By PYMNTS – A new breed of artificial intelligence is emerging, capable of tasks it was never explicitly taught. This technology, known as zero-shot learning, is pushing the boundaries of AI capabilities and raising questions about the future of machine intelligence.

OpenAI’s GPT-4, for example, is a language model that showcases zero-shot learning abilities. Without any specific legal training, GPT-4 scored in the 90th percentile on the Uniform Bar Exam. The model also exhibited zero-shot translation capabilities, accurately translating between language pairs it had never seen before, such as Slovenian to Swahili.

Zero-shot learning allows AI systems to perform tasks or recognize objects without prior specific training. Traditional machine learning models require extensive datasets for each new task, but zero-shot learning algorithms apply existing knowledge to novel situations, mimicking human-like inference.

The roots of zero-shot learning trace back to a 2009 paper by Carnegie Mellon University researchers titled “Zero-shot Learning with Semantic Output Codes.” Since then, the field has advanced rapidly, with breakthroughs occurring in the past five years.

The impact of zero-shot learning extends beyond language processing. Researchers published a study in Nature Biomedical Engineering demonstrating CheXzero, an AI model that detects various diseases from chest X-rays using zero-shot learning. The model successfully identified conditions on which it had never been trained, such as pneumothorax (collapsed lung), by leveraging its understanding of related medical concepts and image features.

Google DeepMind unveiled Gato, a generalist AI agent that showcases zero-shot learning across diverse domains. In one demonstration, Gato played a new Atari game it had never seen before, applying strategies learned from other games to achieve a high score within minutes of exposure to the new game.  Read On: