meta_pixel
Tapesearch Logo
Log in
The Journal.

Artificial: Episode 2, Selling Out

The Journal.

The Wall Street Journal

News, Daily News, Business News

4.25.8K Ratings

🗓️ 10 December 2023

⏱️ 31 minutes

🧾️ Download transcript

Summary

OpenAI’s breakout product, ChatGPT, had humble origins. What started as a small research project ballooned into something much bigger: a groundbreaking large language model. But developing that technology was expensive, and to fund it, OpenAI would make a big compromise. Further Reading: - Elon Musk Tries to Direct AI—Again - The Contradictions of Sam Altman, AI Crusader Further Listening: - Artificial: Episode 1, The Dream - The Hidden Workforce That Helped Filter Violence and Abuse Out of ChatGPT Learn more about your ad choices. Visit megaphone.fm/adchoices

Transcript

Click on a timestamp to play from that location

0:00.0

Melanie Subaya is 28 years old. She lives in New York and she's an artificial intelligence researcher.

0:16.0

As a kid, were you always interested in computers?

0:20.0

I actually

0:25.0

I really loved reading and writing and that's actually how I got into natural language

0:27.6

processing eventually and that then kind of became my interest in

0:31.0

loving computer science.

0:32.0

And when Melanie was in college in 2016, she

0:36.3

tried to bring together her two interests, computer science and literature. Her idea was to build an AI model that could write short stories,

0:47.4

which at the time was a tall order.

0:51.1

How good of a writer was AI back then?

0:53.2

Terrible, absolutely terrible.

0:55.0

Could it string a sentence together?

0:57.0

It could.

0:58.2

It was hard to get beyond a sentence and definitely beyond a couple sentences was very hard.

1:05.4

For her senior thesis, Melanie built what's called a language model.

1:10.2

She fed a computer around 100,000 examples of short stories, really short ones, just five sentences long.

1:18.2

And then she asked the model to write its own story. She would give it the first line and the AI would fill in the rest.

1:27.2

For some reason the model loved to end the story with somebody getting very nervous and going to buy a car.

1:37.2

And so... with somebody getting very nervous and going to buy a car and so I'm sorry is that how where did it learn that for some reason that was what it landed on So an example is Tyrone was working at his job in a local restaurant. He was very nervous about his first job. He was very nervous about the job. He was very nervous about it. He went to the store to quite a few examples. Can we hear another one?

2:07.0

Yeah. Frank had surprised the whole family when he came home that day.

2:11.0

When he got home, he was able to get a new car. He was very

2:14.3

happy to be able to get his own. He was very happy with his new job. He was very

...

Please login to see the full transcript.

Disclaimer: The podcast and artwork embedded on this page are from The Wall Street Journal, and are the property of its owner and not affiliated with or endorsed by Tapesearch.

Generated transcripts are the property of The Wall Street Journal and are distributed freely under the Fair Use doctrine. Transcripts generated by Tapesearch are not guaranteed to be accurate.

Copyright © Tapesearch 2026.