The shadows that follow the AI generative models

What happens when the AI generating your content confidently lies or amplifies harmful societal biases? This talk explores the ethical shadows of generative models.

#1about 6 minutes

Understanding how generative AI models create content

Generative AI models are trained on vast datasets to create new content like images, music, and code from user prompts.

#2about 5 minutes

The challenge of correctness and model hallucination

AI models can provide incorrect or outdated information and even "hallucinate" facts due to their training data and susceptibility to being tricked.

#3about 1 minute

Understanding the security risk of prompt injection

Prompt injection is a security vulnerability where malicious user input can manipulate a model's output, similar to SQL injection attacks.

#4about 2 minutes

Assigning ownership and responsibility for AI content

Determining who is responsible for AI-generated content, such as prize-winning art or offensive jokes, raises complex legal and ethical questions.

#5about 4 minutes

How training data creates biased AI models

AI models can perpetuate and amplify societal biases present in their training data, leading to stereotypical or discriminatory outputs.

#6about 2 minutes

The serious threat of malicious and illegal AI use

Generative AI can be exploited for illegal activities like creating non-consensual deepfake pornography, spreading fake news, and perpetrating scams.

#7about 5 minutes

The role of regulation and ethics in AI development

Governments are introducing regulations like the EU AI Act to mitigate risks, while a growing focus on AI ethics aims to guide responsible development.