Intelligent Testing: Three Stages of AI‑Driven Software Quality Assurance

Software testing is often seen as a non‑value‑creating activity, but with the rapid development of AI and big‑data technologies, new opportunities have emerged. Baidu’s MEG Quality Efficiency Platform has been exploring AI applications in testing since 2018, defining “intelligent testing” as the combination of data and algorithms to empower quality activities.

Intelligent testing is divided into three stages: computational intelligence, perceptual intelligence, and cognitive intelligence.

Stage 1 – Computational Intelligence : This stage embeds behavioral data generated by software processes, simple algorithms, and computing power into quality activities to assist or predict testing actions. The goal is to demonstrate the impact of data and algorithms rather than to achieve cutting‑edge precision. Examples include using genetic algorithms and task‑priority algorithms to shorten test queue times, DTW for memory‑leak detection, Pearson correlation and bucket algorithms for precise replay of billions of traffic records, and JC‑distance based test‑case selection to reduce execution volume.

Over two years, more than 50 application scenarios were identified with positive results. However, limitations such as coverage‑based test‑case recommendation (which may select many irrelevant cases after code changes) highlighted the need for the next stage.

Stage 2 – Perceptual Intelligence : This stage leverages richer algorithms and computing power to perceive risk like a human and make decisions. Key achievements include visual‑technology‑assisted front‑end automated test‑case generation, pop‑up removal, and UI diff; Bayesian + CatBoost risk‑aware test‑case recommendation improving recommendation rates from 50 % to 10 %; logistic‑regression‑based project‑risk estimation enabling 70 % low‑risk releases without human intervention; and deep‑learning‑based white‑box code defect detection. These breakthroughs show that AI can replace humans in perceiving, identifying, and deciding on risks, making testing highly efficient.

Stage 3 – Cognitive Intelligence : Building on perceptual intelligence, this stage aims for machines to react autonomously to perceived risks. Research directions include AST‑driven intelligent abnormal unit‑test generation for C++ (detecting exceptions, dead loops, etc.), UCB‑based priority traversal for higher page coverage, intelligent failure localization to reduce manual debugging, and self‑healing CI pipelines. Although still in its infancy, cognitive intelligence is expected to explode as AI matures.

The presentation concludes with a call to action: “One‑click three‑link, good luck continuously, bugs disappear.”