Sunday Reads - Links I Found Interesting 6/16


PyTorch Hub is an awesome new step forward in research reproducibility for machine learning and artificial intelligence. Makes it super easy to publish your pre-trained research models and for others to download them and test them. I was shocked how uncommon it was for people to publish their models when I first started reading machine learning papers.

An enzymatic pathway in the human gut microbiome that converts A to universal O type blood

This was a metagenomic study that identified two enzymes that may help convert A and B blood to O increasing the supply of blood available.


This is a cool ArXiv access point with upvoting and down voting, ability to save notes, see twitter commentary on paper and more.

The Good Social Internet

Social media often sucks. The social internet is a magical place full of rich relationships, new connections, intriguing ideas, and true community.

What do I mean when I say the social internet? It is all of the internet designed for sharing, connecting, and engaging with other people. This may include social media, but it also includes personal blogs, forums, discord servers, random IRC channels. In short, places where people congregate around shared interests.

I have already conceded that social media can be an interesting part of that, so it is easy to wonder why it was separated out. The simple reason is that the dynamics of most social media are very different from the dynamics of other social internet applications. For one there seems to be a fundamental push vs pull difference in the way that you normally come to view the content.

Social media has a habit of pushing content on to you. This may be literally in the form of push notifications or more subtly via the magic of algorithmic infinite newsfeeds. You get to see more and more content, much of which you never would have chosen, all pushed on you. This does offer the opportunity for fortuitous discovery, but it also fundamentally changes our relationship with the internet.

In the old social internet you pulled the information you desired, rather than having content you may want pushed on you. If I wanted to see someone's recent thoughts I would navigate to their blog. If I wanted to discuss a topic with people I would go to that forum. If I wanted to stay in touch with people I would have to remember to reach out to them or drop in a channel or server where I knew they could be reached. The internet was on demand, instead of demanding.

Now perhaps you reject my premise that the internet is no longer on demand. Before you do so I want you to install RescueTime or a similar program on every computer you use, and combine the time from that with the Screen Time from your phone and add in the number of hours you spend watching Netflix or Hulu. Likely you will find that the majority of your waking hours are spent staring at a glowing rectangle. If you truly feel that each and every time you wanted that content, then you are the exception that proves the rule. Most will find that they have without realizing it found that vast portions of their life, sometimes more than they spend sleeping, have been frittered away.

The simple objection here is to my usage of the word fritter. Perhaps you feel your life is substantively enhanced by this time spent with the glowing rectangles rather than a harm. However, before you jump to that conclusion spend a minute thinking on the opportunity cost. Remember that every minute you spend using one of these devices is a minute you cannot spend doing something else. If you still feel this has been a net benefit for you then are in a fantastic place. If however you find that there are other things you wish you spent your time on then I recommend you read Digital Minimalism by Cal Newport. I also think it is time for us to start challenging the status quo and acting to take back the social internet.

There are some initiatives that are already working on this concept. The first that comes to my mind is the awesome IndieWeb movement. They advocate simply for people publishing their content on their own website and then syndicating it elsewhere if they desire. This gives people the power over their own voice again, and even if they are deplatformed or demonetized or obsoleted they still maintain control of the original content and are free to move it to new platforms with greater ease. Considering examples like this I was left thinking about what the ideal platform for syndication could look like that will still empower people to have this same kind of control over their content. The idea I keep coming back to is somewhere between a classic forum, Reddit, and Medium.

Imagine a stripped down Medium allowing only text, images, and hyperlinks, that retains the syndication abilities that Medium has. Except instead of the articles being organized into a publication or on your personal profile they are instead organized around debates/discussions. Posts would surface to the top based on some mix of voting and maybe some version of the old forum 'reputation'. No one will be able to see their reputation because it will be compromised of several things including: you will have a general site rep which will lightly affect your postings in all categories and topics, a category rep which will more significantly affect your postings in topics in that category, and a topic rep that will significantly affect your visibility on that topic.

