomly-initialized, feed-forward networks contain subnetworks that—when trained in isolation—reach test accuracy comparable to the original network.
Explore the quote
AI QUOTE

Elias Frantar and colleagues

We propose GPTQ, a new one-shot weight quantization method based on approximate second-order information.
Explore the quote
AI QUOTE

Yoav Freund and Robert Schapire

The model we study can be interpreted as a broad, abstract extension of the well-studied on-line prediction model to a general decision-theoretic setting.
Explore the quote
AI QUOTE

Batya Friedman and colleagues

Value sensitive design accounts for human values in a principled and comprehensive manner throughout the design process.
Explore the quote
AI QUOTE

Yao Fu and colleagues

We propose complexity-based prompting, a simple and effective example selection scheme for multi-step reasoning.
Explore the quote
AI QUOTE

Scott Fujimoto and colleagues

We show that this problem persists in an actor-critic setting and propose novel mechanisms to minimize its effects on both the actor and the critic.
Explore the quote
AI QUOTE

Kunihiko Fukushima

A neural network model for a mechanism of visual pattern recognition is proposed in this paper.
Explore the quote
Back to A–Z ↑

G20 quotes

AI QUOTE

Yarin Gal and colleagues

However such tools for regression and classification do not capture model uncertainty.
Explore the quote
AI QUOTE

Deep Ganguli and colleagues

These generative models have an unusual combination of predictable loss on a broad training distribution, and unpredictable specific capabilities, inputs, and outputs.
Explore the quote
AI QUOTE

Bill Gates

The development of AI is as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone.
Explore the quote
AI QUOTE

Timnit Gebru

The machine learning community has no standardized way to document how and why a dataset was created.
Explore the quote
AI QUOTE

Jonas Gehring and colleagues

We introduce an architecture based entirely on convolutional neural networks.
Explore the quote
AI QUOTE

Robert Geirhos and colleagues

Shortcuts are decision rules that perform well on standard benchmarks but fail to transfer to more challenging testing conditions.
Explore the quote
AI QUOTE

Kristalina Georgieva

AI will transform the global economy. Let’s make sure it benefits humanity.
Explore the quote
AI QUOTE

Ross Girshick and colleagues

Since we combine region proposals with CNNs, we call our method R-CNN: Regions with CNN features.
Explore the quote
AI QUOTE

Ian Goodfellow and colleagues

We propose a new framework for estimating generative models via an adversarial process.
Explore the quote
AI QUOTE

Aaron Grattafiori and colleagues

This paper presents a new set of foundation models, called Llama 3.
Explore the quote
AI QUOTE

Alex Graves and colleagues

Recurrent neural networks (RNNs) are a powerful model for sequential data.
Explore the quote
AI QUOTE

Thomas L. Griffiths and colleagues

Tree of Thoughts enables exploration over coherent units of text that serve as intermediate steps toward problem solving.
Explore the quote
AI QUOTE

Jean-Bastien Grill and colleagues

We introduce Bootstrap Your Own Latent, a new approach to self-supervised image representation learning.
Explore the quote
AI QUOTE

Barbara Grosz

A computer system should make us feel smarter, not dumber and work seamlessly with us, like a human partner.
Explore the quote
AI QUOTE

Aditya Grover and Jure Leskovec

Here we propose node2vec, an algorithmic framework for learning continuous feature representations for nodes in networks.
Explore the quote
AI QUOTE

Albert Gu and colleagues

Foundation models, now powering most of the exciting applications in deep learning, are almost universally based on the Transformer architecture and its core attention module.
Explore the quote
AI QUOTE

Anmol Gulati and colleagues

In this work, we achieve the best of both worlds by studying how to combine convolution neural networks and transformers.
Explore the quote
AI QUOTE

Chuan Guo and colleagues

We discover that modern neural networks, unlike those from a decade ago, are poorly calibrated.
Explore the quote
AI QUOTE

António Guterres

The fate of humanity must never be left to the ‘black box’ of an algorithm.
Explore the quote
Back to A–Z ↑

H30 quotes

AI QUOTE

David Ha and Jürgen Schmidhuber

We explore building generative neural network models of popular reinforcement learning environments.
Explore the quote
AI QUOTE

Tuomas Haarnoja and colleagues

The actor aims to maximize expected reward while also maximizing entropy.
Explore the quote
AI QUOTE

Dylan Hadfield-Menell and colleagues

We propose a formal definition of the value alignment problem as cooperative inverse reinforcement learning.
Explore the quote
AI QUOTE

Danijar Hafner and colleagues

We present Dreamer, a reinforcement learning agent that solves long-horizon tasks from images purely by latent imagination.
Explore the quote
AI QUOTE

William L. Hamilton and colleagues

Here we present GraphSAGE, a general, inductive framework that leverages node feature information to efficiently generate node embeddings for previously unseen data.
Explore the quote
AI QUOTE

Song Han and colleagues

We introduce deep compression, a three stage pipeline: pruning, trained quantization and Huffman coding.
Explore the quote
AI QUOTE

Awni Hannun and colleagues

We present a state-of-the-art speech recognition system developed using end-to-end deep learning.
Explore the quote
AI QUOTE

