What Is AI? History, Current Trends, and Future Challenges Explained

01 What Is AI?

According to Wikipedia, artificial intelligence (AI) is a General Purpose Technology that spans many economic sectors, offering significant technical complementarity and spillover effects. In short, AI is a foundational technology of the 21st century that permeates everyday life.

02 AI History, Present, and Future

Early AI was influenced by 20th‑century mathematical philosophy, giving rise to two schools: symbolic AI based on formalism and logicism, and connectionist/behaviorist AI rooted in constructivism. These early approaches promised great potential but later proved far more complex than expected.

AI experienced cycles of rise and fall until 2006, when Geoffrey Hinton and Ruslan Salakhutdinov published a paper solving the vanishing‑gradient problem in deep networks, sparking the deep‑learning wave in academia and industry.

In 2012, Hinton’s team won the ImageNet ILSVRC competition with a 10% lower error rate than the runner‑up, further accelerating deep‑learning adoption.

The surge was powered by massive datasets (e.g., ImageNet) and GPU‑accelerated training, providing the compute power needed for large‑scale models.

03 AI Application Drivers

Data abundance : IoT devices, smartphones, and wearables generate massive data; IDC predicts global data will grow from 33 ZB in 2018 to 175 ZB by 2025, making data the “oil” of AI.

Algorithm breakthroughs : Beyond CNN/RNN models, reinforcement learning, GANs, and large‑scale language models like BERT have surpassed human performance on benchmarks (e.g., GLUE 80.4%, MultiNLI 86.7%). Vision tasks such as image classification already exceed human accuracy, while speech recognition reaches a 3.71% character error rate comparable to professional stenographers.

AI chips : Specialized hardware (e.g., NVIDIA Tesla V100, Google TPU v3, Huawei Ascend 310/910) delivers orders‑of‑magnitude performance gains, breaking Moore’s‑law limits and enabling ZB‑scale data analysis.

04 Will AI Surpass Humans?

Some algorithms now outperform humans on specific datasets, prompting the question of whether AI will exceed human capabilities. However, many problems are not mathematically formalizable, and only a small subset of AI‑solvable problems lie within the “blue circle” of the illustrated diagram.

Mathematical limits (undecidable problems, non‑computable functions) mean that not all questions can be solved by any algorithm or Turing machine. Consequently, AI can excel in narrow domains but cannot universally replace human intelligence.

Even when AI achieves high accuracy, it can make glaring mistakes—e.g., misclassifying adversarially perturbed images or confidently labeling noisy patterns as specific objects—demonstrating that AI predictions are not infallible.

Thus, AI should be viewed as a powerful tool that augments human productivity rather than a replacement for human cognition.

05 Three Major Challenges for AI Beginners

1. Broad prerequisite knowledge : AI development requires Python, Linux, image processing (OpenCV), and solid mathematics.

2. Lack of GPU resources : Training models on personal GPUs is costly.

3. Limited community support : Non‑CS students often find fewer learning resources and peer groups compared to traditional software development.

06 Conclusion

Fortunately, Huawei Cloud offers the ModelArts AI development platform, providing low‑cost, low‑threshold access to AI tools, tutorials, and community resources to help overcome the challenges mentioned above.