Artificial intelligence (AI) and machine learning (ML) have quickly progressed from niche technology trends to frequent integration with business operations, new products and services, and customer service innovations across industries. According to Grand View Research, the artificial intelligence software market reached $62.3 billion in 2020 and is expected to grow exponentially, hitting $997.8 billion by 2028.
As the AI software market expands, there are several software trends we expect to see in the next several years: watch for increased automation, more intelligent security practices, and a better understanding of AI ethics.
Trends to Watch in Artificial Intelligence Software
- The Evolution of AIOps and MLOps
- AI/ML to Automate Basic Cybersecurity Tasks
- Growing Number of Data Quality Solutions
- Examining AI Ethics
Artificial intelligence and machine learning for IT operations, known as AIOps and MLOps respectively, are likely the two fastest-growing operational practices among major enterprises. They support a growing drive toward automation and consolidation of back-office operations, using machine learning and big data analytics to automate network monitoring, troubleshooting, and other network management tasks.
By using AI and ML to consolidate network management tools and limit the need for human action on basic operations, network administrators are free to spend more time on strategic network efforts. AIOps is growing quickly as people see the saved time and costs that it brings.
Certain tasks can and should be automated with AI and ML to decrease user error, and free up time for your team’s cybersecurity experts to focus on more complex issues. These are some of the top cybersecurity areas that can be automated:
- Day-to-day security management
- Threat spotting/network monitoring
- Security log reading
- Alerts for escalated threats
The key to successfully automating cybersecurity with AI and ML is developing and constantly improving upon the training data you feed into these systems. Without detailed protocols and training, your cybersecurity AI will miss key management, auditing, and alerting tasks that could jeopardize your network’s safety or make your ML scripts more susceptible to breach.
High-quality training data is the only way companies can really take advantage of AI solutions, which is why many are investing time and resources into cleaning up their data. This focus is not only on the legibility of data, but also on the overall compliance and scalability of that data:
- Data governance tools are helping enterprises to ensure their training data adheres to all appropriate data protection laws and regulations.
- Data annotation tools make qualitative, quantitative, structured, and unstructured data legible for ML technologies.
- Smart data fabrics, data lakes, and data warehouses continue to grow as more enterprises recognize the need for big data storage space that also offers high levels of searchability.
“Some of the hottest tools these days are the ones typically considered to be the least sexy – quality profiling tools and data governance tools,” O’Connor said. “Tools that automate insights into the quality of data and enable that quality to be significantly improved through automation can have an exponential impact on the quality of analytical insights.”
As AI software grows in its capabilities and widespread use, developers, enterprises, and users alike have developed concerns about the ethics behind these tools. These are the areas of concern that have already arisen in AI ethics and some that will likely pose a problem for AI software vendors and users in the future:
Voice Recognition Software
According to recent studies, the voices of many BIPOC and non-native English speakers are not always picked up by the voice recognition of smart speakers and other natural language processing (NLP) software. The studies explain that voice recognition technology from Amazon, Apple, Google, IBM, and Microsoft misidentified 35% of words from Black individuals, with a significantly smaller margin of error for white users.
The study notes that the majority of developers behind this technology are white, and thus did not account for vocal or dialectical differences in the development of voice recognition technology.
Facial Recognition Software
Facial recognition technology poses consequential problems for BIPOC, transgender, and nonbinary communities. Simple — though still harmful — racist profiling and misgendering can occur for users, causing problems like medical misdiagnosis. Errors with this technology have even limited medical studies among certain communities.
But there are greater safety concerns with this type of technology when government and law enforcement groups use computer vision. If an individual’s facial profile “matches” a certain minority group’s template profile in the system, many police and surveillance technologies are trained to watch their actions more closely in public. Innocents are even targeted for arrest in group gatherings.
The growing use of computer vision and its inherent biases pose concerns for the safety of minority and disadvantaged groups in settings like airport security and public protests — both spaces where minority groups are already discriminated against.
AI in the Wrong Hands
AI powers and simplifies many business processes, and often aids in humanitarian efforts like medical diagnosis and treatment. But what happens when powerful, humanoid technology gets into the hands of terrorists, warring factions, and other malevolent actors?
We’ve already seen the earliest developments of adversarial machine learning, or the practice of hacking into ML technology and changing its internal script or actions. Such a breach can have negligible impact in some cases, but it can also have dire consequences: driving a self-driving car off the road or activating an AI-powered military drone, for example.
There’s also the development of generative adversarial networks and synthetic content generation, colloquially known as “deepfakes.” This artificial manipulation and production of media raises many concerns about copyright, creative ownership, and the spread of disinformation. Adversarial AI capabilities continue to grow, so the question becomes: are we creating technology that does more damage than good?
How AI Impacts Human Resources: HRIS Trends for 2021: The Future of HR Management