Shibo Hao and colleagues

RAP repurposes the LLM as both a world model and a reasoning agent.
Explore the quote
AI QUOTE

Moritz Hardt and colleagues

We introduce the notion of equal opportunity as a nondiscrimination criterion for supervised learning.
Explore the quote
AI QUOTE

Demis Hassabis

Step one: Solve intelligence. Step two: Use it to solve everything else.
Explore the quote
AI QUOTE

Stephen Hawking

Success in creating AI could be the biggest event in the history of our civilisation.
Explore the quote
AI QUOTE

Kaiming He and colleagues

We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously.
Explore the quote
AI QUOTE

Pengcheng He and colleagues

We propose a new model architecture, DeBERTa, that improves the BERT and RoBERTa models using two novel techniques.
Explore the quote
AI QUOTE

Dan Hendrycks and colleagues

We propose a new test to measure a text model’s multitask accuracy.
Explore the quote
AI QUOTE

Matteo Hessel and colleagues

This paper examines six extensions to the DQN algorithm and empirically studies their combination.
Explore the quote
AI QUOTE

Jonathan Ho and colleagues

We present high quality image synthesis results using diffusion probabilistic models.
Explore the quote
AI QUOTE

Tin Kam Ho

The essence of the method is to build multiple trees in randomly selected subspaces of the feature space.
Explore the quote
AI QUOTE

Sepp Hochreiter and Jürgen Schmidhuber

We briefly review Hochreiter’s analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory.
Explore the quote
AI QUOTE

Jordan Hoffmann and colleagues

We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget.
Explore the quote
AI QUOTE

Ari Holtzman and colleagues

Our findings motivate Nucleus Sampling, a simple but effective method to draw the best out of neural generation.
Explore the quote
AI QUOTE

Sara Hooker

This essay introduces the term hardware lottery to describe when a research idea wins because it is suited to the available software and hardware.
Explore the quote
AI QUOTE

John Hopfield

Computational properties can emerge as collective properties of systems having a large number of simple equivalent components.
Explore the quote
AI QUOTE

Eric Horvitz

The value of what’s uniquely human will rise to become even more valued than it is today.
Explore the quote
AI QUOTE

Andrew G. Howard and colleagues

We present a class of efficient models called MobileNets for mobile and embedded vision applications.
Explore the quote
AI QUOTE

Jeremy Howard and colleagues

We propose Universal Language Model Fine-tuning, an effective transfer learning method that can be applied to any task in NLP.
Explore the quote
AI QUOTE

Edward J. Hu and colleagues

As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes less feasible.
Explore the quote
AI QUOTE

Jie Hu and colleagues

In this work, we focus instead on the channel relationship.
Explore the quote
AI QUOTE

Gao Huang and colleagues

In this paper, we embrace this observation and introduce the Dense Convolutional Network.
Explore the quote
Back to A–Z ↑

I3 quotes

AI QUOTE

Sergey Ioffe and Christian Szegedy

Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch.
Explore the quote
AI QUOTE

Geoffrey Irving and colleagues

We propose training agents via self play on a zero sum debate game.
Explore the quote
AI QUOTE

Phillip Isola and colleagues

We investigate conditional adversarial networks as a general-purpose solution to image-to-image translation problems.
Explore the quote
Back to A–Z ↑

J6 quotes

AI QUOTE

Andrew Jaegle and colleagues

The model leverages an asymmetric attention mechanism to iteratively distill inputs into a tight latent bottleneck.
Explore the quote
AI QUOTE

Eric Jang and colleagues

We present an efficient gradient estimator that replaces the non-differentiable sample from a categorical distribution with a differentiable sample.
Explore the quote
AI QUOTE

Chao Jia and colleagues

A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss.
Explore the quote
AI QUOTE

Albert Q. Jiang and colleagues

We introduce Mixtral 8x7B, a Sparse Mixture of Experts language model.
Explore the quote
AI QUOTE

Michael I. Jordan

Artificial Intelligence is the mantra of the current era.
Explore the quote
AI QUOTE

John Jumper and colleagues

We validated an entirely redesigned version of our neural network-based model, AlphaFold.
Explore the quote
Back to A–Z ↑

K19 quotes

AI QUOTE

Peter Kairouz and colleagues

FL embodies the principles of focused data collection and minimization.
Explore the quote
AI QUOTE

Jared Kaplan and colleagues

Performance has a power-law relationship with each of the three scale factors: model size, dataset size, and computation.
Explore the quote
AI QUOTE

Andrej Karpathy and colleagues

We present a model that generates natural language descriptions of images and their regions.
Explore the quote
AI QUOTE

Vladimir Karpukhin and colleagues

Retrieval can be practically implemented using dense representations alone.
Explore the quote
AI QUOTE

Tero Karras and colleagues

We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature.
Explore the quote
AI QUOTE

Alex Kendall and colleagues

We propose a principled approach to multi-task deep learning which weighs multiple loss functions by considering the homoscedastic uncertainty of each task.
Explore the quote
AI QUOTE

