Published Sep 25, 2023

LLMs for Evil

Dive into the world of adversarial machine learning with researcher Maximilian Mozes as he unpacks the dual nature of large language models, addressing the inherent threats they pose such as misinformation and data poisoning, while exploring sophisticated strategies to safeguard these powerful technologies.

Episode Highlights

Topics covered

Episode Highlights

Related Episodes

Emergent Deception in LLMs
Answers 383 questions
Do LLMs Make Ethical Choices
Answers 383 questions
Deploying LLMs
Answers 383 questions
I LLM and You Can Too
Answers 383 questions
LLMs for Data Analysis
Answers 383 questions
LLMs in Music Composition
Answers 383 questions
Which Professions Are Threatened by LLMs
Answers 383 questions
AI Platforms
Answers 383 questions
Evaluating Jokes with LLMs
Answers 383 questions
Process Mining with LLMs
Answers 383 questions
LLMs in Social Science
Answers 383 questions
The Limits of NLP
Answers 383 questions
NLP for Developers
Answers 383 questions
Program Aided Language Models
Answers 383 questions
ELMo
Answers 383 questions

Dexa/Kyle Polich

LLMs for Evil

Topics covered

Popular Clips

NLP Advancements

Exciting AI Applications

Language Model Labeling

Language Model Threats

Language Models in Cybersecurity

Adversarial Machine Learning

Threats of Large Language Models

Future Data Trends

Language Model Limitations

Safeguarding Language Models

Language Model Privacy

Illicit Uses of AI

AI Safety Solutions

Episode Highlights

Adversarial Machine Learning

Safeguarding Strategies

Threats and Vulnerabilities

Related Episodes

Emergent Deception in LLMs

Do LLMs Make Ethical Choices

Deploying LLMs

I LLM and You Can Too

LLMs for Data Analysis

LLMs in Music Composition

Which Professions Are Threatened by LLMs

AI Platforms

Evaluating Jokes with LLMs

Process Mining with LLMs

LLMs in Social Science

The Limits of NLP

NLP for Developers

Program Aided Language Models

ELMo

LLMs for Evil

Topics covered

Popular Clips

Episode Highlights

Adversarial Machine Learning

Safeguarding StrategiesMaximilian Mozes explores the preventative measures and techniques to safeguard large language models (LLMs) from misuse. He discusses strategies like reinforcement learning, content filtering, and red teaming to enhance model security and prevent harmful outputs.

Safeguarding Strategies

Threats and VulnerabilitiesMaximilian Mozes, a PhD student at University College London, explores the dual-edged nature of large language models (LLMs) in his research. He highlights the potential threats posed by LLMs, including privacy risks, misinformation, and data poisoning.

Threats and Vulnerabilities

Related Episodes