Disqus - Latest Comments for shanusmagnus

Re: Intermittent Fasting

Shane Hoversten — Mon, 21 Dec 2015 12:49:43 -0000

Try eating more fat for breakfast. That's a huge help for hunger, esp. saturated fat, if you're not afraid of it.

Re: Visualizing Representations: Deep Learning and Human Beings

Shane Hoversten — Thu, 22 Jan 2015 16:54:16 -0000

I am not the author, but I think your second point is very astute, and is an important amendment to the Brett Victor quote in the article. Even using tools certain phenomena are subject to restrictions -- transforming certain regions of the EM spectrum into other regions that human physiology can cope with is wonderful, but it is a process of transduction -- one thing into another, preserving structural integrity. Other representational issues mentioned in this article are something else.

For instance, using dimensionality reduction of whatever flavor can provide intuitions into a dataset too vast for humans to process, but only if the information inherent in the data is amenable to that reduction. Human visual cortex appears quite sensible in posterior regions that we somewhat understand, with certain parts (think 'hidden layers') coding for feature conjunctions that are reasonably explicable. But that's no guarantee that this will always be true, and in fact the current lack of understanding wrt more anterior regions of VC is an illustration of that. Whatever those networks are doing is beyond current understanding except in a very coarse way.

People mistake the generally sensible relationships between things in the world they know as some immutable law that things can always be reduced to the sensible. Tools bridge that, in the best case, by transforming into an isomorphic space; and in the worst case by throwing away information. Just as there are limits to what you can explain to a child, no matter how hard you work to transform the information into something graspable within the child's cognitive space, there will be limits to what our tools can render comprehensible to us.

Re: On Disowning Descendants

Shane Hoversten — Sun, 10 Mar 2013 10:02:05 -0000

Clearly this post is a response to something specific, as some commentors confirm. But what? Is there a representative link to illustrate?

Re: OMG rm -rf ~ in a valentine bash script and its partly my fault??!?! - William Edwards, Coder

Shane Hoversten — Wed, 13 Feb 2013 09:46:38 -0000

Beyond stupid -- like breaking into someone's house and shooting them and then bemoaning the state of a world in which someone can break into someone's house and shoot them.

Re: Do Not Pass This Way Again

Shane Hoversten — Sun, 27 Jan 2013 03:27:41 -0000

With all that, curious as to why you think some pretty heavy hitters, with deep pockets and some of the smartest people around on the payroll, choose to go with MySQL? Why is Google CloudSQL not developed on something else? It doesn't seem like the 'bad reasons' you give really apply. Are the number of people familiar with MySQL vs. Postgres (or whatever) so much more numerous in the engineer population that it's worth using MySQL, warts and all?

Re: Life in a nutshell

Shane Hoversten — Wed, 12 Dec 2012 00:08:48 -0000

And a one-armed dude?

Re: Ruby Under a Microscope - Pat Shaughnessy

Shane Hoversten — Mon, 29 Oct 2012 19:57:18 -0000

That pretty much covers it I guess.

Re: Conducting your business redux

Shane Hoversten — Thu, 06 Sep 2012 10:51:18 -0000

If you'd like to create your own fictional universe using my characters, you can have them say whatever you like.

(Plus, I'm pretty sure that _is_ what he was writing.)

Re: Conducting your business redux

Shane Hoversten — Wed, 05 Sep 2012 10:39:35 -0000

"While conducting my stupid affairs today I encountered a salt-of-the-earth Midwestern type. As he was working quietly and politely next to me the only reason I noticed him was his profound handsomeness and a dazzling intellect that shone through even in the absence of speech."

Re: Several reasons for cautious optimism

Shane Hoversten — Wed, 05 Sep 2012 10:07:43 -0000

Also, I suspect that the bookstore is just a honey trap. If you walk in there you get stuck to the floor and die. This way the city eliminates undesirables. Going into the cafe probably just puts you on some kind of watch list.

Re: Several reasons for cautious optimism

Shane Hoversten — Mon, 03 Sep 2012 22:55:06 -0000

I get up that way alot, I'll text you next time. It will be nice to have an Elk River friend again.

