State of AI Report Notes Continued Challenges


The misunderstandings business leaders have about the inner workings of AI and machine learning processes could continue to hamper adoption.

The disconnect between IT and the business world is shrinking year by year. Companies are finally getting the hang of what artificial intelligence (AI) can do and how it can be applied to use cases that drive business value.

A recent report from Appen, “The State of AI and Machine Learning Report 2022,” notes what’s currently going well for companies at this stage of digital transformation and some obstacles that still lie ahead. Here’s what you need to know.

Data Sourcing is still a challenge for most companies

AI and machine learning are data-hungry. Sourcing quality data in real-time in significant enough quantities to power ML and artificial intelligence initiatives remains a challenge for many of the businesses responding to the study. However, there’s a substantial disconnect between the responses from IT leaders and business leaders, with just 24% of the latter identifying this challenge. 42% of IT leaders note the challenge.

This could still point to a lack of clarity around what makes up a successful artificial intelligence implementation. According to IT leaders, model evaluation is nearly as important an obstacle as data sourcing, so business leaders may not have much behind-the-scenes knowledge yet for these processing-heavy tasks. As companies collaborate and communicate through data-literate means, those opinions could nudge closer together.

Data accuracy is critical to the success of any AI project, and that isn’t in question, according to all respondents. However, less than 10% of companies report data accuracy above 90%, so something is going amiss at this critical step. Companies will need to reconcile their data collection, possibly through a data operating system or other holistic solution, to ensure data for their AI projects.

See also: AI’s Achilles Heel: Data Quality

Forget taking jobs—Humans are still vital to the AI lifecycle

Human-in-the-loop machine learning is an extremely popular response, with 81% stating that humans are critical to data collection. A whopping 97% of respondents believe that human-in-the-loop model evaluation is critical for performance.

Companies are recognizing that humans still offer a valuable barrier to model inaccuracy. After some high-profile cases of model bias, fewer companies are willing to take on black box AI, preferring some measure of human oversight to reduce mistakes and ensure that models continue top performance. Since artificial intelligence is a differentiator for business, this isn’t surprising. The fast track to digital transformation involves implementing these use cases, and humans will continue to provide the foundation.

See also: Study Outlines Industry and Public AI Perceptions

Perceptions are changing, and that’s a good thing

More companies are willing to integrate AI into their existing businesses, and very few believe themselves to be behind their peers in these initiatives. Instead, tech leaders seem to agree that they’re at least on par with others in their industry—in the U.S., 42% of respondents believed this.

More respondents in the US believed themselves to be ahead of their peers in their industry, with 54% responding in this manner versus 44% in Europe. Because so much of the United States is a hotbed for startups and innovation labs specifically focused on AI, this could be the reason for the greater confidence.

What’s clear is just how important AI is overall to the global business picture. Most companies believe that it’s a critical part of current and future business operations and are eager to do what they can to ensure they don’t fall behind their peers.

And as always, responsible AI is an important principle

Companies are aware that artificial intelligence poses ethical risks. They are also aware of the potential distrust from consumers for AI projects. An overwhelming majority of respondents believe that responsible AI is the foundation of all AI projects, with 49% strongly agreeing with that statement and 44% somewhat agreeing with that statement.

Data ethics isn’t just about AI accuracy and relieving bias. It’s also a nod to the responsibility companies have to ensure consumers trust the AI initiatives their favorite companies engage in and to ensure everyone’s safety from supply chain to consumer.

Diversity and inclusion are a big part of the business conversation, and that includes all AI initiatives. Companies are willing to spend more time and resources ensuring their AI meets ethical requirements. Trust in AI is a significant challenge moving forward—not from companies but from their customers. Moving forward, more companies seem to be considering these questions and more as they build their AI projects.

AI offers business powerful solutions but is still in the early stages

Even though AI has been around for a while, true implementation for business value is still relatively new. The misunderstandings business leaders have about the inner workings of AI and machine learning processes could continue to hamper adoption. However, as more of the workforce achieve data literacy, this gap could begin to close.

Understanding that responsible AI includes more than trustworthy data sourcing or basic human oversight is key. Business leaders are beginning to understand the proactive work they must do to ensure that consumers are comfortable and informed about the artificial intelligence projects businesses take on. This will go a long way to ensuring that companies don’t repeat past mistakes in responsible AI implementation.

The lynchpin is and always will be humans. Human intervention and oversight helps ensure more transparent processes and improve the operational implementation of these new technologies. Soon, we may all benefit even more from the rich experiences companies can deliver through the use of artificial intelligence and the resolution of critical global challenges.

Elizabeth Wallace

About Elizabeth Wallace

Elizabeth Wallace is a Nashville-based freelance writer with a soft spot for data science and AI and a background in linguistics. She spent 13 years teaching language in higher ed and now helps startups and other organizations explain - clearly - what it is they do.

Leave a Reply

Your email address will not be published. Required fields are marked *