Connect with us


MIT has discovered bugs in datasets used for AI testing



A team of researchers led by experts from the Massachusetts Institute of Technology (MIT) examined ten datasets most commonly used to test machine learning systems. The scientists found that about 3.4% of the data was inaccurate or mislabeled. This could cause problems in artificial intelligence systems that use these datasets.

The datasets, each linked to over 100,000 works, include text content, images, and videos from newsgroups, the Amazon store, YouTube, and the IMDb movie database. Errors include negative product reviews, mistakenly marked as positive, incorrect descriptions of what is shown in the illustrations, inaccurate descriptions of the content of sound recordings.

It is significant that the researchers also used machine learning methods and related software tools to detect possible errors.

It remains to add that some errors can be considered insignificant, and sometimes we should rather talk about the ambiguity of the input data. However, in one of the datasets – the QuickDraw test – there are errors in about 10% of the dataset. What AI can learn from such data is anyone’s guess.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Latest News

News8 mins ago

Sony Xperia 1 III was shown right before the announcement. Official images of the flagship released

Sony will be unveiling the new Xperia phones at a special event this week. Right before the announcement, AndroidNext published...

Gaming19 mins ago

Waiting for the next generation of Nintendo Switch? Together, Nintendo introduced the Nintendo Switch Lite in a new, darker blue color.

While many Nintendo fans have been waiting for a new version of the Switch game console with a better display...

News52 mins ago

US blamed for global chip shortage and rising electronics prices

Huawei Technologies blamed the U.S. for a chip shortage that shook the global industry, saying Washington’s sanctions against Chinese companies...

Tablets1 hour ago

2560 x 1600 resolution, 120 Hz frame rate, 4096-touch stylus support. All the details about the screen of the tablet Xiaomi Mi Pad 5

One of the users of the social network Weibo found interesting documents in the TCL database: they mention a 10.95-inch...

News1 hour ago

Xiaomi Mi Pad 5 will receive a keyboard and support for VR content

The Chinese company Xiaomi is currently preparing to launch tablets on the market under the names Xiaomi Mi Pad 5...

Components1 hour ago

AMD Finally Unveils Ryzen 5000G Desktop APUs, But It’s Not Easy To Buy Again

AMD has finally unveiled the Ryzen 5000G Desktop APUs. As expected, there are three models in the lineup. Unfortunately, as...

Phones2 hours ago

This is what the iPhone 13 looks like with a new camera and reduced bangs. Accurate 3D renders published

Resource MySmartPrice published images of a three-dimensional model of the iPhone 13 smartphone, which was created on the basis of...

News2 hours ago

Xiaomi intrigues with a huge new TV that is “bigger than life”

The Indian office of Xiaomi has published a teaser for a new product in the TV category. In the teaser,...

Most Popular