To improve machine learning, the answer might be taking a different approach to robot education rather than just feeding models more data
The key to developing flexible machine-learning models that are capable of reasoning like people do may not be feeding them oodles of training data. Instead, a new study suggests, it might come down to how they are trained. These findings could be a big step toward better, less error-prone artificial intelligence models and could help illuminate the secrets of how AI systems—and humans—learn.
But a novel training protocol that is focused on shaping how neural networks learn can boost an AI model’s ability to interpret information the way humans do, according to a study published on Wednesday in Nature. The findings suggest that a certain approach to AI education might create compositional machine learning models that can generalize just as well as people—at least in some instances.
The language consisted of nonsense words that “translated” into sets of colorful dots. Some of these invented words were symbolic terms that directly represented dots of a certain color, while others signified functions that changed the order or number of dot outputs. For instance, dax represented a simple red dot, but fep was a function that, when paired with dax or any other symbolic word, multiplied its corresponding dot output by three. So “dax fep” would translate into three red dots.
The model’s performance is particularly remarkable, given its small size. “This is not a large language model trained on the whole Internet; this is a relatively small transformer trained for these tasks,” says Armando Solar-Lezama, a computer scientist at the Massachusetts Institute of Technology, who was not involved in the new study. “It was interesting to see that nevertheless it’s able to exhibit these kinds of generalizations.
日本 最新ニュース, 日本 見出し
Similar News:他のニュース ソースから収集した、これに似たニュース記事を読むこともできます。
Stable and Efficient: Scientists Develop New Method To Create Next-Gen Solar CellsScience, Space and Technology News 2023
続きを読む »
New method to analyze implications of large-scale flood adaptationIn a test of their new analysis tool, researchers show where 'moving up' or 'moving over' may make the most sense for those affected by the 2022 Pakistan flood, and what costs it would entail.
続きを読む »
New method produces homogeneous polystyrene microparticles in a stable dispersionDispersions of polymer particles in a liquid phase (latexes) have many important applications in coatings technology, medical imaging, and cell biology.
続きを読む »
Scientists develop new method to create stable, efficient next-gen solar cellsNext-generation solar materials are cheaper and more sustainable to produce than traditional silicon solar cells, but hurdles remain in making the devices durable enough to withstand real-world conditions.
続きを読む »
New method proposed to improve the ocean observational network in the tropical western PacificENSO, short for the El Niño and Southern Oscillation, is the most influential interannual oscillation, and significantly impacts global climate. The Tropical Pacific Observation System (TPOS), including moored buoys, plays an important role in understanding, monitoring, and forecasting ENSO events.
続きを読む »
- New Brunswick, New Jersey job with Rutgers Cancer Institute of New JerseyPostdoctoral positions at Rutgers Cancer Institute of New Jersey, Rutgers University-State University of New Jersey (New Brunswick campus) are available immediately.
続きを読む »