Profile cover photo
Profile photo
Gordon Irlam
55 followers
55 followers
About
Gordon's posts

Post has attachment

Post has attachment
I'm slowly getting up to speed on AI. I read a book on neuroscience, and have now started a book on deep learning. Meanwhile the Whitehouse has issued a request for information on AI. I am planning to submit some comments on neuromorphic computing, and I am looking for feedback on my draft comments from all of you before I do.

Post has attachment
I compared 11 different asset allocation calculators on the web. For roughly the same scenario I found the recommended asset allocation ranged from 35% to 100% stocks. Caveat emptor.

Post has attachment
AACalc is a new open source asset allocation calculator that I wrote. My old asset allocator was slow, complex, and precise. This one is fast, simple, and approximate. It takes into account Social Security benefits, and produces high and low estimates as a result of uncertainties in the equity risk premium.

Post has attachment
I wrote a paper exploring asset allocation's little secret: the standard error of the equity risk premium is so large that a 95% confidence interval spans almost the entire range of possible asset allocations. This is not saying asset allocation is uncertain as the returns you experience are uncertain because of your limited lifetime. Rather, there is intrinsic uncertainty as to what returns you will experience due to the limited historical record of returns.

Post has attachment
I wrote an open source pricing tool for income annuities. It uses actuarial life tables and the latest yield curve from the Fed to compute fair SPIA and DIA prices. Using it I find Money's Worth Ratios for actual quotes are typically close to 100%.

Post has attachment
It is not for the faint of heart, but I have just released the source code to my AACalc asset allocator and consumption planner under an open source GNU Affero license. Briefly, this is a retirement planning stochastic dynamic programming tool that I have been working on for the past few years. Unlike Monte Carlo simulation, which takes a strategy and works forward in time to see how well it performs, stochastic dynamic programming works backwards in time to compute the optimal strategy to be employed.

Post has attachment
After a 2-3 year absence I've finally updated my Back of the Envelope Guide to Philanthropy website. AI risk remains the top issue. New evaluations: farm animal welfare, GiveWell, urban land reform, sleep research, and political donations. Notable changes: OneWorld Health appears to have largely failed taking $180m with it. New cross cutting pages on: global catastrophic risks, lobbying, and international development lobbying.

Post has attachment
I've developed an actuarially fair SPIA pricing calculator. I am finding the SPIA quotes I have received are quite competitive, with Money's Worth Ratios in the 90-100% range when measured against Treasury bonds.

I was confused by the "+1" button in Google Plus. I just spent an hour trying to figure it out. For anyone else similarly confused here is what I discovered.

+1 is NOT used for a recommender system to indicate "show me more postings like this".

+1 is primarily used as a simple means of providing feedback to the author that you appreciated their posting. It is primarily just a counter.

However, it also has some aspects of the share button: "If the post was shared with people in your extended circles who also have you in their circles, they may also see your +1 in their streams". Extended circles are your "circle's circles", that is your friend's friends. I think this means if You and YourEnemy have a CommonFriend, and YourEnemy includes You in their circles, but You don't include them, YourEnemy will be able to see your +1 activity (at least on public posts). Google doesn't make it clear if YourEnemy will simply see the +1'd post, or if they will be able to see that You +1'd it.

Lastly. if you +1 a product or brand (what exactly defines these isn't clear) then Google may use your name and image in ads. This can be opted out. For me the default was opted out, but that may be because I use Google Apps for Business.

I find it all very confusing (and I find Facebook even more confusing). Why can't they separate out "thank", "share", and "endorse"?
Wait while more posts are being loaded