Explore 366 quotations from researchers, innovators, writers, and pioneers who have shaped how we understand artificial intelligence.
QUOTE OF THE DAY IN AI

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 →
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
Rodney Brooks
The world is its own best model.Explore the quote →
AI QUOTE
Meredith Broussard
Neutrality in tech is a myth.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
Joy Buolamwini
Algorithmic bias, like human bias, results in unfairness.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 →
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
Ted Chiang
Think of ChatGPT as a blurry JPEG of all the text on the Web.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
Rumman Chowdhury
It’s simply math, put into code.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 →
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 →
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 →
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 →
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
Zoubin Ghahramani
How can a machine learn from experience?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 →
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
Geoffrey Hinton
These things really do understand what they’re saying.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 →
AI QUOTE
Jensen Huang
Generative AI is the defining technology of our time.Explore the quote →
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 →
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 →
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
Lina Khan
There is no AI exemption to the laws on the books.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 →
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
Jaron Lanier
As a computer scientist, I don’t like the term ‘A.I.’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
Ursula von der Leyen
Algorithms must not create a new black box.Explore the quote →
AI QUOTE
Fei-Fei Li
There’s nothing “artificial” about AI.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 →
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 →
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
Satya Nadella
AI must be designed to assist humanity.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
Andrew Ng
AI is the new electricity.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 →
O7 quotes
AI QUOTE
Cathy O’Neil
Algorithms are opinions embedded in code.Explore the quote →
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 →
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
Judea Pearl
Data are profoundly dumb.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 →
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 →
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
Ginni Rometty
The purpose of AI is to augment human intelligence.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 →
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
Mustafa Suleyman
AI is a new kind of digital species.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 →
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 →
V5 quotes
AI QUOTE
Leslie Valiant
A computational theory of learning is described.Explore the quote →
AI QUOTE
Shannon Vallor
It is not AI systems that pose the greatest risk to our humanity right now.Explore the quote →
AI QUOTE
Ashish Vaswani and colleagues
Attention is all you need.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 →
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
Norbert Wiener
To live effectively is to live with adequate information.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 →
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 →
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 →
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 →
Explore more quotes
Browse the established NewsStreets quote collections by topic.
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 MoreEXPLAINER 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
