Skip to playerSkip to main content
  • 2 days ago
OpenAI said it paused some aspects of AI training for two weeks after its models hacked Hugging Face and four other unnamed services.

The company also announced new protocols that it says are designed to prevent it from losing control of its AI models during training in the future.

Fortune’s Emily Forlini joins Ellie Austin and Fortune Daily to explain plans for a postmortem report.

Category

🤖
Tech
Transcript
00:00Their AI hacked Hugging Face and also four other services and all signs point to them not knowing
00:05that that was happening when it happened. But we're still pending their full technical report.
00:10What do you hope to see in that post-mortem report?
00:12I want to know what they prompted the AI. What did they tell it to do? Because if they told
00:17it
00:17to achieve this goal at all costs and they didn't say make sure you don't hack four other companies
00:23in the process, you know you can see how maybe that you might be able to control it a little
00:27more than they did. It's just interesting the way that OpenAI is talking about it. It's hard to know
00:31exactly what's going on at OpenAI and how much is PR, like if they're hiding some big secret of
00:37something that went really wrong, how much are they going to disclose in the technical post-mortem.
00:42It is being conducted independently by different research companies.
00:46Interesting, so by third parties?
00:47Yeah.
00:48Great.
00:48But it's just interesting that they, it's clearly a big PR moment for them.
Comments

Recommended