What Will AI Look Like by 2023? Key Trends, ML/DL Insights, and DevOps Strategies

Introduction

In recent years AI has moved from sci‑fi fantasy to a pervasive technology driving retail, logistics, education, healthcare and more. China’s AI adoption has tripled since 2018, driven by cost reduction, efficiency gains and improved customer experience, yet growth remains chaotic.

Part 1: AI Overall Trend Forecast

From now to 2023 AI will continue rapid development, but uneven across domains. Key predictions:

By 2021 AI processor revenue will triple due to cross‑cloud and edge deployments.

By 2021 AI use cases will drive exponential data‑storage growth, requiring new architectures.

By 2021 cross‑delivery models (cloud, data‑center, edge) will foster new inference and retraining ecosystems.

By 2022 only 15% of AI projects involving edge or IoT will succeed.

By 2023 AI compute consumption will be five times that of 2018, making AI a primary factor in infrastructure decisions.

By 2023 70% of AI workloads will run in containers or serverless models that need DevOps.

By 2023 40% of I&O teams in large enterprises will use AI‑enhanced automation to boost IT productivity, flexibility and scalability.

These findings show AI will follow its own pace while being market‑tested.

Key Insights from Trend Analysis

Early AI success is driven by strategic adoption of new algorithms and APIs.

Maturing ecosystems let early adopters boost productivity and automation across business lines.

Over‑optimistic expectations can cause project overruns.

Enterprises seek multi‑dimensional ML/DL models for higher accuracy and business impact.

Most AI initiatives fail; successful ones can deliver revolutionary user experiences.

Recommendations for infrastructure leaders include aligning strategy with business needs, combining development, purchase and outsourcing with mature tools, and hiring or up‑skilling AI experts.

Part 2: ML and DL Development

ML and DL are critical for autonomous vehicles, medical applications, etc. The institute predicts that by 2022 over 75% of enterprises will replace traditional ML with DL.

However, only 15% of AI projects involving edge or IoT will succeed, reflecting high complexity.

Key characteristics of ML/DL use cases:

Edge/IoT scenarios face complex data‑analysis challenges.

Embedded applications lack business‑line support outside a few industries.

DL gains traction in asset management and resource planning, while classic ML remains preferred for audit‑heavy sectors.

DL excels in perception tasks (image, translation, NLP) that require massive raw data.

Classic ML needs less data and compute, offering transparency.

Further analysis highlights that DL deployments on edge devices still need powerful hardware and optimization, and that successful AI projects rely on close collaboration between IT and business functions.

Advice for the market impact of ML/DL includes focusing on quick‑ROI IoT cases, leveraging internal engineering teams, fostering collaboration among analytics, business, and domain experts, learning the strengths and limits of each algorithm, and preferring simple ML solutions when appropriate.

*This article is part one of Suning Retail Technology Institute’s “AI Key Technology Forecast and Its Application in I&O Team Building and DevOps Systems”. The next part will follow.