Omar Khattab and colleagues

We propose Demonstrate-Search-Predict, a framework that relies on passing natural language texts in sophisticated pipelines between an LM and an RM.
Explore the quote
AI QUOTE

Douwe Kiela and colleagues

We introduce Dynabench, an open-source platform for dynamic dataset creation and model benchmarking.
Explore the quote
AI QUOTE

Been Kim and colleagues

TCAV uses directional derivatives to compute models’ conceptual sensitivity across entire classes of inputs.
Explore the quote
AI QUOTE

Yoon Kim

We report on a series of experiments with convolutional neural networks trained on top of pre-trained word vectors for sentence-level classification tasks.
Explore the quote
AI QUOTE

Diederik P. Kingma and Jimmy Ba

We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments.
Explore the quote
AI QUOTE

Thomas N. Kipf and colleagues

We present a scalable approach for semi-supervised learning on graph-structured data.
Explore the quote
AI QUOTE

Alexander Kirillov and colleagues

We introduce the Segment Anything project: a new task, model, and dataset for image segmentation.
Explore the quote
AI QUOTE

Nikita Kitaev and colleagues

We introduce two techniques to improve the efficiency of Transformers.
Explore the quote
AI QUOTE

Günter Klambauer and colleagues

We introduce self-normalizing neural networks to enable high-level abstract representations.
Explore the quote
AI QUOTE

Pang Wei Koh and colleagues

We present WILDS, a curated benchmark of 10 datasets reflecting a diverse range of distribution shifts that naturally arise in real-world applications.
Explore the quote
AI QUOTE

Takeshi Kojima and colleagues

We show that LLMs are decent zero-shot reasoners by simply adding ‘Let’s think step by step’ before each answer.
Explore the quote
AI QUOTE

Alex Krizhevsky and colleagues

We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet contest.
Explore the quote
Back to A–Z ↑

L33 quotes

AI QUOTE

Balaji Lakshminarayanan and colleagues

We propose an alternative to Bayesian neural networks that is simple to implement and readily parallelizable.
Explore the quote
AI QUOTE

Remi Lam and colleagues

We introduce a machine learning-based method called GraphCast, which can be trained directly from reanalysis data.
Explore the quote
AI QUOTE

Zhenzhong Lan and colleagues

We present two parameter-reduction techniques to lower memory consumption and increase the training speed of BERT.
Explore the quote
AI QUOTE

Quoc V. Le and colleagues

We consider the problem of building high-level, class-specific feature detectors from only unlabeled data.
Explore the quote
AI QUOTE

Yann LeCun

Our intelligence is what makes us human, and AI is an extension of that quality.
Explore the quote
AI QUOTE

Jan Leike and colleagues

How do we create agents that behave in accordance with the user’s intentions?
Explore the quote
AI QUOTE

Brian Lester and colleagues

We explore prompt tuning, a simple yet effective mechanism for learning soft prompts to condition frozen language models.
Explore the quote
AI QUOTE

Sergey Levine and colleagues

We develop a method that can be used to learn policies that map raw image observations directly to torques at the robot’s motors.
Explore the quote
AI QUOTE

Mike Lewis and colleagues

We present BART, a denoising autoencoder for pretraining sequence-to-sequence models.
Explore the quote
AI QUOTE

Patrick Lewis and colleagues

Providing provenance for their decisions and updating their world knowledge remain open research problems.
Explore the quote
AI QUOTE

Aitor Lewkowycz and colleagues

To help close this gap, we introduce Minerva, a large language model pretrained on general natural language data and further trained on technical content.
Explore the quote
AI QUOTE

Guohao Li and colleagues

To address the challenges of achieving autonomous cooperation, we propose a novel communicative agent framework named role-playing.
Explore the quote
AI QUOTE

Junnan Li and colleagues

We propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks.
Explore the quote
AI QUOTE

Xiang Lisa Li and colleagues

Fine-tuning is the de facto way to leverage large pretrained language models to perform downstream tasks.
Explore the quote
AI QUOTE

Yujia Li and colleagues

To address this gap, we introduce AlphaCode, a system for code generation that can create novel solutions to these problems that require deeper reasoning.
Explore the quote
AI QUOTE

Percy Liang and colleagues

We present Holistic Evaluation of Language Models to improve the transparency of language models.
Explore the quote
AI QUOTE

J.C.R. Licklider

Men will set the goals, formulate the hypotheses, determine the criteria, and perform the evaluations.
Explore the quote
AI QUOTE

Timothy Lillicrap and colleagues

We adapt the ideas underlying the success of Deep Q-Learning to the continuous action domain.
Explore the quote
AI QUOTE

Stephanie Lin and colleagues

We propose a benchmark to measure whether a language model is truthful in generating answers to questions.
Explore the quote
AI QUOTE

Tsung-Yi Lin and colleagues

Our novel Focal Loss focuses training on a sparse set of hard examples.
Explore the quote
AI QUOTE

Aixin Liu and the DeepSeek-AI team

We present DeepSeek-V2, a strong Mixture-of-Experts language model characterized by economical training and efficient inference.
Explore the quote
AI QUOTE

