Opens profile photo
Follow
dynomight
@dynomight7
{{🦕,🧗}{🤔,💪},🧨} acetone👍aspartame👍alcohol👎correlations👎ethylene👍gratitude👎humidifiers👎lists👍mbti👍p2p meth👍potatoes👍pm2.5👎ubi👎wirecutter👎
dynomight.netJoined May 2020

dynomight’s Tweets

Also: Why are there so many reports of weird stuff flying around in the sky?
  • Alien aircraft
    5.7%
  • Govt. conspiracy
    11.4%
  • Other known explanation
    37.1%
  • Unknown, but not aliens
    45.7%
35 votesFinal results
Show this thread
Given this, I maintain that it's very unlikely that someone could walk enough on vacation to lose a significant amount of fat. But it's totally plausible that you could burn off your glycogen and create "illusory" weight loss.
1
Show this thread
Trying to understand the "I lost weight on vacation" issue, I thought it might be useful to have a "glycogen explainer" with all the necessary numbers in a convenient place.
Quote Tweet
A study is a great idea. But let me pre-register my doubt that it's the walking. Arithmetic is good: - Walking for an hour burns ~100 calories. - 1 lb of fat is 3500 calories. - Even if you walk an extra 4 hours a day that's only 3500/(4*7*100) = 0.8 lb of fat per week. twitter.com/mold_time/stat…
Show this thread
1
3
Show this thread
2. You deplete your glycogen stores before fat. Since glycogen is stored with lots of water this means that small caloric deficits can lead to "fake" weight loss, and people might just be getting confused by this. (h/t )
1
5
Show this thread
Two caveats! 1. The number of calories burned while walking is surprisingly uncertain. 100/hour does indeed seem to be on the lower side. (I figure most people aren't speedwalking everywhere on vacation, but adjust as you see fit.)
1
5
Show this thread
A study is a great idea. But let me pre-register my doubt that it's the walking. Arithmetic is good: - Walking for an hour burns ~100 calories. - 1 lb of fat is 3500 calories. - Even if you walk an extra 4 hours a day that's only 3500/(4*7*100) = 0.8 lb of fat per week.
Quote Tweet
ok at some point we need to put together a study to systematically collect all the "I lost weight on a vacation" data points twitter.com/elchefe/status…
Show this thread
15
31
Show this thread
Is there anyone who claims to "solve" the hard problem of consciousness? (As opposed to disputing the idea that the problem exists.)
2
3
Square profile picture
This is a story about science, academia, bureaucratic maneuvering, ambition, politics, capitalism, the “deep state,” secret emails, and slippery ethical slopes. And also alcohol. Read in the latest issue of Asterisk.
1
12
Interested in being a beta reader? I'm writing a thing about LLMs, data scaling, compute scaling, how much data is in the world, etc. If you'd like to help how much I embarrass myself, let me know. You should know AI and/or be good at "applied arithmetic".
Image
5
9
AI experts: Do we know how close current large language models are to overfitting? That is, if we took GPT-3 or PaLM and scaled them up, do we expect big benefits are possible without also scaling the training data?
1
1
Say you're talking to someone (either 1 on 1 or in a small group) and someone proposes an idea. You're pretty sure the idea has major downsides they aren't considering. When this happens in real life, how often do you bring up those downsides?
  • [0% - 25%)
    10.4%
  • [25% - 50%)
    29.9%
  • [50% - 75%)
    34.3%
  • [75% - 100%]
    25.4%
67 votesFinal results
My prediction is that there would be a long arms race of LLM-detectors and LLM-avoiders. Even if the avoiders win the race (plausible) it seems likely this that this would make the final text "worse" than an LLM that doesn't have to worry about being detected.
Quote Tweet
Am I missing something, or are all these attempts at recognising LLM outputs obviously destined to fail? It's dramatically easier to train an LLM for rewording a text than creating the text in the first place; then add a loss func that incorporates detection avoidance. twitter.com/mezaoptimizer/…
Show this thread
1
1
Well done everyone! Though I think I should have left Michael Jackson out of this poll. Anyway, I love AC/DC but who would have thought that their top album (Back in Black) has sold 1.6x as many copies as the Beatles' top album (White Album)?
Quote Tweet
Without cheating, guess the ranking of how many copies were sold of the top-selling album from each of these artists A) AC/DC B) The Beatles C) Michael Jackson D) Bruce Springsteen
Show this poll
2
2
Without cheating, guess the ranking of how many copies were sold of the top-selling album from each of these artists A) AC/DC B) The Beatles C) Michael Jackson D) Bruce Springsteen
  • A > B > C > D
    9.9%
  • D > C > B > A
    19.1%
  • B > D > A > C
    24.4%
  • C > A > D > B
    46.6%
131 votesFinal results
1
5