Skip to main content

Introduction to AI

Introduction to Generative Artificial Intelligence​

Discriminative AI​

  • Discriminative AI: makes decisions and predictions by analyzing and classifying data.
    • Acts as an evaluator.
  • Generative AI: acts as a creator.
    • creates new content based on what it learns during training.
    • innovator
  • Examples of Gen AI:
    • text generation like chatGPT
    • image generation
    • music synthesis
  • AI simplify complex ideas

How generative AI works​

  • Gen AI learns from datasets
    • text
    • images
    • code
  • Gen AI learns the features and patterns it's fed.
  • It's like a chef understanding ingredients to generate recipes for specific needs.
  • Building blocks of Gen AI: large datasets
  • Gen AI:
    • identifies patterns
    • learns rules
    • extracts information
  • Gen AI can
    • complete
    • improve
    • debug
    • optimize

Generative AI in software development​

  • AI in planning phase:
    • Identify the functional requirements and architectual styles from business needs.
  • AI in Development phase:
    • proposes boilerplates
    • writes complex algorithms
    • optimize current code
  • AI in testing phase:
    • automates unit and integration tests.
  • AI in deployment phase:
    • recommends configurations for deployment
    • monitors for potential issues
  • AI in maintenance phase:
    • monitors for bug detection
    • automatic documentation updates

Boost in productivity​

  • Code suggestions
  • Code autocomplete or completion ideas
  • Method generation

Limitations​

  • Accuracy issues
    • may contain bugs or out-dated suggestions
    • fail to account for
      • rounding inconsistencies
      • empty carts
      • invalid input
      • etc
  • Developers should still consider edge cases

Ethical concerns​

  • suggestions might contain copyright or proprietary code that you shouldn't use or need permission to use.

Waning Personal skills​

  • relying on AI might dampen down your programming skills over time.

Tips​

  • Use AI for efficiency
    • Automate repetitive tasks
    • Generate initial drafts
  • Use human oversight to ensure code meets project standards.

Ethical Considerations and impact​

Data Privacy​

  • Datasets can include personal data
    • AI could potentially remember and reproduce sensitive info.
  • Example: customer service AI trained on transcripts of real customer interactions.
    • Names of customers might be removed
    • But addresses and customer's credit card numbers can potentially be revealed in the AI response.
      • Thus, unintentional privacy breaches can occur.
  • AI can generate unsecured code.
    • For example, unsecured code can be generated to manage user's sensitive data to be stored in the database, leading to possile breaches.
  • AI can suggest proprietary code in a public repo, which another developer might unknowingly use, violating property rights.
  • AI models need to improve on these issues.

Generating harmful or misleading content​

  • Generative AI can be misused to create harmful outputs, ranging from fake news articles to malicious code.
    • e.g. deep fakes can trick people
    • e.g. hackers can taken advantage of developers that utilize malicious code generated from Gen AI.

The societal impacts of GenAI​

Effect on job markets​

  • Prevents junior engineers from joining job market.
  • Needed retraining and upskilling of employees to stay competitive in AI driven market.
  • Legal challenges if AI copies proprietary code or is really similar to it.

The need for regulation​

  • European Union’s AI Act (2024): Categorizes AI systems based on risk levels and ensures that high-risk applications meet strict standards.
  • United States’ AI Bill of Rights (2024): Provides guidelines to safeguard privacy, civil rights, and transparency in AI systems.
  • China’s AI Regulations (2023-2024): Emphasizes innovation and national security while enforcing tight controls on AI development.
  • Workers may face job displacement if they lack access to reskilling opportunities, highlighting the need for legal provisions that support workforce development in AI-driven industries.
  • Unchecked biases in GenAI systems could perpetuate discrimination and reinforce inequalities, emphasizing the importance of regulations to ensure fairness and inclusivity.
  • Privacy violations, threats to civil liberties, and a lack of transparency could arise without proper safeguards, leading to mistrust and potential societal instability.
  • GenAI-generated social media content and deepfakes could manipulate public perception, fueling misinformation and causing significant societal disruption.
  • High-risk applications, such as AI tools used in healthcare, require stringent safety and accuracy standards to prevent harm and ensure public trust.

Critical thinking: Sensitive applications of GenAI​

  • News generation
    • The AI could prioritize sensational headlines over factual accuracy, undermining trust in news sources.
    • To address these risks, it is essential for news organizations to implement strict human oversight when using AI, ensuring that journalistic integrity and accuracy remain at the forefront.
  • Deepfakes
    • It’s essential to develop and implement tools that can accurately detect AI-generated content, ensuring the responsible use of this technology and protecting public trust to mitigate these dangers.