Hanxiao Liu and colleagues

This paper addresses the scalability challenge of architecture search by formulating the task in a differentiable manner.
Explore the quote
AI QUOTE

Haotian Liu and colleagues

We present the first attempt to use language-only GPT-4 to generate multimodal language-image instruction-following data.
Explore the quote
AI QUOTE

Yinhan Liu and colleagues

We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it.
Explore the quote
AI QUOTE

Ze Liu and colleagues

This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision.
Explore the quote
AI QUOTE

Jonathan Long and colleagues

Our key insight is to build fully convolutional networks that take input of arbitrary size and produce correspondingly-sized output.
Explore the quote
AI QUOTE

Ilya Loshchilov and Frank Hutter

We propose a simple modification to recover the original formulation of weight decay regularization.
Explore the quote
AI QUOTE

Ada Lovelace

The Analytical Engine has no pretensions whatever to originate anything.
Explore the quote
AI QUOTE

David G. Lowe

This paper presents a method for extracting distinctive invariant features from images.
Explore the quote
AI QUOTE

Scott M. Lundberg and colleagues

We present a unified framework for interpreting predictions, SHAP (SHapley Additive exPlanations).
Explore the quote
Back to A–Z ↑

M16 quotes

AI QUOTE

Marlos C. Machado and colleagues

We show how diverse the evaluation methodologies in the ALE have become with time, and highlight some key concerns when evaluating agents in the ALE.
Explore the quote
AI QUOTE

Aman Madaan and colleagues

Like humans, large language models do not always generate the best output on their first try.
Explore the quote
AI QUOTE

Aleksander Madry and colleagues

To address this problem, we study the adversarial robustness of neural networks through the lens of robust optimization.
Explore the quote
AI QUOTE

Gary Marcus

Deep learning must be supplemented by other techniques if we are to reach artificial general intelligence.
Explore the quote
AI QUOTE

Maja J. Matarić

Socially assistive robotics focuses on assisting users through social rather than physical interaction.
Explore the quote
AI QUOTE

John McCarthy

It is the science and engineering of making intelligent machines, especially intelligent computer programs.
Explore the quote
AI QUOTE

Leland McInnes and colleagues

UMAP is a novel manifold learning technique for dimension reduction.
Explore the quote
AI QUOTE

H. Brendan McMahan and colleagues

We term this decentralized approach Federated Learning.
Explore the quote
AI QUOTE

Amil Merchant and colleagues

We increase the number of known stable crystals by almost an order of magnitude.
Explore the quote
AI QUOTE

Thomas Mesnard and the Gemma Team

This work introduces Gemma, a family of lightweight, state-of-the-art open models built from the research and technology used to create Gemini models.
Explore the quote
AI QUOTE

Tomas Mikolov and colleagues

We propose two novel model architectures for computing continuous vector representations of words from very large data sets.
Explore the quote
AI QUOTE

Ben Mildenhall and colleagues

Because volume rendering is naturally differentiable, the only input required to optimize our representation is a set of images with known camera poses.
Explore the quote
AI QUOTE

Margaret Mitchell

Part of the issue is that the existential risk concern has drowned out a lot of the other types of concerns.
Explore the quote
AI QUOTE

Melanie Mitchell

An essential first step toward teaching machines ethical concepts is to enable machines to grasp humanlike concepts in the first place.
Explore the quote
AI QUOTE

Volodymyr Mnih and colleagues

The agent is not given any information about the rules or purpose of the game.
Explore the quote
AI QUOTE

Igor Mordatch and Pieter Abbeel

In this paper we investigate if, and how, grounded compositional language can emerge as a means to achieve goals in multi-agent populations.
Explore the quote
Back to A–Z ↑

N10 quotes

AI QUOTE

Moin Nadeem and colleagues

We present StereoSet, a large-scale natural dataset in English to measure stereotypical biases in four domains: gender, profession, race, and religion.
Explore the quote
AI QUOTE

Neel Nanda and colleagues

We argue that progress measures can be found via mechanistic interpretability: reverse-engineering learned behaviors into their individual components.
Explore the quote
AI QUOTE

Arvind Narayanan

The fact is that it is hard to predict the future using AI, and its ability to do that is not improving.
Explore the quote
AI QUOTE

Nils J. Nilsson

Intelligence is that quality that enables an entity to function appropriately and with foresight in its environment.
Explore the quote
AI QUOTE

Helen Nissenbaum

Contextual integrity demands that information gathering and dissemination be appropriate to that context and obey its governing norms.
Explore the quote
AI QUOTE

Safiya Umoja Noble

The greater challenge before us will not be access to the internet, but freedom from machine-based decision making and control over our lives.
Explore the quote
AI QUOTE

Mehdi Noroozi and colleagues

In this paper we study the problem of image representation learning without human annotation.
Explore the quote
AI QUOTE

Mohammad Norouzi and colleagues

This paper presents a simple and computationally efficient approach to incorporate task reward into a maximum likelihood framework.
Explore the quote
Back to A–Z ↑

