## Artificial Intelligence’s Promise and Peril

## Source details

**Canonical URL:** [Artificial Intelligence’s Promise and Peril](https://www.imf.org/-/media/files/publications/fandd/article/2023/december/08-09-b2b-final.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/fandd/article/2023/december/08-09-b2b-final.pdf.md)
- [Structured JSON version](/-/media/files/publications/fandd/article/2023/december/08-09-b2b-final.pdf.json)

---

### What is generative AI?
- GenAI represents the most impressive advance in machine-learning technologies yet, marking a significant leap in AI’s ability to understand and interact with complex data patterns.
- It is poised to unleash a new wave of creativity and productivity but raises important questions for humanity.

### Technological milestones
- 1950: Alan Turing envisioned machines reaching mastery in an “imitation game.”
- 1960s: ELIZA generated human-like responses and served as a precursor to modern chatbots.
- 1980s (two decades after the 1960s): artificial neural networks appeared, enabling machines to understand nuances of language and recognize images, but progress was limited by data and computing power.
- 2000s: Deep learning emerged as the third wave of AI, enabled by continually increasing data and computing power.
- 2014: Generative adversarial networks (GANs) introduced the “generator” and “discriminator” dual-network competition that revolutionized AI’s replication of complex patterns.
- 2017: The paper “Attention Is All You Need” introduced attention mechanisms that allowed AI to pay attention to relevant parts of input, improving generation of human-like content.
- November 2022: ChatGPT was launched by OpenAI, followed by GenAI chatbots from other big-tech firms.

### Economics and finance applications
- Traditional AI (advanced analytics, machine learning, predictive deep learning) has long been used for crunching numbers, gauging market trends, and customizing financial products.
- GenAI’s distinguishing capability: to delve deeper and interpret complex data in a more creative manner, producing not just forecasts but alternate scenarios, insightful charts, and snippets of code.
- Public sector adoption:
  - Governments are beginning to employ GenAI to improve citizen services and overcome workforce shortages.
  - Central banks see potential to sift through vast amounts of banking data to refine economic forecasts and better monitor risks, including fraud.
- Private sector adoption:
  - Investment firms use GenAI to detect subtle shifts in stock prices and market sentiment and to propose more creative investment options.
  - Insurance companies explore generative models to create personalized policies aligned with individual needs and preferences.

### Risks and concerns
- Hallucination: AI can create nonsensical and untrue facts (a phenomenon called “hallucination”) and may not know the meaning behind words; ChatGPT’s knowledge is limited to its latest training date.
- Bias and transparency: Traditional challenges—amplification of existing biases in training data and lack of decision transparency—have renewed urgency.
- Weaponization and misinformation:
  - GenAI can tell stories that resonate with individuals’ preexisting beliefs, reinforcing echo chambers and ideological silos.
  - Malicious actors can leverage GenAI to manipulate politics, markets, and public opinion.
  - March 2022: An AI-generated video purported to show Ukrainian President Volodymyr Zelenskyy surrendering to Russian forces, illustrating potential for manipulation.
  - “GenAI creations can be so convincing that they create a false sense of reality. This has the potential to spread misinformation, incite panic, and even destabilize economic or financial systems.”
- Job displacement: GenAI’s advancement may automate tasks previously performed by humans, potentially leading to job losses and requiring strategies for employment and retraining.
- Existential risk concerns: Leading AI experts, including ChatGPT’s creator, cosigned a letter warning that “mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” Turing’s earlier warning is recalled: “there is a danger that machines will eventually take control of our lives.”

### Recommendations and governance themes
- The piece emphasizes the need for:
  - Vigilant oversight.
  - New regulatory frameworks.
  - An unwavering commitment to ethical, transparent, controllable innovations that harmonize with human values.
  - Strategies for employment and retraining to address potential job displacement.
- The conclusion: GenAI is a monumental shift that cannot be uninvented and demands policy action consistent with the risks and opportunities outlined.

*Hervé Tourpe is head of the IMF’s Digital Advisory Unit.*

---


_Source: https://www.imf.org/-/media/files/publications/fandd/article/2023/december/08-09-b2b-final.pdf_