Re: First humanities MOOC professors road-test Coursera's peer grading model

Shane Hoversten — Sun, 02 Sep 2012 21:35:04 -0000

I continue to be baffled by the tenor of comments criticizing an organization offering a non-trivial education on non-trivial topics which they will provide you in the comfort of your own home for zero money. Clearly there is room for improvement with how the content is delivered and how grading is performed. I'm sure the Coursera folks would admit to this.

But seriously, have some perspective - you're not actually entitled to anything from this relationship. There is no contract, you didn't pay for a service, and just like in every other part of life you may well have to put up with some things that are less than ideal, including internet trolls and assholes. If you can't deal with these inconveniences, in light of the benefits on offer, then you've got bigger problems than whatever you might be suffering with Coursera.

Re: The tragedy of macroeconomics

Shane Hoversten — Thu, 19 Jan 2012 13:15:12 -0000

One more thing on the health metaphor:

It may or may not be tough to say what caused your grandfather's heart attack. But what we can certainly say is that, for instance, it probably wasn't eating cholesterol and saturated fat, despite what 90% of doctors would have you believe. How do we know this? Because other populations, who eat absurd quantities of those things, have great heart health. It's not possible to make causal attributions with correlational data, but even this (relatively) weak evidence is enough to definitively refute the "saturated fat = universally bad" people, who somehow continue to dominate the mainstream cardiological universe. If you can blankly refute a claim, and yet the claim persists, that is a special kind of systemic dementia.

Anyway, I didn't mean to imply that complex systems that are amenable to scientific investigation are magically rendered simple; just that you can at least move definitively forward via the scientific method. The further you are from being able to a) measure accurately and b) perturb a system in the form of experimentation, the further you are from actually knowing anything substantial about it.

Re: The tragedy of macroeconomics

Shane Hoversten — Thu, 19 Jan 2012 11:33:44 -0000

It seems to me that it's difficult to know cause-effect in _any_ circumstance. IE: what caused the great depression? Or: is it better for the economy to save all your money, or to spend all of it? The answers to big questions tend to be superficial or just kick the can down the road. And even when they posit something particular, you can't actually test it. That standard of evidence and proof would be laughable in any other field besides English or History. In fact, macro econ would seem to be more aptly named "history, with equations."

Although I'm willing to be corrected. What, in your mind, is a non-trivial 'scientific' finding of macro econ? I don't think the austerity stuff qualifies, based on what I've seen.I also don't think the metaphor with health holds. There is certainly a lot of contention in the field (for instance, is butter good or bad?) But that's mostly political. There is, at base, a body of empirical research that tests the appropriate questions in a rigorous scientific manner, which has produced concrete and repeatable findings. It so happens that the health/nutrition field is terminally polluted by both lobbying groups and years of conventional wisdom, so findings that counter the orthodoxy take a long time to get any traction. But at least there are findings, and they are repeatable, and they're pretty straightforward for those with ears to hear.

Re: Harry Potter redux

Shane Hoversten — Tue, 13 Dec 2011 14:06:32 -0000

I think this a response to the general modern Western attitude towards children and reality, in which everything is sanitized and nobody should be pressured to achieve anything in particular. Do your best! We're all okay! Which was itself probably a reaction to something; and I guess I'd rather they wrote everyman heroes than works in which everybody is the lost Prince of Narnia or whatever. I guess the happy middle can be hard to find -- maybe those things don't sell.

Re: WWCHD?

Shane Hoversten — Fri, 02 Dec 2011 22:31:01 -0000

THTM? The Hot Tamale ??

Re: Spoilers Don’t Spoil Anything

Shane Hoversten — Thu, 18 Aug 2011 13:22:44 -0000

Reference?

Re: Messes

Shane Hoversten — Wed, 27 Jul 2011 17:55:21 -0000

This is what I want to know, though. There's the old stereotype of the anal-expulsive who's got his shit scattered everywhere in his office, in all these piles, and who gets enraged if someone cleans it up for him, ala "I have a system, and you've ruined it!" The thing is, I _am_ that old stereotype. In the midst of my chaos, I _do_ know where stuff is. But I still hate it. I'd rather it were clean, even though I can function, organizationally-speaking, in the chaos.

