bfinn

Former tech entrepreneur (co-founder of the music software company Sibelius). Among other things I now play the stock market, write software to predict it, and occasionally advise tech startups. I have degrees in philosophy.

Wiki Contributions

Comments

Christiano, Cotra, and Yudkowsky on AI progress

On a detail:

what would the chess graph look like if it was measuring pawn handicaps?

I figured out from a paper a while back (sorry, can't recall where!) that 1 pawn = 100 elo points, at least at high levels of play. Grandmaster Larry Kaufman suggests the elo value e.g. of a knight handicap varies with the playing level:

https://en.wikipedia.org/wiki/Handicap_(chess)#Rating_equivalent

In Defence of Optimizing Routine Tasks

It would be handy to provide a list of tasks that many people do often and that can be optimized (and how to) - i.e. normal everyday things, rather than programming. Particularly ones relevant to this post, i.e. where optimizing seems like more trouble than it's worth. (Examples of the former abound - eg buying ready meals or Deliveroo instead of cooking, paying someone to do some of your admin. No good examples of the latter occur to me right now, but there must be some.)

Optimizing also includes dropping. In particular, dropping = not doing something at all, not because it's of negative value, but because it's lower value than other things you could do, i.e. a poor use of your time, and not easily outsourced. There may be things you could & should drop even though doing so involves time/effort and keeping going involves little.

The Power of Pomodoros

Cf when I first went on a diet some years ago, which worked spectacularly well, the main outcome (other than losing weight) was I learned to notice when I was full and didn't need to eat more.

The Power of Pomodoros

Though it would be odd (and require a strong explanation) if no productivity techniques do work. Which is like saying, no medicines really work, it's all just placebo effect etc. Since productivity fails for particular reasons (e.g. procrastination), and presumably techniques (like medicines) can be designed to fix or at least mitigate those reasons.

Shoulder Advisors 101

Great post. Two comments:

The popular saying "What would Jesus do?" suggests many devout Christians use Jesus precisely as a shoulder advisor - no doubt frequently & with intense seriousness. Hence they may well have useful insights into the technique.

Also:

It's important to be clear that the experience of "hearing the voices" actually happens, in many people. This is not a metaphor, and it is not hyperbole or exaggeration. I'm not saying that people tend to hallucinate actual sounds—that probably would be schizophrenia.

I read a book about Jaynes' bicameral mind theory which mentioned research in the 1980s into shoulder advisers (in effect) of institutionalised paraplegics. It seems many paraplegics, who couldn't move since birth, and I think couldn't speak either, had frequent auditory hallucinations of a person, often known to them (eg a parent), who would regularly talk to them, providing commentary, advice, comfort or criticism. They would hear this as a seemingly real, external voice (but not visible). Similar I assume to schizophrenics hearing voices. The book gave this as evidence of remnants of the bicameral mind in modern humans.

Can you control the past?

This is probably too trivial a point to mention, but FWIW:

“Try, for example, to make the two whiteboards different. Imagine that you’ll get ten million dollars if you succeed. It doesn’t matter: you’ll fail. Your most whimsical impulse, your most intricate mental acrobatics, your special-est snowflake self, will never suffice”

I know you specified that both AIs are internally deterministic and have identical inputs (rooms, whiteboards etc), but if I were them I’d try and seek out some indeterminacy elsewhere. Eg go out of the door, get some dice (or quantum random number generator), and toss them to decide what to write on the whiteboard, and thereby get two different results. Or just draw the night sky on the whiteboard (which will differ if they’re light-years apart); note this doesn’t require any indeterminacy in the universe, merely a lack of universal symmetry to break the Twin Earth setup.

To which you’d respond, well let’s say the AI can’t do that (it can’t move, etc.)

(And ex hypothesi it can’t create indeterminacy within the room, eg by tossing the board marker, because the rooms are identically set up, including air molecules etc.)

ADDED: though maybe it could say something Basilisk-like that would (or might sometimes) persuade someone else to provide an indeterministic or night-sky-like input. But if its only means of communication or action is writing on the whiteboard, and only the first thing it writes counts (or it can only write once), that wouldn’t work.

A Guide for Productivity

Thanks for this. Having read many productivity books I can confirm this is an excellent summary (and better than reading most of them). And includes some useful points I hadn't seen elsewhere (eg the idea of evaluating every subtask of a project after doing it, not just the whole project). So I hope more people will read it.

