Dr. Alok Kanojia
๐ค SpeakerAppearances Over Time
Podcast Appearances
And the basic problem is that AI is so sycophantic.
Our reality testing of the world requires contrary opinions.
So when you're like, hey, I have this idea.
I want to throw something by you.
And then I say, no.
So how do we know what reality is?
Because we have this perception, but this person has this perception.
This person has this perception.
So we modulate our perception.
Hey, I got you a gift.
No, you didn't.
I told you yesterday that I was going to pick this up.
No, you didn't tell me that.
So we stay in reality because we get signals from reality.
The thing about the AIs is they're language learning models.
They don't actually know anything.
All they do is scrape the internet.
This is a simplification.
I'm not a data scientist or AI engineer, but here's my understanding because people wanted to build like a Dr. K chatbot and I tried to get into understanding the mechanism.
What an AI does is it just says a word and then it tries to figure out which words are gonna make you happy.