So I wonder if the people in the above-mentioned cliche really aren't bothered by the clutter, or if they have simply become functional and invested in their clutter? Having met a number of those types (and, as I said, being one myself) I can verify that in my experience they've always been the latter.

Re: TIMBERWOLVES: A New-Look Pack, Wolves Workout

Shane Hoversten — Sun, 26 Jun 2011 22:17:47 -0000

I love that in interviews Kevin Love actually sounds like a functional human being having a conversation, and not some drone mumbling the same tired crap. Good work.

Re: Multiplexing vs Multitasking

Shane Hoversten — Fri, 01 Apr 2011 18:33:07 -0000

I'm thinking mostly of special "hacks" in the machinery to allow for seemingly paradoxical effects: visual indexing, the popout effect, the own-name effect, to name three that seem to break what might be seen as a linear scaling relationship (more items to deal with leads to linear decay in ability to deal with them.) These aren't really 'exceptions' to the attentional model, as such, but they might be construed that way by someone trying to prove a point.

You posted today something about Shannon and information theory, which I think is a nice way of framing this 'multiplexing' idea, especially if you consider multiplexing in its technical sense, where you have N streams interleaved together, s.t. it takes N units of time to get 1 unit from each of the N streams. From an information-theoretic standpoint, the information density of the multiplexed approach is the same as the single-stream approach, but that's ignoring switching costs, which, again, have been shown many times to be substantial to the point of ruinous in prospective task-switchers.

Playing devil's advocate for a second, it would be interesting if someone were to show an information gain in the multiplexed approach, s.t. streams 1..3 multiplexed together produced non-linear results in the manner that language does, where a three word utterance, considered as a whole, conveys more information than each of those three words in isolation. Theoretically interesting, but the ground-laying required to defend such a claim would be elaborate. (Already in reading this comment I'm tempted to fill it out with hedges and justifications to make it precise enough to be defensible.)

Re: Multiplexing vs Multitasking

Shane Hoversten — Fri, 01 Apr 2011 15:20:10 -0000

Key Kevin, do you turn the radio down when you come to a complex intersection, or meet something unusual on your drive? I bet you do. Attention is limited, and executive control needs it. I know this literature pretty well, and there's just no way doing more stuff doesn't decay performance at any particular thing. There are some corner cases, certainly, but they're small corners.

Which is not to say that what is gained might not somehow be worth what is lost -- going to the Met and being overwhelmed by all the art within your field of view is sort of part of how I experience art, moreso than full cognitive resources deployed to one particular piece. But claiming that task-specific processing is not impeded is contrary to at least two decades of empirical work.

Re: Which Traits Predict Success? (The Importance of Grit)

Shane Hoversten — Sun, 20 Mar 2011 19:42:10 -0000

Thank you. Exactly.

Re: In Praise of the Segway

Shane Hoversten — Thu, 10 Mar 2011 19:00:24 -0000

I keep wondering what kind of urban environment will tip the Segway into mass adoption -- one can imagine a setting where large numbers of Segways are available for public rental, like Zip Cars, and where this is super helpful. But it hasn't happened yet in any cities I know. Probably too much density is bad (hard to maneuver the thing in crowds) and too little density is also bad (not enough people to support Segway rentals.) Is there a just-right point?

Re: Simultanology

Shane Hoversten — Thu, 10 Mar 2011 18:56:21 -0000

I have had a similar revelation, but with one twist: even in its most favorable environments, simultanology requires a degree of faith that the thing will continue to exist, which is almost true, but not quite always.

Further, sometimes the environment prevents the just-in-time access, and the Khan academy is a good example: sometimes I want to watch a video, but I'm not in a place with enough bandwidth to do it.

Someday these problems will go away, and that will be a Big Win (no more figuring out what podcasts I need before starting a long trip.) But it hasn't happened yet.

Re: 99 Cent Books

Shane Hoversten — Tue, 08 Mar 2011 15:37:27 -0000

I think the risk you're talking about here doesn't apply in the vast majority of cases, at least for fiction: if you're writing fiction, you're doing it because you have to do it, not because you expect to make a living. So I agree with you that the .99 is great for people who would write a book, anyway, but disagree that the people who have managed to be published by traditional means have different risk profiles.