In the third of a series of articles, Gareth Smith, General Manager, Software Test Automation at Keysight looks at key trends in AI in 2024.
As AI becomes increasingly embedded in software, the systems will become more autonomous, which increases risk and complexity and makes testing a real challenge. As a result, a fixed set of tests (scripts) will no longer suffice when evaluating intelligent systems. Instead, AI will be needed to automatically and continuously test applications. The future of software testing is autonomous test design and execution, says Smith.

As AI permeates every system and complexity and sophistication soar, there is a risk that quality will go down. This is a result of the sheer number of permutations, which makes testing everything impossible. This means decisions will need to be made around how, what, and when to test to ensure quality is maintained.

Automated, AI-driven test tools optimize experiences on 5G 

AI regulation also needs to be deep and wide. There is universal acceptance of the need to regulate AI. However, what the regulation should encompass will be subject to much debate due to the breadth and complexity of the technology involved. It will take a seismic event with significant negative consequences before the necessary funding is available. Only then will clear standards and best practices come into effect. If regulation doesn’t happen in the near future, it increases the risk that it will no longer be possible to rein AI in.

As the risks associated with AI are recognized, enterprises will need to appoint an AI and security compliance officer to the C-Suite. Over time, this role will merge with the CSO. With live learning, it will be vital to have guardrails in place to keep AI on track. Constant checks and balances will be essential to validate that an intelligent system is behaving and hasn’t gone rogue. Live surveillance will become standard. However, as these systems develop, it will also be necessary to test that they haven’t learned how to look like they are behaving while undertaking nefarious activity. Reinforcement learning and similar techniques can inadvertently drive the AI to cover its tracks to reach its goal and will be a huge challenge to address before the end of the decade. These problems will create a slew of new opportunities for companies that can help clean up, control, and put guardrails in place for AI.

Keysight teams to add AI into 6G

Currently, AI systems are tested by the companies building them. As the risks are increasingly understood, having an independent body to verify that an AI system is compliant is essential. Gaining an AI certification (i.e., AI driver’s license) will be the first step. However, just like your car, it will require a regular test to ensure it remains ethical, responsible, free of bias, and meets the necessary country and industry standards. In the longer term, this may result in an NFT label on each AI system to validate that it’s fit for purpose and meets all the required criteria.

Bosch rolls out generative AI in manufacturing

Citizen developers have long been touted as the answer to the IT talent shortage. However, the rapid growth of AI-powered solutions is fueling a new generation of business developers. These domain experts will increasingly be involved in the SDLC as they understand the goals and operations of the enterprise. This will give rise to a new wave of no-code systems that enable business users to define goals and then have AI technology close the gap. The operational knowledge ensures that the software meets the specific needs of the organization and mitigates the risk.

Sustainability

There has been significant hype around how AI systems will transform our lives, but little attention has focused on the compute power required. In 2024, AI’s impact on sustainability will enter the spotlight, and organizations will start to monitor the carbon footprint of their entire technology infrastructure as they strive to meet net-zero targets. As a result, companies will need to decide where and how to judiciously use AI rather than thinking it can be deployed everywhere. And when it comes to testing software and applications, businesses will also have to pivot from testing everything to predicting the tests that matter most to reduce the environmental impact.