Now how topics and categories will be formed has been a little bit more difficult for me to figure out. I have gone back and forth on whether or not it should be community created or if there would need to be more stringent proofing by the company. What I think I have settled on is the company creates the broad categories, then new topics are launched in each category daily. What topic will be launched next in each category is decided by voting. Seems to be the simplest solution.

Now we come to the trickiest problem:moderation. We have partially eliminated some of the most dangerous effects of deplatforming by ensuring that people still have access to their content (would be allowed to download archive even if they get banned from posting), but now must figure out how to handle issues. The upvoting and downvoting and reputation systems partially alleviate this problem by burying posts that are undesirable. This combined with automated screening for certain offenses, spamming among them can help also. However, you do still need humans who are moderating and looking for certain content. Child pornography cannot be allowed. Doxing cannot be allowed. Unfortunately, this means we will need humans watching our site for the most dangerous of content, but much of the objectionable and distasteful content needs to be allowed to stand in order for the site to truly be about discussions and debates.

Finally, I do not know whether or not a system like this could ever actually work. It is just an idea that has intrigued me for a long time.

Original post:

Sunday Reads: Links I Found Interesting

Mapping human microbiome drug metabolism by gut bacteria and their genes

A fascinating look at how the microbiome may affect drug metabolism. Important to remember that the game does not end at pharmacogenomics and we need to be paying attention to the complex interplay of numerous complex systems to understand drug action.

Deep learning can predict microsatellite instability directly from histology in gastrointestinal cancer

Thanks to the 'magic' of deep learning we may be able to better predict which patients are going to respond to immunotherapy in gastrointestinal cancer with cheaper tests. More people treating their cancer certainly sounds good to me.

CCR5-∆32 is deleterious in the homozygous state in humans

The gene that was CRISPR-ed in those Chinese babies makes it more likely you die. This isn't even accounting for the potential off-target effects. Turns out the thing we all knew was unethical is in fact unethical. Who da thunk?

Principles of and strategies for germline gene therapy

Following in the same vein as the previous article we take a theoretical look at the potential for these germline therapies.

The Sweetgreenification of Society

Interesting Substack post about the increasing stratification of society through the lens of boutique businesses.

RNA sequence analysis reveals macroscopic somatic clonal expansion across normal tissues

From one, many. Our bodies are a huge mess of different mutations each of which could or could not be maybe contributing to diseases. Thinking of yourself as having one genetic identity is flawed.

A Jaunt Down Financial Fraud Lane

A fun article taking a look at some of the numerous scams in the cryptocurrency ecosystem. I am partial to the disaster that is EOS.

Sunday Reads: Links I Found Interesting

EfficientNet: Improving Accuracy and Efficiency through AutoML and Model Scaling

Google did that thing they do again where they make vast steps in artificial intelligence and machine learning. Efficiency of these image recognition networks is up to 10* greater. Most of this gain is because they try to use "scaling coefficients" so that the network scales in a predictable way. I'm just mad because it's a TensorFlow model and not PyTorch so I can't drop it into any of my existing image recognition notebooks.

A promising step forward for predicting lung cancer

Another Google blog post about how they are doing incredible things. Man what I wouldn't do to work for Google Brain. (This research is also being done at the same university I am doing my capstone with, so hey maybe they can sneak me in) Okay so in this article they describe a state of the art result for predicting lung cancer using improved volumetric predictions of CT scans. They instead of looking at individual slices in the image are instead reconstructing 3-d structures to improve the accuracy. This both is and is not a crazy step forward. Being able to use the 3-d structure seems to be truly revolutionary, but some of the radiologists performed equally as well as it. Seems that it will be a useful assistance tool for now.

Moving Camera, Moving People: A Deep Learning Approach to Depth PredictionI promise this won't be all Google, but again what they are doing right here is incredibly cool and a little bit scary. They have found a way to approximate the 3-d size, shape, and depth of moving people even when the camera is moving. This work has really cool implications for AR and VR and a little bit terrifying uses for a potential police state. There are many places where face recognition has been banned or people are considering banning it, however, combining a 3-d map of a person with existing effective identification techniques like gait tracking can serve as a proxy for facial recognition in those areas. Combined with facial recognition it could provide an even stronger match limiting false positives, and avoiding false negatives.

