Thesis Week 9 (Spring)

Work log:


6 minutes late, oops. Walking always takes more time than I expect. I feel energetic so I’m not starting a green tea cycle until tomorrow. I slept really well, despite waking up a few times in the middle of the night.

I need to turn in a homework assignment today, so I’m spending the whole work day on that instead of working on my thesis.

Okay the two server queueing is actually pretty tricky. Taking lunch break at 11:52 because I need time to think about this.

Back from lunch a bit late because I got caught up talking with the roommates. I think I know how to do N server queueing now.

Done! And on time too. I spent another 30 minutes mucking around with a GPU fan duct Taylor helped me make.


Three minutes late. Slightly better. I guess walking down the block and back takes like 8 minutes. Oh, I did have to look for a while for my front door key, I think one of the cats stole it and hid it in one of his many hordes.

I’m a bit at a loss on what to work on this week. I don’t have any feedback from Dr. Yampolskiy, maybe I could ask him for feedback.

Looking back at last week, I realize I need to work on the AI safety section. It’s just bullet points right now, there’s a lot more to write into it.

It’s 10:50, I read “Complex Value Systems are Required to Realize Valuable Futures”. Not super relevant, I’m really reading it to procrastinate writing. However, the concept of an “Outcome pump” made using time travel is especially interesting, and a great metaphor for a non-sentient optimization process.

Hmm ten minutes later and still not sure what to do. I had an idea though - start on the web portion of it. There’s a lot of programming and such to do there, might as well get a start on that while I’m not sure what else to do.

This is going well - I accidentally went 15 minutes into lunch.

Lunch break!

Back from lunch. Oof, I have a headache. No idea why - maybe lack of sleep? I did stay up later than usual last night. Sleep schedule maintenance is so hard. Anyway, back to coding.

This headache is pretty bad, I’m going to quit early and take a nap.

(Postscript: I ended up taking a nearly 3 hour nap, apparently I was sleep deprived.)


Left at 10 till 10 to have time to walk, only to find my imperfectfoods delivery, so I was 5 minutes late (but now I have lots of fresh vegetables in the fridge so it’s worth it).

Anyway, what to do today. I really need to come up with a system for giving suggestions based on risk factors presented. However, that might be completely impossible/pointless.

This is going to be the most important and hardest to defend section. I have to make sure it doesn’t seem like I’m just making up all this stuff. I’m going to start with just making stuff up, and see what I can develop it into or back up any of my points with actual research.

I managed to come up with a few recommendations based on inputs to the classification inputs. I’m tempted to go back into designing the website now, and add in these recommendations. Yeah, I’ll do that.

Done! It works pretty well, also It’s quitting time so bye!

TODO: better web design, hints for all the inputs


I skipped my walk, time got away from me this morning. Oh well. I guess I’ll keep working on coming up with ad-hoc conditional recommendations to start with. It feels mundane, but actually me a few months ago would be extremely proud to know I reached this point. The large amount of time between wanting and achieving makes it hard to feel like anything has happened, but so much has happened and I should be proud of what I’ve accomplished.

Actually, it’s time to expend my AI safety evaluation. It’s only a page and a half and it could be at least 3 pages.

Oops spent 30 minutes reading the preview to “The Most Human Human”, about the Turing test, after it was cited in this paper that I’m citing: I also leaned about “super-forecasters” who have an unusually good ability to predict the future. Little progress has been made but also I don’t entirely feel like I’ve been wasting time.

Time to break for lunch. I’m really feeling the oof today.

Hello, I’m back. I’m going to fill out these two subsections (Scalable Oversight and Safe Exploration), then send this draft to Dr. Yampolskiy as this weeks status update, with a request for suggestions. I really am in need of suggestions, my progress is slow right now and I don’t really know what to do each day.

This isn’t on topic, but two totally ideas I learned about just came together in my head: one, that there are millions of living things (bacteria, virus, pollen, spores, etc.) in every tiny volume of air around us, and two, bacteria (and all living things) have complex horizontal gene transfer and ecological interactions which some people personify as “Gaia”, the living thing that is the Earth’s ecosphere. (Sadly the second idea is unfalsifiable at our current understanding of life and intelligence). Anyway, taken together, this presents strange revelation, that with every breath I’m inhaling millions of living things, which are a tiny piece of the body of a god. Anyway, that kind of thinking will get you kicked out of academia (just kidding only serious) so I’ll leave it at that.

It’s 1:30, I’m going to go on a short walk and see if that’ll improve my ability to work. The work words are just not making their way onto the page today.

I’m back at 2:00, time to work now. I don’t think I feel all that different, really.

Despite being such a rough week mentally, I still managed to write about 700 words. Slow and steady doesn’t win the race but it does get you to the finish line in one piece. (Btw the tortoise and the hare is a story about the folly of arrogance, not the virtue of deliberateness).

Anyway, it’s almost time to quit for the week. Status update sent, time to update the blog and log off.