DeepLearning.AI

Benchmark Tests Are Meaningless: The problem with training data contamination in machine learning

The universe of web pages includes correct answers to common questions that are used to test large language models. How can we evaluate new models if they’ve studied the answers before we give them the test?

Datasets

Benchmark Tests Are Meaningless: The problem with training data contamination in machine learning

The universe of web pages includes correct answers to common questions that are used to test large language models. How can we evaluate new models if they’ve studied the answers before we give them the test?

Datasets

Balancing Web Data Distributions: Automated method organizes large datasets for better model performance

Datasets that were scraped from the web tend to be unbalanced, meaning examples of some classes (say, cats) are plentiful while examples of others (say, caterpillars) are scarce.

Datasets

Balancing Web Data Distributions: Automated method organizes large datasets for better model performance

Datasets that were scraped from the web tend to be unbalanced, meaning examples of some classes (say, cats) are plentiful while examples of others (say, caterpillars) are scarce.

Datasets

Data Disappears: Creative workers don't want AI developers to train models on their work

The latest advances in AI are built on freely available training data. What will happen if it becomes off-limits? Creative workers don’t want AI developers to train models on their works without permission or compensation, or at all. Data is vanishing as they scramble to lock it down.

Datasets

Data Disappears: Creative workers don't want AI developers to train models on their work

The latest advances in AI are built on freely available training data. What will happen if it becomes off-limits? Creative workers don’t want AI developers to train models on their works without permission or compensation, or at all. Data is vanishing as they scramble to lock it down.

Datasets

More Scraped Data, Greater Bias: Research shows that training on larger datasets can increase social bias.

How can we build large-scale language and vision models that don’t inherit social biases? Conventional wisdom suggests training on larger datasets, but research challenges this assumption.

Datasets

More Scraped Data, Greater Bias: Research shows that training on larger datasets can increase social bias.

How can we build large-scale language and vision models that don’t inherit social biases? Conventional wisdom suggests training on larger datasets, but research challenges this assumption.

Datasets

News Outlet Challenges AI Developers: The New York Times forbids the use of its work in training datasets.

The New York Times launched a multi-pronged attack on the use of its work in training datasets. The company updated its terms of service to forbid use of its web content and other data for training AI systems.

Datasets

News Outlet Challenges AI Developers: The New York Times forbids the use of its work in training datasets.

The New York Times launched a multi-pronged attack on the use of its work in training datasets. The company updated its terms of service to forbid use of its web content and other data for training AI systems.

Datasets

Sample-Efficient Training for Robots: Reinforcement learning from human feedback to train robots

Training an agent that controls a robot arm to perform a task — say, opening a door — that involves a sequence of motions (reach, grasp, turn, pull, release) can take from tens of thousands to millions of examples...

Datasets

Sample-Efficient Training for Robots: Reinforcement learning from human feedback to train robots

Training an agent that controls a robot arm to perform a task — say, opening a door — that involves a sequence of motions (reach, grasp, turn, pull, release) can take from tens of thousands to millions of examples...

Datasets

Stable Biases: Stable Diffusion may amplify biases in its training data.

Stable Diffusion may amplify biases in its training data in ways that promote deeply ingrained social stereotypes.

Datasets

Stable Biases: Stable Diffusion may amplify biases in its training data.

Stable Diffusion may amplify biases in its training data in ways that promote deeply ingrained social stereotypes.

Datasets

Finer Tuning: Surgical fine-tuning modifies layers based on data differences.

Fine-tuning a neural network typically involves retraining every layer on new data. But research shows that networks may perform better when fine-tuning modifies only a subset of layers.

Datasets

Benchmark Tests Are Meaningless: The problem with training data contamination in machine learning

Benchmark Tests Are Meaningless: The problem with training data contamination in machine learning

Balancing Web Data Distributions: Automated method organizes large datasets for better model performance

Balancing Web Data Distributions: Automated method organizes large datasets for better model performance

Data Disappears: Creative workers don't want AI developers to train models on their work

Data Disappears: Creative workers don't want AI developers to train models on their work

More Scraped Data, Greater Bias: Research shows that training on larger datasets can increase social bias.

More Scraped Data, Greater Bias: Research shows that training on larger datasets can increase social bias.

News Outlet Challenges AI Developers: The New York Times forbids the use of its work in training datasets.

News Outlet Challenges AI Developers: The New York Times forbids the use of its work in training datasets.

Sample-Efficient Training for Robots: Reinforcement learning from human feedback to train robots

Sample-Efficient Training for Robots: Reinforcement learning from human feedback to train robots

Stable Biases: Stable Diffusion may amplify biases in its training data.

Stable Biases: Stable Diffusion may amplify biases in its training data.

Finer Tuning: Surgical fine-tuning modifies layers based on data differences.

Subscribe to The Batch