Speech2Face: Learning the Face Behind a Voice

Okay we are finally away from Google, but into something even more terrifying. This neural network when fed a small sample of speech is able to generate a qualitatively accurate facial guess. The model seems quite adept at identifying both race and gender. Scary stuff.

Defund Crypto

This fun parody site created by Joshua Davis, Kyle Gibson, and the pseudonymous Cas Piancey mercilessly lampoons the tomfoolery of Kik's attempt to challenge the SEC. For the record, I do not think promising an Ethereum public DApp and delivering a one node Stellar fork is a good thing.


This interesting service will likely not be appreciated by the privacy minded. While they do have a strong privacy policy its purpose is to bring a ton of your disparate personal data together and find interesting correlations. Whether it is useful or just noise remains to be seen.

Key Excerpts and Commentary from NYAG/Tether Court Transcript

Shit went down

First of all link to the transcript:

Second of all congrats to @lawmaster and The Block for a great scoop. Now let’s get down to business.

Bitfinex is still being hesitant to hand over documents to the NYAG.  They have struggled to get access to documents relating to the transfer from Tether to Bitfinex, and this suggests to me that either the documents don’t exist or there is a very good reason they are not being shared.

Screen Shot 2019-05-21 at 5.03.56 PM

This is directly contrary to what Bitfinex has claimed to the public wherein they have claimed that they have been fully cooperative. Archive link:, I’m sure that there is no reason to think that Bitfinex is hiding something. No reason at all.

Shortly after this we learn very interesting things, Tether’s lawyer admits to Tether investing in Bitcoin:

Screen Shot 2019-05-21 at 5.13.37 PM

Luckily we have a sharp judge here who quickly gets to the meet of the issue and correctly points out that this seems contrary to the nature of a “Stablecoin”.

The Tether lawyer responds by confirming what we all suspected since the ToS change is that other assets includes cryptocurrencies:

The Tether lawyer then continues basically saying they will not produce documents and will instead appeal and challenge every single step of the way:

The Tether lawyer then also says that they do not think there is any amount of dollars they need to keep in reserve:

The Tether lawyer then takes the classic Tether defender tactic of it’s okay because banks do it too:

The judge quickly ascertains the issue with this and points out that this effectively means there is no reserves:

The Tether lawyer responds by saying it’s okay, if they need to they’ll earn money some other way, pay it back, and just delay redemptions:

Screen Shot 2019-05-21 at 5.22.23 PM

A little further down the NYAG reveals that Bitfinex/Tether executives get lump sum payouts from the unsegregated Tether accounts where no reserves have to be kept:

Screen Shot 2019-05-21 at 5.25.12 PM

The NYAG also reveals the juicy tidbit that the largest redemption ever was less than $25 million:

Screen Shot 2019-05-21 at 5.29.07 PM

Why is this particularly juicy? Well let’s take a quick trip over to their treasury address on Omni: here it does not take long to find bigger transactions coming in than that like this: So somethign is seriously amiss here.

Now relevant to this entire document is the issue of disclosure. Tether claims that they are not in the wrong because once they started using other assets they disclosed it.  However, is that true? I will contend it is not. Let us consider Tether’s own website:

In 2015 Tether openly admits to exchanging Bitcoins for Tethers without KYC. Now it is possible, but in my opinion unlikely that they still had sufficient fiat reserves at that point, but I think it is plausible to doubt that and to believe that Bitcoins have often been a part of the backing.

Screen Shot 2019-05-21 at 5.48.34 PM

In conclusion: Tether has paid executives dividends out of non-segregated accounts, does not feel a need to keep cash reserves, is buying bitcoins with reserves, and cannot handle a rush to redeem.  Their largest claimed transaction is also smaller than multiple apparent redemptions on the blockchain.

Original post:

Loading more posts…