I really don't get why people take Bender seriously as a source for information about AI. She's been consistently wrong about model capabilities for years now, and seems to have nothing but contempt for the computer scientists, technologists, and machine learning experts that are driving this technology. Look at her bluesky with its screeds against "mathy maths" "tech bros" etc
I do not know that person, but her views in the linked video seem great and vastly underrepresented in the current media landscape.
She lists many valid AI risks, discounts the doomsday PR and instead talks about the accountability gap that current AI frontier companies profit from.
I'm sure her Twitter timeline is awful, as you noted. But do you have thoughts on the arguments she presents in the linked video?
Contrarians gonna contrarian. Zitron is the same way. The only thing that will convince these people is if an incident actually happens - and even this is questionable, because clearly they don't understand the OpenAI incident well enough to take it seriously.
There’s a small cottage industry of people who now have careers built on telling people that AI is a grift. Bender, Ed Zitron, Gary Marcus, etc. And there is a large audience who wants to hear this message.
What is troubling to me is I can’t think of a single prominent AI skeptic who has changed their mind. Are there any?
Wait, the Emily Bender who just undertook an increasingly astonishing credibility conflagration on the basis that LLMs can't do arithmetic unless the answers are in their training data? That Emily Bender? Gosh, I'm inclined to agree that AI x-risk is mostly fake, but with friends like these...
We must stop conflating the risk of AI and the risk of AI companies.
The AI is harmless. Without human input to turn it on, it is inert. The same danger as a piece of driftwood.
AI employees do not have to turn the AI on. It's not compulsory! If it is risky, stop doing it.
I do not see any members of the public with 200,000 GB300s, secretly training dangerous frontier models.
Generally speaking, if you need as much electricity as a large city, it's quite hard to hide.
The problem is not the AI, it is bad people in AI harming others; if there was no AI perhaps they would have done the same thing another way.
15 comments
[ 0.28 ms ] story [ 18.1 ms ] threadShe lists many valid AI risks, discounts the doomsday PR and instead talks about the accountability gap that current AI frontier companies profit from.
I'm sure her Twitter timeline is awful, as you noted. But do you have thoughts on the arguments she presents in the linked video?
The AI is harmless. Without human input to turn it on, it is inert. The same danger as a piece of driftwood.
AI employees do not have to turn the AI on. It's not compulsory! If it is risky, stop doing it.
I do not see any members of the public with 200,000 GB300s, secretly training dangerous frontier models.
Generally speaking, if you need as much electricity as a large city, it's quite hard to hide.
The problem is not the AI, it is bad people in AI harming others; if there was no AI perhaps they would have done the same thing another way.
What is troubling to me is I can’t think of a single prominent AI skeptic who has changed their mind. Are there any?
I think that is correct. Alignment is a red herring.
edit: typo
(https://bsky.app/profile/colin-fraser.net/post/3mwrfschsn22p)
AI employees do not have to turn the AI on. It's not compulsory! If it is risky, stop doing it.
I do not see any members of the public with 200,000 GB300s, secretly training dangerous frontier models.
Generally speaking, if you need as much electricity as a large city, it's quite hard to hide.
The problem is not the AI, it is bad people in AI harming others; if there was no AI perhaps they would have done the same thing another way.