O7 quotes

AI QUOTE

Barack Obama

You’re harnessing the power of artificial intelligence—big data, robotics, automation—for the good of all of us.
Explore the quote
AI QUOTE

Ziad Obermeyer and colleagues

The algorithm assigned Black patients the same level of risk at lower levels of health.
Explore the quote
AI QUOTE

Catherine Olsson and colleagues

We present six complementary lines of evidence, arguing that induction heads may be the mechanistic source of general in-context learning.
Explore the quote
AI QUOTE

Aaron van den Oord and colleagues

Learning useful representations without supervision remains a key challenge in machine learning.
Explore the quote
AI QUOTE

Ian Osband and colleagues

We propose bootstrapped DQN, a simple algorithm that explores in a computationally and statistically efficient manner through use of randomized value functions.
Explore the quote
AI QUOTE

Long Ouyang and colleagues

Making language models bigger does not inherently make them better at following a user’s intent.
Explore the quote
Back to A–Z ↑

P18 quotes

AI QUOTE

Nicolas Papernot and colleagues

A model may inadvertently and implicitly store some of its training data.
Explore the quote
AI QUOTE

Kishore Papineni and colleagues

We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent.
Explore the quote
AI QUOTE

Alicia Parrish and colleagues

We introduce the Bias Benchmark for QA (BBQ), a dataset of question sets constructed by the authors that highlight attested social biases.
Explore the quote
AI QUOTE

Deepak Pathak and colleagues

In such cases, curiosity can serve as an intrinsic reward signal to enable the agent to explore its environment.
Explore the quote
AI QUOTE

Jaideep Pathak and colleagues

FourCastNet generates a week-long forecast in less than 2 seconds, orders of magnitude faster than IFS.
Explore the quote
AI QUOTE

Shishir G. Patil and colleagues

We release Gorilla, a finetuned LLaMA-based model that surpasses the performance of GPT-4 on writing API calls.
Explore the quote
AI QUOTE

William Peebles and Saining Xie

We explore a new class of diffusion models based on the transformer architecture.
Explore the quote
AI QUOTE

Xue Bin Peng and colleagues

We show that well-known reinforcement learning (RL) methods can be adapted to learn robust control policies.
Explore the quote
AI QUOTE

Jeffrey Pennington and colleagues

We propose a new global log-bilinear regression model that combines the advantages of the two major model families in the literature.
Explore the quote
AI QUOTE

Ethan Perez and colleagues

LM-based red teaming is one promising tool for finding and fixing diverse, undesirable LM behaviors before impacting users.
Explore the quote
AI QUOTE

Bryan Perozzi and colleagues

We present DeepWalk, a novel approach for learning latent representations of vertices in a network.
Explore the quote
AI QUOTE

Matthew E. Peters and colleagues

We introduce a new type of deep contextualized word representation.
Explore the quote
AI QUOTE

Sundar Pichai

AI is too important not to regulate, and too important not to regulate well.
Explore the quote
AI QUOTE

Lerrel Pinto and colleagues

Our experiments clearly show the benefit of using large-scale datasets (and multi-stage training) for the task of grasping.
Explore the quote
AI QUOTE

Matthias Plappert and colleagues

Our policies transfer to the physical robot despite being trained entirely in simulation.
Explore the quote
AI QUOTE

Michael Poli and colleagues

We propose Hyena, a subquadratic drop-in replacement for attention.
Explore the quote
AI QUOTE

Ofir Press and colleagues

We therefore introduce a simpler and more efficient position method, Attention with Linear Biases.
Explore the quote
Back to A–Z ↑

Q1 quote

AI QUOTE

Yujia Qin and colleagues

To bridge this gap, we introduce ToolLLM, a general tool-use framework encompassing data construction, model training, and evaluation.
Explore the quote
Back to A–Z ↑

R27 quotes

AI QUOTE

Alec Radford and colleagues

Language models begin to learn these tasks without any explicit supervision when trained on a new dataset of millions of webpages.
Explore the quote
AI QUOTE

Rafael Rafailov and colleagues

The resulting algorithm, which we call Direct Preference Optimization, is stable, performant, and computationally lightweight.
Explore the quote
AI QUOTE

Colin Raffel and colleagues

We introduce a unified framework that converts all text-based language problems into a text-to-text format.
Explore the quote
AI QUOTE

Gina Raimondo

We need to ensure AI is serving people, not the other way around.
Explore the quote
AI QUOTE

Samyam Rajbhandari and colleagues

We develop a novel solution, Zero Redundancy Optimizer, to optimize memory, vastly improving training speed while increasing the model size.
Explore the quote
AI QUOTE

Inioluwa Deborah Raji and colleagues

We introduce a framework for algorithmic auditing that supports artificial intelligence system development end-to-end.
Explore the quote
AI QUOTE

Pranav Rajpurkar and colleagues

We present the Stanford Question Answering Dataset, a new reading comprehension dataset consisting of 100,000+ questions.
Explore the quote
AI QUOTE

Prajit Ramachandran and colleagues

Using a combination of exhaustive and reinforcement learning-based search, we discover multiple novel activation functions.
Explore the quote
AI QUOTE

