inputcoffee

Arbitrage opp:

Write a program to take the opposite side of any bet that you can hedge in the public market.

For example: if someone bets against Tesla, and you can take the opposite side, but (more) cheaply offset your risk in the option market.

You are wise not to. Even with the help its a long way from years of grad school:

http://www.slate.com/articles/life/culturebox/2012/02/the_my...

I agree with your description of he dynamic at play. It raises two questions:

1. Is the money that the startup attracts responsible for its success? In other words if a mediocre company goes through Y Combinator and then attracts a $55 million round, is it more likely to succeed than a great company that does not? (Let’s day the mediocre company doesn’t squander the cash wastefully but slowly looks for the product market fit)

2. Are there fundamentals that can be distinguished from an “observer effect.” Suppose everyone believes that a company coming out of Stanford is more likely to succeed than one coming out of (say) Babson. Does believing it make it true because the company attracts more money in each round?

These two thoughts are variations on a theme of the role of signaling in picking out fundamentals.

Edit: I should also point out that GV might also use “true” fundamentals like search results, trends, etc

The part of this logical chain that I don’t follow is the low cost of living attracting companies.

I can see why low cost of living ought to attract companies but in practice the new companies seem to gravitate towards already high cost of living cities.

This may be because the other considerations (workers, education, infrastructure etc) May outweigh low cost of living alone. Or there might be another reason I cannot think of.

The article has uncovered sources that claim the algorithm makes the ultimate decision and other sources that claim it doesn’t.

If it does make the ultimate decision and not just for political reasons then this is very interesting. Having input data that is sufficiently informative is important on a number of levels. Firstly this means that it is possible to pick winners on the basis of other VCs etc Secondly it means one doesn’t have to be personally concerned with the story if others can vet it for you.

If the machine doesn’t make the ultimate decision then — as others point out — it’s just old fashioned screens and checklists with a new interface.

Great question.

In theory, jobs put an upper limit on housing usage. People almost always follow jobs.

There are many beautiful cities in the world but only SF is near Silicon Valley.

You can find cheap housing in nice areas if they don’t depend on jobs (college towns, retirement communities).

People will present counter examples at the margins of course, but generally the relationship holds.

You misunderstand the sense of "true" here.

The point is not that the "true" price in the sense of what it ought to be. It is the "true" price in the sense of what the market is actually paying for the service.

When people put out the number $23, they compare it with the cost of power and wonder why, if it is so much lower, it doesn't already swamp the current solution.

So the purpose of the comment is to explain that it is not a complete solution, and not reflective of the total ("true") bill.

Yes, and any estimate of solar only costs is composed largely of financing charges which is another way of saying capital costs. These include the installation.

That is true, and existing infrastructure does (mostly) handle the gaps.

How much does "existing infrastructure" charge to handle the gaps? This calculation differs wildly from as low as $6 to as high as $15, but I think $12 is the right answer.

So add $12 to solar only as a rough rule of thumb. Note that this $12 still comes with transmission issues, and environmental impact.

The true price of solar is the price to produce reliable, continuous energy. That requires something to smooth out the signal and store the energy, most probably a battery.

The solar + battery price is also going down, though not as fast as the solar only piece. But that's the metric we should be following. When that hits $30 per MWh, it'll be game over.

This is great.

I wish we could do our own arbitrary style analysis on the data set sort of the way one can do a factor analysis on a portfolio.

I would look at words that are common between Marukami and McCarthy compared to the rest of the corpus for instance.

Not sure why you're getting down-voted. You're right, the author didn't close the loop. Typically we would expect to see an insight after you apply the ML technique.

So they extracted features, clustered the pages and found... what?

I am sure they learned something but it might be proprietary.

You need a bank account or credit card for paypal.

I guess you can pay for this anonymously.

This looks interesting. Is it basically a hosting site you can pay for with crypto?

This makes other things possible: anyone can pay for it using crypto etc?

I can imagine some use cases, but since you've been thinking about this a lot longer, can you say a little about:

1. Use cases

2. Long term vision for different paths along which you might see this evolve?

Thanks in advance

I concede the point: the "best" explanation is best in a context.

The best explanation for a PhD candidate about to defend their thesis the next day is not the same as it is for, say, me.

But that reinforces the point I am making.

This was a pre-internet book though.

The internet has made explanations hyper competitive. At this point I can find dozens -- perhaps hundreds -- of explanations of Godel's theorem.

