Here are five important AI stories from the week.
Traditionally, problem-solving required humans to design hypotheses and test them one-by-one in a painstakingly long and arduous process. Now artificial intelligence is able to problem-solve at scale leveraging large data sets, cloud-based infrastructure, and the latest machine learning algorithms. In other words, we are moving to a new age of intelligence, rooted in fast-paced radical empiricism led by AI.
Bias in data used to train AI systems is a major problem; for example, machine learning models to screen resumes may perpetuate a gender skew in favor of male applicants. While AI systems may be relatively new, identifying and addressing bias in data is not a new problem. Rather, bias is an age-old problem, one that humans are well-equipped to address, provided they pay sufficient attention to it instead of feeding data blindly into a machine learning algorithm.
With generative models, AI is now able to learn the statistical properties of music and generate new samples based on the underlying distribution it has learned. Such models wow but also have limits, which this article explores.
AI could empower and augment humans in work that they perform, or AI could replace humans. Most likely, AI will do both. While AI that replaces humans wows audiences, it is harder to build and potentially less impactful than AI that augments humans. This article addresses the differences between complementary AI and substitutive AI.
No longer just a theoretical discipline, AI is now being applied in the real world. This is the age of applied AI, or AI 2.0. Finding use cases for AI - given AI’s unique strengths and weaknesses - and building effective AI solutions are a challenge. In this article, Mike Loukides breaks down the keynote speeches that explore this topic at the recent O’Reilly AI Conference In New York City.
More Stories Worth Reading…