Aditya Ramesh and colleagues

We describe a simple approach for this task based on a transformer.
Explore the quote
AI QUOTE

Roshan Rao and colleagues

We introduce the Tasks Assessing Protein Embeddings, a set of five biologically relevant semi-supervised learning tasks.
Explore the quote
AI QUOTE

Ali Razavi and colleagues

We explore the use of Vector Quantized Variational AutoEncoder models for large scale image generation.
Explore the quote
AI QUOTE

Benjamin Recht and colleagues

We view our results as evidence that current accuracy numbers are brittle and susceptible to even minute natural variations in the data distribution.
Explore the quote
AI QUOTE

Joseph Redmon and colleagues

We present YOLO, a new approach to object detection.
Explore the quote
AI QUOTE

Nils Reimers and colleagues

We present Sentence-BERT, a modification of the pretrained BERT network that use siamese and triplet network structures.
Explore the quote
AI QUOTE

Shaoqing Ren and colleagues

An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position.
Explore the quote
AI QUOTE

Danilo Jimenez Rezende and Mohamed Shakir

We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions.
Explore the quote
AI QUOTE

Marco Tulio Ribeiro and colleagues

We propose LIME, a novel explanation technique that explains the predictions of any classifier in an interpretable and faithful manner.
Explore the quote
AI QUOTE

Robin Rombach and colleagues

We apply them in the latent space of powerful pretrained autoencoders.
Explore the quote
AI QUOTE

Olaf Ronneberger and colleagues

The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization.
Explore the quote
AI QUOTE

Frank Rosenblatt

A perceptron is first and foremost a brain model, not an invention for pattern recognition.
Explore the quote
AI QUOTE

Baptiste Rozière and colleagues

We release Code Llama, a family of large language models for code based on Llama 2.
Explore the quote
AI QUOTE

Cynthia Rudin

Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead.
Explore the quote
AI QUOTE

David E. Rumelhart and colleagues

We describe a new learning procedure, back-propagation, for networks of neurone-like units.
Explore the quote
AI QUOTE

Daniela Rus

On an individual level, AI will offer opportunities to make our lives safer, more convenient, and more satisfying.
Explore the quote
AI QUOTE

Olga Russakovsky and colleagues

The ImageNet Large Scale Visual Recognition Challenge is a benchmark in object category classification and detection.
Explore the quote
AI QUOTE

Stuart Russell

The machine’s only objective is the satisfaction of human preferences.
Explore the quote
Back to A–Z ↑

S38 quotes

AI QUOTE

Chitwan Saharia and colleagues

We present Imagen, a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding.
Explore the quote
AI QUOTE

Keisuke Sakaguchi and colleagues

We introduce WinoGrande, a large-scale dataset of 44k problems, inspired by the original WSC design.
Explore the quote
AI QUOTE

Victor Sanh and colleagues

Our smaller, faster and lighter model is cheaper to pre-train.
Explore the quote
AI QUOTE

Teven Le Scao and the BigScience Workshop

As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model.
Explore the quote
AI QUOTE

Timo Schick and colleagues

In this paper, we show that LMs can teach themselves to use external tools via simple APIs and achieve the best of both worlds.
Explore the quote
AI QUOTE

Julian Schrittwieser and colleagues

MuZero learns a model that, when applied iteratively, predicts the quantities most directly relevant to planning.
Explore the quote
AI QUOTE

Florian Schroff and colleagues

We present a system, called FaceNet, that directly learns a mapping from face images to a compact Euclidean space.
Explore the quote
AI QUOTE

John Schulman and colleagues

PPO strikes a balance between ease of implementation, sample complexity, and ease of tuning.
Explore the quote
AI QUOTE

John Searle

Programs are neither constitutive of nor sufficient for minds.
Explore the quote
AI QUOTE

Ramprasaath R. Selvaraju and colleagues

We propose a technique for producing visual explanations for decisions from a large class of CNN-based models, making them more transparent.
Explore the quote
AI QUOTE

Rico Sennrich and colleagues

We introduce a simpler and more effective approach, making the NMT model capable of open-vocabulary translation.
Explore the quote
AI QUOTE

Noam Shazeer

We propose a variant called multi-query attention, where the keys and values are shared across all of the different attention heads.
Explore the quote
AI QUOTE

Jonathan Shen and colleagues

This paper describes Tacotron 2, a neural network architecture for speech synthesis directly from text.
Explore the quote
AI QUOTE

Noah Shinn and colleagues

We propose Reflexion, a novel framework to reinforce language agents through linguistic feedback.
Explore the quote
AI QUOTE

Mohammad Shoeybi and colleagues

We implement a simple, efficient intra-layer model parallel approach that enables training transformer models with billions of parameters.
Explore the quote
AI QUOTE

Reza Shokri and colleagues

We quantitatively investigate how machine learning models leak information about the individual data records on which they were trained.
Explore the quote
AI QUOTE

David Silver and colleagues

AlphaGo achieved a 99% win rate against the strongest other Go programs, and defeated European champion Fan Hui 5–0.
Explore the quote
AI QUOTE

