I don’t find this argument convincing. I don’t think Sam did a great job either but that’s also because he has to be super coy about his company/plans/progress/techniques etc.
The Jordan Peterson comment was making fun of Lex and a positive comment for Sam.
Besides, I can think Sam did kinda bad and Elizier did kind of bad but expect Elizier to do much better!
.
I’m curious to know your rating on how you think Eliezer did compare to what you’d expect is possible with 80 hours of prep time including the help of close friends/co-workers.
I would rate his episode at around a 4⁄10
Why didn’t he have a pre-prepared well thought list of convincing arguments, intuition pumps, stories, analogies, etc. that would be easy to engage with for a semi-informed listener? He was clearly grasping for them on the spot.
Why didn’t he have quotes from the top respected AI people saying things like “I don’t think we have a solution for super intelligence.”, “AI alignment is a serious problem”, etc.
Why did he not have written notes? Seriously… why did he not prepare notes? (he could have paid someone that knows his arguments really well to prepare notes for him)
How many hours would you guess Eliezer prepared for this particular interview? (maybe you know the true answer, I’m curious)
How many friends/co-workers did Eliezer ask for help in designing great conversation topics, responses, quotes, references, etc.?
This was a 3-hour long episode consumed by millions of people. He had the mind share of ~6 million hours of human cognition and this is what he came up with? Do you rate his performance more than a 4/10?
I expect Rob Miles, Connor Leahy, or Michaël Trazzi would have done enough preparation and had a better approach, and could have done an 8+/10 job. What do you think of those 3? Or even Paul Christiano.
Eliezer should spend whatever points he has with Lex to get one of those above 4 on a future episode is my opinion.
I don’t find this argument convincing. I don’t think Sam did a great job either but that’s also because he has to be super coy about his company/plans/progress/techniques etc.
The Jordan Peterson comment was making fun of Lex and a positive comment for Sam.
Besides, I can think Sam did kinda bad and Elizier did kind of bad but expect Elizier to do much better!
.
I’m curious to know your rating on how you think Eliezer did compare to what you’d expect is possible with 80 hours of prep time including the help of close friends/co-workers.
I would rate his episode at around a 4⁄10
Why didn’t he have a pre-prepared well thought list of convincing arguments, intuition pumps, stories, analogies, etc. that would be easy to engage with for a semi-informed listener? He was clearly grasping for them on the spot.
Why didn’t he have quotes from the top respected AI people saying things like “I don’t think we have a solution for super intelligence.”, “AI alignment is a serious problem”, etc.
Why did he not have written notes? Seriously… why did he not prepare notes? (he could have paid someone that knows his arguments really well to prepare notes for him)
How many hours would you guess Eliezer prepared for this particular interview? (maybe you know the true answer, I’m curious)
How many friends/co-workers did Eliezer ask for help in designing great conversation topics, responses, quotes, references, etc.?
This was a 3-hour long episode consumed by millions of people. He had the mind share of ~6 million hours of human cognition and this is what he came up with? Do you rate his performance more than a 4/10?
I expect Rob Miles, Connor Leahy, or Michaël Trazzi would have done enough preparation and had a better approach, and could have done an 8+/10 job. What do you think of those 3? Or even Paul Christiano.
Eliezer should spend whatever points he has with Lex to get one of those above 4 on a future episode is my opinion.
@habryka curious what you think of this comment