Towards an expanded AI « Machine Learning Times
Originally published in Communications from the ACM, April 2022, Vol. 65 No. 4, pages 56-57
Despite the great successes of artificial intelligence (AI) and deep learning, critical assessments have been made of current deep learning methods.8 Deep learning is data-intensive, has limited knowledge transfer capabilities, does not adapt quickly to changing tasks or distributions, and insufficiently integrates global or prior knowledge.1,3,8,14 While deep learning excels in natural language processing and vision benchmarks, it often underperforms in real-world applications. Deep learning models have been shown to fail with new data, new applications, deployments in the wild, and stress tests.4,5,7,13,15 Therefore, practitioners doubt these models and hesitate to use them in real applications.
Current AI research has attempted to overcome the criticisms and limitations of deep learning. AI research and machine learning in particular are aiming for a new level of AI – “broad AI” – with vastly improved and broader capabilities for skill acquisition and problem solving.3 We contrast “broad AI” with “narrow AI”, which are the currently applied AI systems. Broad AI significantly outperforms narrow AI in the following key properties: knowledge transfer and interaction, adaptability and robustness, abstraction and advanced reasoning, and efficiency (as illustrated in the attached document). figure). A broad AI is a sophisticated and adaptive system, which successfully performs any cognitive task by virtue of its sensory perception, previous experience and learned skills.
To continue reading this article, Click here.