Herbert Simon and Allen Newell

There are now in the world machines that think, that learn and that create.
Explore the quote
AI QUOTE

Karen Simonyan and Andrew Zisserman

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting.
Explore the quote
AI QUOTE

Karan Singhal and colleagues

The resulting model, Med-PaLM, performs encouragingly, but remains inferior to clinicians.
Explore the quote
AI QUOTE

Jake Snell and colleagues

We propose prototypical networks for the problem of few-shot classification.
Explore the quote
AI QUOTE

Jasper Snoek and colleagues

Machine learning algorithms frequently require careful tuning of model hyperparameters, regularization terms, and optimization parameters.
Explore the quote
AI QUOTE

Jascha Sohl-Dickstein and colleagues

The essential idea is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process.
Explore the quote
AI QUOTE

Kihyuk Sohn and colleagues

We demonstrate the power of a simple combination of two common SSL methods: consistency regularization and pseudo-labeling.
Explore the quote
AI QUOTE

Jiaming Song and colleagues

To accelerate sampling, we present denoising diffusion implicit models.
Explore the quote
AI QUOTE

Aarohi Srivastava and colleagues

To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench).
Explore the quote
AI QUOTE

Nitish Srivastava and colleagues

The key idea is to randomly drop units (along with their connections) from the neural network during training.
Explore the quote
AI QUOTE

Nisan Stiennon and colleagues

In this work, we show that it is possible to significantly improve summary quality by training a model to optimize for human preferences.
Explore the quote
AI QUOTE

Emma Strubell and colleagues

These accuracy improvements depend on the availability of exceptionally large computational resources that necessitate similarly substantial energy consumption.
Explore the quote
AI QUOTE

Sainbayar Sukhbaatar and colleagues

We introduce a neural network with a recurrent attention model over a possibly large external memory.
Explore the quote
AI QUOTE

Yutao Sun and colleagues

We propose Retentive Network as a foundation architecture for large language models.
Explore the quote
AI QUOTE

Rishi Sunak

AI will bring new knowledge, new opportunities for economic growth, new advances in human capability.
Explore the quote
AI QUOTE

Mukund Sundararajan and colleagues

We identify two fundamental axioms—Sensitivity and Implementation Invariance—that attribution methods ought to satisfy.
Explore the quote
AI QUOTE

Ilya Sutskever and colleagues

We present a general end-to-end approach to sequence learning that makes minimal assumptions on the sequence structure.
Explore the quote
AI QUOTE

Richard S. Sutton

The biggest lesson from 70 years of AI research is that general methods leveraging computation are ultimately the most effective.
Explore the quote
AI QUOTE

Latanya Sweeney

This writing investigates the delivery of these kinds of ads by Google AdSense using a sample of racially associated names.
Explore the quote
AI QUOTE

Christian Szegedy and colleagues

Here we explore ways to scale up networks in ways that aim at utilizing the added computation as efficiently as possible.
Explore the quote
Back to A–Z ↑

T7 quotes

AI QUOTE

Mingxing Tan and colleagues

Carefully balancing network depth, width, and resolution can lead to better performance.
Explore the quote
AI QUOTE

Max Tegmark

The real problem is that we lack a convincing plan for AI safety.
Explore the quote
AI QUOTE

Josh Tobin and colleagues

With enough variability in the simulator, the real world may appear to the model as just another variation.
Explore the quote
AI QUOTE

Ilya Tolstikhin and colleagues

We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons.
Explore the quote
AI QUOTE

Hugo Touvron and colleagues

We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters.
Explore the quote
AI QUOTE

Alan Turing

We can only see a short distance ahead, but we can see plenty there that needs to be done.
Explore the quote
AI QUOTE

Alan Turing

We can only see a short distance ahead, but we can see plenty there that needs to be done.
Explore the quote
Back to A–Z ↑

V5 quotes

AI QUOTE

Shannon Vallor

It is not AI systems that pose the greatest risk to our humanity right now.
Explore the quote
AI QUOTE

Petar Veličković and colleagues

We present graph attention networks, novel neural network architectures that operate on graph-structured data.
Explore the quote
AI QUOTE

Oriol Vinyals and colleagues

We present a general end-to-end approach to sequence learning that makes minimal assumptions on the sequence structure.
Explore the quote
Back to A–Z ↑

W16 quotes

AI QUOTE

Sandra Wachter and colleagues

Explanations can, in principle, be offered without opening the black box.
Explore the quote
AI QUOTE

Guanzhi Wang and colleagues

We introduce Voyager, the first LLM-powered embodied lifelong learning agent in Minecraft.
Explore the quote
AI QUOTE

Hongyu Wang and colleagues

We introduce BitNet, a scalable and stable 1-bit Transformer architecture designed for large language models.
Explore the quote
AI QUOTE

Xuezhi Wang and colleagues

We propose a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting.
Explore the quote
AI QUOTE

Yizhong Wang and colleagues

We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off their own generations.
Explore the quote
AI QUOTE

Yuxuan Wang and colleagues

We present Tacotron, an end-to-end generative text-to-speech model that synthesizes speech directly from characters.
Explore the quote
AI QUOTE

