Mathematics PhD; algebraic geometry, data science and statistical learning theory.
https://kaleb.ruscitti.ca
kaleb
Karma: 139
Your categories are not essentially gendered, although I understand why we feel that way. For example, in your travel-packing example my wife would be considered rugged while I would be considered elegant, under your definitions. I also think that in traditional Chinese culture, both of your definitions would be considered masculine. (Sorry women, I guess you get nothing lol)
I also think that we apply these strategies unequally in different parts of our lives. I’d guess if you have to give a research talk at a conference, you’d take an ‘elegant’ approach of “let me prepare my talk well and try to anticipate possible questions the audience will have” instead of “let me do the minimal prep and then just power through any technical difficulties or difficult questions’.
Maybe our gender socialization leads us to favour different strategies in different situations along gendered lines?
In the leftist political sphere, this distinction is captured by the names “reformers” vs “revolutionaries”, and the argument about which approach to take has been going on forever.
I believe it would be worthwhile for us to look at some of those arguments and see if previous thinkers have new (to us) perspectives that can be informative about AI safety approaches.
I believe the implication of decrying something as populist is that it would not actually be good for the masses. So a populist promise is one that the masses want and think will be good for them, but in actuality would be bad for them. Then a populist politician is one that promises popular policies regardless of whether they will be net good or bad.
I think whether or not someone considers populism bad is downstream of whether or not they think the masses can accurately assess what would be good for them.
If I understand your presentation correctly, it sounds like you think this new era of FE will mostly be net negative. Is that right, or have I misunderstood you?
If you do think it will be net negative, do you feel like there is any action society should be taking to rein in the consequences? Your key takeaways section suggests that we should just brace for the wave.
I am a twenty-something mathematics researcher, so I am your target audience. Have you been making unprecedented work on your projects? Or can you point me to others that have? I am pretty happy to implement AI tools into my workflow, but I’d guess they speed up my output by only 1.25x at best. I’d be surprised to learn of people getting the 5x+ speedup you’re suggesting and if they are then I’d like to know how they do it.
Your screenshot looks like a desktop or laptop, do you ever manage your to-do from your phone? I’ve been trying various calendar and to-do systems for months, and my biggest issue is that I want to add things quickly on my phone and then manage them and do planning on my desktop, but I haven’t found a software for this that works frictionlessly enough.
Backing up the Oppenheimer quote with an anecdote:
I chat with my colleagues in the Pure Math department here about our motivations for doing research, and for almost all of them it just about getting paid to do something they enjoy, and they pick their research area based on beauty and enjoyment. Pure Math probably has a sampling bias towards this sort of person though.
Personally, I’ve found it increasingly distasteful over the last few years, as I have been thinking more about how I can impact the world with my research.
I added in a section about Benford’s law, a surprising functorial prior on first digits of numbers in randomly compiled data!
However, I have the impression that one critique of the paper at the time was that functorial seems to just encompass a bunch of known cases in statistics like equivariance and exchangability. It’s hard to cook up a natural example that isn’t covered in one of those cases.
Actually, unlike Celsius and Fahrenheit which are measured in degrees, Kelvin is just measured in Kelvin, an interesting quirk of the different historical circumstances around their definitions.
Some ~5 years ago I played a bit of Baba is You with a friend. I did Tiny Pond on my own before the workshop (I was Raemon for the purposes of our event), but I found it a bit too easy. In particular, I remembered how all the mechanics in the level worked, so my certainty about each step was very high. I collaborated with Jenn in the second round and felt similarly.
Even though the surprise anticipation part of the exercise was moot for me, I still think that the practice of trying to discover and write out the entire solution before solving it was worthwhile, and I plan to do it again on more Baba is You levels.Many of the others in the workshop felt that this exercise is artificial because ‘you’d never solve research questions like this’, you would always play around with things in advance. I think if I were to run the workshop again, I’d emphasize more how for some problems, playing around in advance is prohibitively expensive, so planning far in advance is important. I think your web version of the game that tracks development time is a nice step towards addressing this part of things.
On a similar note, in my experience doing research, a long-term plan might include cycles of experimentation and plan-making. Maybe you should have a section of the plan-making exercise where you set out 2-3 small experiments you want to run, running them, and then return to long-term plan-making? Or maybe that defeats the purpose of the surprise-anticipation portion?
Personally, I conjecture that as a math PhD student and puzzle game enjoyer, Baba is You is too similar to the exact type of training that I have had for years at this point. I want to try doing a plan-making and surprise anticipation exercise where I have to do something totally foreign to me, like making a chair out of wood!
Thanks Raemon for your effort putting this together.
Yes; I ordered a) before b) as I thought a) is an easier action, but that is subjective.
Shipping domestically via water is the most environmentally friendly option available.
Can someone elaborate on this? I would imagine that (electrified) rail is the most environmentally friendly option available. Or perhaps you consider rail to not be an available option?
Is one takeaway of your post that we should consider current safety research as more about training human researchers than about the actual knowledge obtained from the research?