Computer Vision

The ability of machines to extract information from visual data.

Lux logo LUX TAKE:

Computer vision technology has become reasonably mature and has reached a tipping point; the technology has moved out of research labs and is being used to successfully address real-world problems in multiple industry sectors. In the next five years, we anticipate rapid adoption of this technology in a plethora of different applications.

What's New

Lux Research analysts and the Lux Intelligence Engine have added the following recent computer vision developments.

New Lux Content

A Digital Transformation Framework: Applying Digital Tools to Improve Business Operations: One of the key elements to a successful digital transformation is laying out a clear vision. This involves designing new digitally-enhanced products and customer-centric digital business models to go with them as well as identifying business processes that are suitable for digitalization. In a previous report, we presented a framework that B2B and B2C companies across different sectors could use to create customer-centric digital business models. In this report, we present another framework that companies can use to identify suitable opportunities for digitalization in internal processes as well as the right digital tools to enable that transformation. We use case studies from various sectors to highlight how to use this framework.

Different approaches to AI: Symbolic reasoning and machine learning: In a previous insight, we developed a working definition of artificial intelligence (AI): AI is the ability of machines to perform complex tasks that have historically required human or animal thinking. In this insight, we provide a high-level overview of different approaches to AI, as shown in Figure 1.​Figure 1: The two main branches of AI: symbolic reasoning and machine learning (source: Lux Research)Some of the first attempts at AI were driven by methods classified as symbolic reasoning, a top-down or hard-coded approach. Such methods attempt to codify expert knowledge into logic-based rules to emulate the way a human solves a problem. This top-down approach dominated the first 40 years of AI research. For example, symbolic reasoning was used to program Deep Blue, which was the first computer to win a game of chess against a reigning world champion. Deep Blue operated by searching through a library of moves and positions from expert chess games and won by using an evaluation algorithm that predicted the likelihood of a move's success. While symbolic reasoning worked for well-constrained environments like playing chess, it broke down under real-world situations like language translation because it lacked the complete information needed to capture all of the nuances of human thought. Under this approach, AI was limited by a human's ability to map out the rules to solve complex tasks.In the past 20 years, advancements in technology and computing (such as new algorithms, the proliferation of sensors, the emergence of powerful processors, cloud computing, and the internet) have given rise to a data-centric approach to AI, commonly referred to as machine learning. Rather than encoding hard logic rules shared by a human expert, machine learning systems use a bottom-up approach whereby they analyze empirical data to develop the ability to "think." A variety of machine learning techniques (such as artificial neural networks and deep learning) have been developed in recent years, but the fundamental premise behind all these algorithms remains the same: Data is used to determine the parameters of a complex function, which can then be used to make new predictions based on input data. In fact, it is this new class of machine learning techniques that has led to the proliferation of AI in areas as diverse as speech recognition, natural language processing, and computer vision. Machine learning-based AI offers the ability to digest vast amounts of multidimensional data, enabling applications like materials informatics, something for which humans would not be able to derive logical rules. Clients should note that almost all AI developments in recent years have been developments in the more modern approach of machine learning, as opposed to symbolic reasoning. While machine learning models have proven to be extremely effective in numerous applications, they suffer from several challenges, such as the lack of explainability, algorithmic bias, and the need for large datasets to train them. In one of our upcoming insights, we will dive deeper into some of the emerging machine learning algorithms that try and address these issues.

About Lux Research

Lux Research is a leading provider of tech-enabled research and advisory solutions, helping clients drive growth through technology innovation. A pioneer in the research industry, Lux uniquely combines technical expertise and business insights with a proprietary intelligence platform, using advanced analytics and data science to surface true leading indicators. With quality data derived from primary research, fact-based analysis, and opinions that challenge traditional thinking, Lux clients are empowered to make more informed decisions today to ensure future success.

Already a Lux Client?

Sign In

Interested in Learning More?

Contact us to learn the benefits of becoming a Lux member.

Or call us now

For North America (Boston Headquarters)
+1 (617) 502-5300

For EMEA (Amsterdam)
+31 (0) 20-3460840

For APAC (Singapore)
+65 6592-6978