Christopher Watkins and Peter Dayan

Q-learning is a simple way for agents to learn how to act optimally in controlled Markovian domains.
Explore the quote
AI QUOTE

Jason Wei and colleagues

We show that instruction tuning — finetuning language models on a collection of tasks described via instructions — substantially improves zero-shot performance on unseen tasks.
Explore the quote
AI QUOTE

Joseph Weizenbaum

No other organism, and certainly no computer, can be made to confront genuine human problems in human terms.
Explore the quote
AI QUOTE

Johannes Welbl and colleagues

We propose a novel task to encourage the development of models for text understanding across multiple documents.
Explore the quote
AI QUOTE

Ronald J. Williams

REINFORCE algorithms are shown to make weight adjustments in a direction that lies along the gradient of expected reinforcement.
Explore the quote
AI QUOTE

Chenfei Wu and colleagues

We build a system called Visual ChatGPT, incorporating different Visual Foundation Models.
Explore the quote
AI QUOTE

Qingyun Wu and colleagues

AutoGen is an open-source framework that allows developers to build LLM applications via multiple agents that can converse with each other to accomplish tasks.
Explore the quote
AI QUOTE

Yonghui Wu and colleagues

We present GNMT, Google’s Neural Machine Translation system, which attempts to address many of these issues.
Explore the quote
AI QUOTE

Zhirong Wu and colleagues

We formulate this intuition as a non-parametric classification problem at the instance-level.
Explore the quote
Back to A–Z ↑

X4 quotes

AI QUOTE

Guangxuan Xiao and colleagues

We propose SmoothQuant, a training-free, accuracy-preserving, and general-purpose post-training quantization solution.
Explore the quote
AI QUOTE

Qizhe Xie and colleagues

We present Noisy Student Training, a semi-supervised learning approach that works well even when labeled data is abundant.
Explore the quote
AI QUOTE

Kelvin Xu and colleagues

We introduce an attention based model that automatically learns to describe the content of images.
Explore the quote
AI QUOTE

Keyulu Xu and colleagues

Here, we present a theoretical framework for analyzing the expressive power of GNNs to capture different graph structures.
Explore the quote
Back to A–Z ↑

Y4 quotes

AI QUOTE

Shunyu Yao and colleagues

We explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner.
Explore the quote
AI QUOTE

Denis Yarats and colleagues

We present DrQ-v2, a model-free reinforcement learning algorithm for visual continuous control.
Explore the quote
AI QUOTE

Chengxuan Ying and colleagues

Our key insight is the necessity of effectively encoding the structural information of a graph into the model.
Explore the quote
AI QUOTE

Jiahui Yu and colleagues

We present the Pathways Autoregressive Text-to-Image model, which generates high-fidelity photorealistic images and supports content-rich synthesis.
Explore the quote
Back to A–Z ↑

Z12 quotes

AI QUOTE

Lotfi A. Zadeh

A fuzzy set is a class of objects with a continuum of grades of membership.
Explore the quote
AI QUOTE

Sergey Zagoruyko and colleagues

We call the resulting network structures wide residual networks.
Explore the quote
AI QUOTE

Manzil Zaheer and colleagues

We propose BigBird, a sparse attention mechanism that reduces this quadratic dependency to linear.
Explore the quote
AI QUOTE

Rowan Zellers and colleagues

Commonsense inference still proves difficult for even state-of-the-art models.
Explore the quote
AI QUOTE

Biao Zhang and Rico Sennrich

RMSNorm is computationally simpler and thus more efficient than LayerNorm.
Explore the quote
AI QUOTE

Richard Zhang and colleagues

We propose a fully automatic approach that produces vibrant and realistic colorizations.
Explore the quote
AI QUOTE

Susan Zhang and colleagues

We present Open Pre-trained Transformers, a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters.
Explore the quote
AI QUOTE

Tianyi Zhang and colleagues

We propose BERTScore, an automatic evaluation metric for text generation.
Explore the quote
AI QUOTE

Lianmin Zheng and colleagues

To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions.
Explore the quote
AI QUOTE

Chunting Zhou and colleagues

Taken together, these results strongly suggest that almost all knowledge in large language models is learned during pretraining.
Explore the quote
AI QUOTE

Denny Zhou and colleagues

Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks.
Explore the quote
AI QUOTE

Deyao Zhu and colleagues

We present MiniGPT-4, which aligns a frozen visual encoder with a frozen advanced language model using one projection layer.
Explore the quote
Back to A–Z ↑
KEEP EXPLORING

Explore more quotes

Browse the established NewsStreets quote collections by topic.

THIS DAY IN AI

November 17
2016: UNREAL Speeds Learning
Read More

Explore All Days in AI

QUOTE OF THE DAY

Yao Fu and colleagues, 2022
“We propose complexity-based prompting, a simple and effective example selection scheme for multi-step reasoning.”
Read More

Explore All Quotes in AI

EXPLAINER OF THE DAY

What Is a Saliency Map in AI?
A saliency map highlights input regions whose small changes have a strong local effect on a model’s selected output.
Read More

Explore All Explainers in AI