But in the pre-internet days I could go to the library, or have my friend Kent explain it to me. (He had to explain first that languages were either complete or true -- he called it sound if I recall. soundness was the property -- and then it was demonstrated that these formal languages were sound, therefore incomplete.)

Back then you couldn't just google Godel and say "oh, that reminds me of music returning to the tonic" or whatever and then google that.

I think because a lot of the ideas have sort of spread out into culture, and been re-explained over the years, the best explanation are not in those books anymore.

Actually, with Tensorflow as a Keras backend, I would expect them to be the same. I am not sure where the performance difference between TF and TF as a backend come from.

I do think that pure TF would be easier to scale up over multiple servers etc. but that's only because I don't know how it would work in Keras. Maybe its easy.

I appreciate how you focus on just two, which are the state of the art in your opinion.

I find it less useful to see comparisons of "top 50 deep learning frameworks for 2018" which include esoteric stuff that is only there for sake of completeness.

This way a person branching out from Tensorflow (I assume its Tensorflow) knows which two frameworks to try out, and what to look for.

As others are pointing out, TF isn't that hard.

Or, rather, it is hard but the difficulty is from getting an intuition for what part of this weird multi layer net is producing this weird behavior and is it an artefact or something interesting, and is the connectivity complete and is should I change the learning rate and activation functions?

The real reason to use Tensorflow is the same reason you might use a Go framework instead of Rails: in your heart you have this hope that this thing will one day grow into a really large project and support lots of people and that will be easier with this scalable, optimized code.

Its not even that you'll hit Google scale, its that you'll hit popular scale and still serve the whole thing out of your Digital Ocean droplet.

Great Viz. Although my favorite is still the video of Andrej Karpathy's Stanford lecture explaining it.

He sort of goes through some other implications which, from an intuition massaging pov, is great.

https://www.youtube.com/watch?v=i94OvYb6noo

For those who are impressed by such things -- as I am -- he is now head of AI or ML or something at Tesla.

I half agree with you. The notation alone, although "understandable" doesn't provide a "deep" understanding.

I think the intuition is only half-transmitted with just the notation.

I do think working through an example sort of completes the intuition.

There are probably others who get frustrated, but I don't think most people do.

I think its fine, and I haven't heard others complain about it over the years.

One of the interesting aspects of Foucault's take on power was that it was constitutive.

To put it another way, the company that reclassifies what it does, or uses a new taxonomy may actually be trying to change itself by doing so and may succeed.

The big difference between the taxonomy a company constructs, and the kind that historians construct, of course is that companies have a real measurable output -- profit -- that they have to achieve.

Speaking as a newbie (python and R mostly) who just wants the blogging software to blog, I use Jekyll because it’s the default choice. Yes there are python options but I don’t want to program the software - my programming energy is already spoken for.

I don’t care about the speed differential.

However if Hugo is actually easier somehow, well that would be tempting.

I mean that in the generic sense when you grow a project (I use python mostly) and run into issues invariably with the size and complexity.

As a hobbyist I love seeing new languages. I look at their features and a few simple examples and immediately imagine they'll solve all the issues I have faced with other languages.

I generally appreciate the massive amount work that goes into something like this.

But when I want to work on a project, I am wary of using something new. I just know in my heart I'll set the wrong flag on the compiler and the binary will end up in a directory I can't find and it'll take me two days to figure out how to do it.

My question is this: if the main strength of this language is that uses Swift-like syntax, and lets you develop native apps in Android and iOS, why not just make an Android compiler for Swift?

The author’s claim only applies to weak form of market efficiency.

The original Fama paper distinguishes between three forms of the EMH: weak, semi strong and strong.

The weak form only alleges that historical price info is fully incorporated in the current price. You can still make money by working hard and figuring things out from outside the price history.

The weak form applies to purely technical trading that extracts value from price history like momentum and the simple moving averages cross over studies you see.

It is certainly an interesting problem. I don't know that I have the solution.

I agree with everything you said about the problem around finding a good reward.

However, coffee drinkers can be picky. It would not motivate me, but then again I am not sure what would other than the benefit of doing my friend and the customer a favor.

That’s compelling as a demonstration of Julia’s expressiveness.

My next thought is: what would it look like in Python?

With and without a deep learning library.

I suspect it would be much shorter with and slightly longer without but I am not certain.

It sounds like these companies are providing frequency modulation, and smoothing our the renewable power signals with the batteries.

So the user still enjoys - and pays for - the storage component while the utility enjoys and pays for the frequency reg.