Artificial Intelligence (AI) Webinar

Real-World Adoption Of AI And Machine Learning

By Meredith Holmes

Bob O'Donnell of TECHnalysis Research, gave a presentation via webinar entitled, "Artificial Intelligence and Machine Learning: AI in the Enterprise: A Look at Real-World Usage." O'Donnell covered highlights of a study he conducted to determine the nature and extent of AI usage among U.S.-based businesses. Data were gathered from an online survey of 504 medium-sized and large companies that are building or using AI applications. The initial sample was 3,700 companies in all types of businesses.

A top expert in technology market research, Bob O'Donnell is president and chief analyst of TECHnalysis, a technology market research and consulting firm he founded.  He writes a biweekly column for the Tech section of USAToday and appears regularly on TV. O'Donnell's book, Personal Computer Secrets, helps both nerds and novices get the most out of their PC. For fun, he plays trombone and writes songs for his band, "The Headliners."

AI in the Enterprise

O'Donnell began by saying, "The goal of this study is to understand not what's theoretically possible, but what businesses are actually doing." Of the 3,700 companies in the initial sample, about 1 in 5 is currently using AI. Of that group more than half (50 percent) report being in full deployment. More large businesses (61 percent) than medium-sized businesses (50 percent) are in full AI deployment. In addition, 72 percent of companies that considered themselves "early adopters" are in full AI deployment. Neither of these findings is surprising, O'Donnell pointed out, since AI is an advanced technology that sophisticated companies with fairly deep pockets are more likely to adopt.

Fifty percent of companies not using AI cited cost as the reason. Other reasons given included  "Lack of in-house expertise" (30 percent), "Don't know enough" (28 percent), and "Don't see the value" (10 percent). Three responses O'Donnell considered significant were "Negative impact on personnel" (23 percent) and "Negative impact on company operation" (18 percent), and "Negative impact on society." He called these "AI fear issues" and said, "Those of you involved in this industry need to know these fears and concerns are real."

Applications

Most companies are using AI for security – especially data security (71 percent), rather than for more futuristic applications, such as robotics and computer vision. Security applications include network security (69 percent), business intelligence (66 percent), and device security (63 percent). "Applications for things like malware detection and network filtering are very straightforward, 'back-office' applications that are critical to the operation of the organization, and deliver real value." said O'Donnell.

Industries Using AI

Tech companies (27 percent), dominate the businesses using AI. Manufacturing (13 percent) was second, and professional, scientific, and tech services (10 percent), third. No utility, real estate, or agriculture businesses in the survey reported using AI. Use of AI in hospitality, construction, and mining (all 1 percent) was also low. The study revealed a gulf between tech companies, who are in a position to understand and finance AI, and most other industries.

Inferencing and Training

Seventy-three percent of companies surveyed build their own AI models and train them. That is, they input data and train the model to recognize patterns in the data and to make distinctions between, for example, a person and a parked car. The resulting algorithms then enable inferencing, or leveraging the ability of the model to make these distinctions on demand. Of interest to licensing professionals is the question of who owns the algorithms.

Goals and Challenges

Asked to rank their top five goals for AI, the majority of companies put "improving efficiency" first; "speeding up and automating tasks," second; and "increasing security," third. Complexity and cost were cited as the top challenges in implementing AI. Another significant obstacle is "uncertainty about AI's impact." However, despite technical challenges, high implementation costs, and uncertainty about AI's impact, O'Donnell concluded, there is great excitement about AI and its applicability in an almost unlimited number of business operations.

Questions and Answers

O'Donnell answered some questions from live audience members:

Q. How are companies handling licensing issues? Who owns the AI model if you're an AI supplier, and what about use of the developed model if you're a service provider?

A. A lot of the early work on AI was done by academics who tend to share results, and a lot is still open-source. But some companies are developing AI models, and AI "stores"—where algorithms are available and on display. I think we'll see a combination of approaches to the IP question. There will be value in developing algorithms that have the "secret sauce"—ability to do specific operations—on multiple databases.

Q. Your study covered AI in the U.S., but what about the rest of the world? Especially China?

A. There is tremendous technical expertise in China, and AI-related research and implementation are progressing rapidly. They also have access to a huge amount of data, like face-tracking data from security cameras. The U.S. has privacy concerns that limit access to this kind of data. There is concern that AI could be politicized, due to different views about use of and access to data. I am concerned that the technology could get regionalized—that is, that worldwide, there will be inconsistent standards and operating systems—that a separate AI subculture could develop in China.

Slides and audio of Bob O'Donnell's January 16 webinar presentation are available on the LES website.

Go to  https://www.lesusacanada.org/page/WebinarRecordings

For information about an expanded 178-slide version of the presentation, contact Bob O'Donnell:

Email—bob@technalysisresearch.com
Twitter—@bobodtech
LinkedIn—Bob O'Donnnell
on the web—www.technalysisresearch.com