Artificial intelligence (AI), machine learning and deep learning are currently some of the hottest topics around. There are good reasons for that. Apps, software, machines and vehicles are getting smarter. We talk to our smartphones. Cars will soon be able to drive themselves. Automatic translation from one language to another is no longer a gimmick. And who would have thought that a computer would ever outdo the world’s best player in the complex board game Go?
The potential of AI
Aspects of data volume and computing power play into the hands of the widespread use of the technology behind AI. Why? Machine learning generates knowledge from experience, in the form of data. These data contain examples of relationships. Machine learning recognizes patterns and then applies these findings to new, previously unknown data to make predictions. As a result, very complex relationships can be calculated.
Machine learning therefore needs as many data and examples as possible. Fortunately, there is plenty of data available. If I want to teach a computer to differentiate pictures of people, cats and cars, I just have to search Google for a few minutes. Companies today have loads of data from machine sensors, production systems or customer transactions. It is often untapped, but still available. Machine learning is also very computationally intensive. With cost-effective storage and a high degree of parallelization of computation, we can now easily apply machine learning to large volumes of data.
This does require some action, however. Most European companies are still in the early stages of this process, as the recent SAS study “The Enterprise AI Promise: Path to Value” shows. There is optimism about the potential of AI, but many of the study respondents are far from convinced that their company has the resources to develop it.
The respondents also felt, however, that technology is not the biggest hurdle. Both relevant software and economically-viable hardware are readily available. Instead, they were concerned about the lack of data science know-how, organizational culture and social acceptance. More than half (55%) of the respondents thought the biggest challenge would be changes in the world of work, the potential job losses and new job profiles. Only 20% of companies surveyed said they had confidence that their data science teams were ready, and 19% did not even have a data science team. Nearly one in two respondents (49 percent) had doubts about the reliability of the results provided by AI (as a black box), suggesting that a managerial rethink is needed. Only every fourth company reporting having a suitable infrastructure for AI (24%).
The news is not all bad
It is not all bad, though. Just under a quarter (23%) of respondents saw AI as offering the opportunity to improve efficiency, productivity and business processes more generally. Almost as many (22%) thought that using this type of technology would benefit customer relationships.
More, AI is already part of everyday business in some places. Machine learning as AI technology is one of the trend themes in the analytics market and is used in predictive maintenance, fraud detection and customer intelligence. Large amounts of data and computing power can be used to make better predictions and more accurate analyses. Plenty of respondents cited self-driving / networked cars (26%) and smart / virtual assistants such as Siri, Alexa or Cortana (24%) as natural areas where AI could quickly make a positive impact.
Sign up to this webinar series and learn about the practicalities of artificial intelligence and making AI business-smart
SAS is right in the middle of it
Machine learning and AI are not new territory for SAS. Cognitive computing combines AI methods, database-based technologies, and human-machine interfaces. The ultimate goal is to give the user the most natural possible application experience with a software system. This is exactly the claim we make for our analytical solutions: that we make them more “human”.
In recent years, visualization has been an important driver of this. A first solution was SAS Visual Analytics, but we have also made advanced analytics available to non-statistics professional users, and we continue to move down this path. An obvious next step is to involve natural language analytics to provide new insights, depending on the situation and the user’s level of knowledge. Dr. Jim Goodnight showed a first example of this with his Alexa demo at the SAS Global Forum. I was so curious that I tried it myself:
Artificial intelligence, as part of computer science, deals with modelling and automating tasks that normally need people and require what we call ‘intelligence’ (e.g., understanding language, recognizing patterns, or applying knowledge gained through experience in new situations). SAS uses AI techniques in text mining, including categorization, subject recognition, or sentiment analysis. Many SAS solutions already have machine learning capabilities, including SAS Enterprise Miner, SAS Visual Analytics, SAS Visual Statistics and SAS Visual Data Mining and Machine Learning.
Versatile use cases
But SAS is doing more than bringing machine learning to data science. The same techniques are also enabling business users to better understand their data intuitively and quickly. Virtually all analytical applications could benefit from better analytics and accessible applications. New applications in IoT and object recognition are bringing large quantities of data, including image data, and real-time stream-based analytics are required. The potential applications are enormous, from predictive maintenance, improving the supermarket experience of the future or production quality, and damage prevention in insurance. No wonder so many sectors and industries are discussing AI.
This blog post was originally posted in German on the regional SAS blog site Mehr Wissen - you can read it here...