These forecasts seemed much more bearish than I was expecting, and I think the reason why is because I believe these forecasts were made before ChatGPT came out.
Can someone give me the TLDR on the implications of these results in light of the fact that Samotsvety's group seemingly/perhaps had much higher odds for AI catastrophe? I didn't read the exact definitions they used for catastrophe, but:
Samotsvety's group (n=13) gave "What's your probability of misaligned AI takeover by 2100, barring pre-APS-AI catastrophe?" at 25% (source https://forum.effectivealtruism.org/posts/EG9xDM8YRz4JN4wMN/samotsvety-s-ai-risk-forecasts)
Whereas XPT gave "AI Catastrophic risk (>10% of humans die within 5 years)" for year 2100 at 2.13%
Without having read the exact definitions for "misaligned AI takeover" and still knowing that Samotsvety's prediction was conditional on pre-APS-AI catastrophe not happening, this still seems like a very large discrepancy. I know that Samotsvety's group was a much smaller n. n=13 vs n=88. How much weight should we give to Samotsvety's group's other predictions on AI timelines given the discrepancy in the risk prediction likelihoods?
Thanks for writing this 🙂 I found value in it. I'd like to write more acknowledging the valuable parts, but due to limited time and energy, I'll only be mentioning the following parts. I hope it's constructive though!
This article ( https://medicalxpress.com/news/2021-02-brain-caffeine-utero.html ) claims "Investigators analyzed brain scans of more than 9,000 nine and ten-year-old participants in the ABCD study. They found clear changes in how the white matter tracks—which form connections between brain regions—were organized in children whose mothers reported they consumed caffeine during pregnancy." I haven't looked into this / am not personally knowledgable about it, but the uncertainty has made me hesitant towards caffeine in utero, during breastfeeding, and even the young toddler phase. Although it's quite possible that if this is even is an example of causation and not correlation, that the effect only happens with caffeine exposure that is earlier in life than the breastfeeding stage.
White noise may be harmful for babies' development as well. Andrew Huberman, a neuroscientist and general science/health communicator who runs a lab at Stanford talks about it here ( https://www.youtube.com/watch?v=JVRyzYB9JSY&t=3350s ) Andrew isn't always the highest signal to noise ratio in terms of true information, but high enough that I have found a lot of value from his podcast especially as a source for leads for further investigation. Again, I haven't looked into any of this / don't have high confidence one way or another.
With keeping the temperature above 70 F, beware of faulty space heaters (if you look at Amazon reviews for space heaters, there are many pictures of burnt marks on electric outlets, etc), and also high (abnormally high?) temperatures are supposedly/apparently correlated with SIDS.
Has anyone done analysis as to what extent opportunity exists for people who take the standard deduction / don't itemize deductions on their taxes? Or is itemizing deductions a pretty strong requirement? If there is any, albeit reduced, opportunity available, and if it's worth it in terms of time investment required, I'd love to participate.
Sorry if this has been covered already. I haven't invested the time to do much more than skim the post, but I think having this question answered without myself and others having to read the whole post thoroughly will have value.
These forecasts seemed much more bearish than I was expecting, and I think the reason why is because I believe these forecasts were made before ChatGPT came out.
Can someone give me the TLDR on the implications of these results in light of the fact that Samotsvety's group seemingly/perhaps had much higher odds for AI catastrophe? I didn't read the exact definitions they used for catastrophe, but:
Samotsvety's group (n=13) gave "What's your probability of misaligned AI takeover by 2100, barring pre-APS-AI catastrophe?" at 25%
(source https://forum.effectivealtruism.org/posts/EG9xDM8YRz4JN4wMN/samotsvety-s-ai-risk-forecasts)
Whereas XPT gave
"AI Catastrophic risk (>10% of humans die within 5 years)" for year 2100 at 2.13%
Without having read the exact definitions for "misaligned AI takeover" and still knowing that Samotsvety's prediction was conditional on pre-APS-AI catastrophe not happening, this still seems like a very large discrepancy. I know that Samotsvety's group was a much smaller n. n=13 vs n=88. How much weight should we give to Samotsvety's group's other predictions on AI timelines given the discrepancy in the risk prediction likelihoods?
Thanks for writing this 🙂 I found value in it. I'd like to write more acknowledging the valuable parts, but due to limited time and energy, I'll only be mentioning the following parts. I hope it's constructive though!
This article ( https://medicalxpress.com/news/2021-02-brain-caffeine-utero.html ) claims "Investigators analyzed brain scans of more than 9,000 nine and ten-year-old participants in the ABCD study. They found clear changes in how the white matter tracks—which form connections between brain regions—were organized in children whose mothers reported they consumed caffeine during pregnancy." I haven't looked into this / am not personally knowledgable about it, but the uncertainty has made me hesitant towards caffeine in utero, during breastfeeding, and even the young toddler phase. Although it's quite possible that if this is even is an example of causation and not correlation, that the effect only happens with caffeine exposure that is earlier in life than the breastfeeding stage.
White noise may be harmful for babies' development as well. Andrew Huberman, a neuroscientist and general science/health communicator who runs a lab at Stanford talks about it here ( https://www.youtube.com/watch?v=JVRyzYB9JSY&t=3350s ) Andrew isn't always the highest signal to noise ratio in terms of true information, but high enough that I have found a lot of value from his podcast especially as a source for leads for further investigation. Again, I haven't looked into any of this / don't have high confidence one way or another.
With keeping the temperature above 70 F, beware of faulty space heaters (if you look at Amazon reviews for space heaters, there are many pictures of burnt marks on electric outlets, etc), and also high (abnormally high?) temperatures are supposedly/apparently correlated with SIDS.
I'm not familiar with Emily Oster's works, or very knowledgable about this subject, but this Redditor seems to be deep in the subject of the scientific literature around daycare, and he/she disagree's with Emily Oster's opinions on daycare/childcare recommendations. See https://www.reddit.com/r/ScienceBasedParenting/comments/mz1bp0/on_cribsheet_on_childcare/ and also https://www.reddit.com/r/ScienceBasedParenting/comments/n3u548/notes_on_the_science_of_childcare/ and then he has also made a post on the slatestarcodex subreddit which I haven't linked to.
Has anyone done analysis as to what extent opportunity exists for people who take the standard deduction / don't itemize deductions on their taxes? Or is itemizing deductions a pretty strong requirement? If there is any, albeit reduced, opportunity available, and if it's worth it in terms of time investment required, I'd love to participate.
Sorry if this has been covered already. I haven't invested the time to do much more than skim the post, but I think having this question answered without myself and others having to read the whole post thoroughly will have value.