Sunday, July 10, 2016

My Perspective on Prisma: It's Burning Money Fast

For the past couple of weeks - the photo-processing app Prisma has been gaining huge attention, and rightfully so - they use CNNs (+3 points for trendy buzz word) and AI (+2.5 points) to process its users' photos and turn them into truly impressive art work, inspired by great artists such as Van Gogh, Picasso, and others (+1.5 points for name dropping).

Here's what it looks like on the famous Success Kid's image:

 Which is it pretty impressive. As usual with these things, well lit highly contrasted photos work best, and it's loads of fun to play with.

Since the app is free, it begs the question: how do these people plan to make money?
Well, here are a few things to consider:

  1. Some of their filters are branded (e.g. by Palmolive), thus generating revenue through advertising.
  2. Since yesterday, the photos come with a water mark on their bottom-right corner, which I guess won't be there in some premium paid version.
  3. As a friend of mine pointed out: it is very hard to do these computations on the device, and as he found out - indeed when you put your device into airplane mode - the app doesn't work, so the heavy lifting must be done in the cloud. To quote my friend: "this is one free app with a huge AWS bill".
  4. In its Product Hunt thread, one of the Co-founders responded to an inquiry about their business model by promising that the app will always be free:

Conclusion: smells like trouble

So while this company is obviously putting out a good product (especially if we ignore the annoying 2-3 seconds wait when applying a filter), I believe balancing their cloud bill with revenue from advertisers seems like a big challenge. Of course, they may be aiming at getting bought quickly by the big guys, but the model of allowing users to heavily drain cloud resources, with a limited ability to monetize (unlike search results, social networks feed, etc.), may mean they won't live long enough for this to happen.

Friday, July 1, 2016

Deliberate Learning: 3 Tips for Acquiring New Skills

Recently, through Freakonomics, I came across the notion of deliberate practice. The idea, originally introduced by Ericsson et al in 1993, and more notably popularized by Malcolm Gladwell's "Outliers", is that the contribution of natural talent to greatness in art, science, and sports, is far smaller than we tend to assume.
The tl;dr version of it is: the thing that differentiates the greatest among us from the rest is not their talent, but the manner in which they practice. Specifically, they practice an awful lot, and do so in a reflective process that efficiently translates the invested time to a consistent improvement in performance.

Mozart - a guy who was great at something
 While I don't expect to become the world's greatest anything, some principles of deliberate practice struck me as familiar, and made me think about the way that I learn. I like learning new things, and having done so in a relatively wide variety of areas (painting, music, education, math, algorithms, poetry, design), I wanted to share what works for me when I come across new stuff to learn. To clarify, this is not deliberate practice. Deliberate practice is about becoming great at something over a long period of time. I want to focus on something much more common and useful for me: becoming reasonably proficient at something quickly.
Let's call it deliberate learning.

1. Choose a mentor and follow blindly, for a limited time

For a limited time, blindly follow you will
This seems to me particularly useful, while being quite challenging for anyone with a critical mind. It's also very obvious and most people do it to some extent, but it's a good idea to do so consciously as it allows us to switch it on when we need to learn something, whereas the natural tendency, especially with age, is to do it less and less, mostly due to ego.
 The idea is to find a teacher, who may be someone teaching you a course at the university, a colleague at work, or even your superior. This person should be very experienced and much better than you at what you're trying to learn, and should be available to you on a regular basis. You should buy into this person's point of view on the subject you're trying to learn: imitate this person's methods, ask for advice, ask questions, and just believe everything they tell you (professionally). For the limited time that you're adopting this person as your mentor, you should also suspend your disbelief about their professional shortcomings. It is a surprisingly powerful approach, as it let's you really run with ideas and concepts along a path someone else already paved, and with their help. After the limited time has passed (e.g., the university term), you can reevaluate all the stuff you used to embrace, and gradually throw away whatever you disagree with, or adapt it to your own style. One last point: this mentor doesn't have to be a person - it may even be a book.

2. Steer toward the boring and the difficult

Boring and difficult? Do go on!
Like the mentor thing, this is about suspending the self for a while. After you finish your learning period on the subject (I know the common wisdom is one should always be learning, but very rarely can one maintain a state of learning about everything) you'll naturally stay away from the stuff that bores you or that you don't understand well. This means that now is the only time you'll have to gain some ground in these areas. This being a learning period, you're also relatively energetic, and can muster the courage to face those areas, so it is very literally now or never. So be on the lookout for boring or difficult topics, and fearlessly dive into them. Your environment (at work or school) knows you're just learning, so it will be more forgiving to failure, which is inevitable when trying new and difficult things.

3. Get people whose judgement you trust to criticize your work frequently

My work sucks? I demand a trial by combat!
This is obviously necessary, and naturally hard on the ego, so encourage yourself with the following as you go:
- Paraphrasing Fight Club - you are not your work, so your work can and will suck sometimes, which doesn't mean you, in general, do.
- Going from terrible to okay is way easier than going from good to great, so if your work is shitty, this is good news as it means with little work you can improve a lot.
- Crummy work counts for experience too, so it wasn't a waste of time.