Is A.I. Poisoning Itself? + Billionaire Cage Fight + Cooking With ChatGPT

2023/6/30

Hard Fork

AI Deep Dive AI Chapters Transcript

People

Casey Newton

Joe Bernstein

Kevin Roos

Topics

Kevin Roos：人工智能生成内容泛滥，导致互联网信息质量下降，用户体验变差。研究表明，人们越来越多地使用AI完成在线任务，这可能会损害社会科学研究的准确性，并污染互联网数据。如果用AI生成的数据训练AI系统，会导致模型退化甚至崩溃。AI模型输出倾向于重复高概率预测，这会导致训练数据偏差，并最终损害模型的质量。大型语言模型本质上是剽窃引擎，过度依赖现有数据会限制其创新能力。解决AI模型崩溃问题的方法包括改进AI生成内容识别系统和重视训练数据质量。目前使用的AI模型主要基于2022年之前的互联网数据进行训练，因此尚未受到AI生成内容的显著影响。AI模型需要人类的创造性输入才能保持高质量输出，人类在互联网上仍然扮演着重要角色。 Casey Newton：研究表明，人们越来越依赖大型语言模型完成在线任务，这可能会损害社会科学研究的准确性，并污染互联网数据。用AI生成的数据训练AI系统会导致模型退化甚至崩溃。少量AI生成数据不会对AI模型造成严重损害，但需要开发新的技术来过滤低质量数据。互联网内容中AI生成内容的比例不断增加，这对消费者和AI开发者都构成挑战。

Deep Dive

Chapters

The podcast discusses the impact of AI-generated content on the internet, highlighting issues like misinformation, data degradation, and the potential for AI models to collapse due to ingesting their own outputs.

Shownotes Transcript

Whether it’s on TikTok or Twitter, A.I.-generated content is already flooding the web. So, what happens when the technology — prone to confidently making things up — starts ingesting itself?

Then, the New York Times reporter Joe Bernstein talks about why Mark Zuckerberg wants to fight Elon Musk in a cage match.

Plus, we put ChatGPT’s recipe generation to the test with A.I. cocktails.

Today’s guests:

Joe Bernstein) is a Styles reporter at The New York Times.
Priya Krishna) is a Food staff reporter at The New York Times.

Additional reading:

Generative A.I. is already changing) the web and potentially harming itself).
The Times’s Cooking team tried cooking recipes made by ChatGPT). The results were mixed).
Mark Zuckerberg has been working out) — and he’d like you to know about it.

Is A.I. Poisoning Itself? + Billionaire Cage Fight + Cooking With ChatGPT 01:08:40 Share

Hard Fork

Deep Dive

Shownotes Transcript

Is A.I. Poisoning Itself? + Billionaire Cage Fight + Cooking With ChatGPT