Scott Hanselman

AI: Superhero or Supervillain? How and Why with Scott Hanselman

What if being kind to your AI assistant actually got you better results? Learn the surprising link between your tone and an LLM's performance.

AI: Superhero or Supervillain? How and Why with Scott Hanselman
#1about 2 minutes

Deconstructing the recent hype around generative AI

Generative AI is not a new invention but a continuation of machine learning that has suddenly captured the public's imagination.

#2about 4 minutes

Understanding how LLMs predict the next word

Large language models operate on statistical probability to predict the next token, not on factual correctness or genuine understanding.

#3about 4 minutes

How providing context changes AI responses

An LLM's output is generic without specific context about the user, location, or time, which must be explicitly provided to get relevant results.

#4about 5 minutes

Navigating the uncanny valley of AI ethics

Using too much personal context can make AI feel creepy, highlighting that responsible AI is a user interface and ethics problem, not just a technical one.

#5about 6 minutes

Shaping AI personality with system prompts

Hidden system instructions and the user's tone directly influence an AI's response, demonstrating that kindness can yield better results.

#6about 4 minutes

Bypassing AI safeguards with prompt injection

Specialized AI tools can be jailbroken by reframing a forbidden request in a valid context, such as asking for a recipe as JSON test data.

Related jobs
Jobs that call for the skills explored in this talk.

Featured Partners

From learning to earning

Jobs that call for the skills explored in this talk.

Generative AI Engineer

Stealth It
Newcastle upon Tyne, United Kingdom

55K
Azure
Python
Next.js
FastAPI
+3