A few comments arising:

  1. You can also of course do a daily & weekly evaluation (possibly combined with gratitude for things that went well - as recommended to increase happiness). Also monthly & annual ones - the latter maybe lasting a whole day or two! At some point I will write up all my thoughts on how to do this.
  2. I strongly recommend the books of Mark Forster. Not so well-known (not apparently being a self-publicist, or in it for the money!), but a very good thinker with many apparently original ideas (not all of which I agree with). His methods have developed over time, so his latest book Secrets of Productive People is probably the best to read, though IIRC it doesn't include everything from his previous books, and goes into less depth on the reasoning behind his ideas.
  3. Re prioritisation, in addition to importance, timing and deadlines need to be considered. It isn't always optimal to do the most important thing first: if that can wait, but there is a less important thing with a tight deadline, that should usually be done first. So importance & timing need to be balanced, in ways that aren't always obvious. I wrote more about this here
  4. Re delegating, the consideration isn't just your area of expertise. That may be right for work, but as you should maximize leisure productivity too (as you point out), you can delegate some leisure things and do others even though you're not expert at them. (Eg sport, cooking, playing an instrument, childcare.) I think the correct consideration for whether to delegate is what is best use of your time - i.e. delegate things that aren't. In work the best use of your time is probably doing things you're expert at; but in leisure it's things like what you enjoy. So if you really enjoy cooking dinner for friends, and have time to, it can make sense to cook dinner yourself rather than delegating it to Deliveroo. Again, I wrote more about this in the above link.
  5. Maybe the reason your post hasn't had more upvotes is it's so long, so hasn't been read. I suggest you could produce a much shorter version, almost a listicle of tips with less explanation (and link to the full version for those who are interested). I know the Internet is full of junk productivity listicles, but yours would be a very good one.
A Better Time until Sunburn Calculator

Interestingly there's a pill you can take which has sun protection effects - polypodium leucotomos extract. It reduces UV damage, equivalent to about SPF 4. So not a replacement for sunscreen, but a useful addition to it, as it protects spots you may have missed. Eg my girlfriend goes on long daily runs, for which it's hard to cover all exposed skin adequately with sunscreen.

It doesn't seem to be widely available, but I get it from Super Smart (order online from Portugal).

The recommended dosage is 1 x 500 mg first thing in the morning (as it takes 30-120 minutes to take effect). If you'll be out in the sun all day, maybe take another one mid-morning. And preferably also take one the day before sun exposure - so not a bad idea to take one (or two) daily all summer.

The lowdown on scientific research into it is here: https://examine.com/supplements/polypodium-leucotomos/ This seems somewhat tentative, but in a more detailed (?subscription only) report elsewhere on the site they recommend it quite strongly.

Being the (Pareto) Best in the World

A v interesting analysis. That diagram is quite illuminating.

Recent editions of the bestselling careers advice book What Colour Is Your Parachute? have a long exercise on identifying your skills and brainstorming niche careers located at their intersection. Presumably there's a certain amount to be said from economic theory about careers that aren't at the Pareto frontier but are nonetheless near it. And possibly heuristics about identifying them, i.e. promising regions.

Me me me reading your analysis I realized how well it applies to me. I was a good programmer as a teenager, and a competent classical musician (though could barely have made a career of it as it's so competitive). And had absorbed entrepreneurial know-how from my parents. So I was above average, but not stellar, at all three. But (unplanned) I put them together to be a music notation software entrepreneur - of which there are only a handful in the world. Yet a niche big enough to be very successful if you're the best of that handful.

DeepMind: Generally capable agents emerge from open-ended play

The game world could include a sign stating the rules/objective of the game in English. Presumably then (perhaps helped by GPT-3) the agents might well learn to find & read it, as the best shortcut to working out what to do. And figure out both the meanings of individual words (nouns, adjectives, verbs) and grammar (word order, conjunctions, etc.) so as to understand whole sentences.*

And thus make the leap from GPT-3 as clever spinner of meaningless symbols like R, E and D to those symbols having a real-world meaning like 'RED'. Or even 'Keep the red cube within line of sight of the yellow pyramid, but don't let the blue agent see either of them'.

Hence leaping from Chinese Room to something more mind-like?

(*Though they might need help, e.g. a pre-coded 'language instinct', as trial and error may take a v long time to figure these out.)

Load More