Saturday, August 29, 2020

Pandemic Atlanta Lockdown - Week 24

Malibu Summer Rose

I was busy with two conferences this week - SigKDD and EuroPar. That also means, my week was different from the regular.

I can remember things when I am drunk. My behavior doesn't change with alcohol. Not sure whether it is good or bad. I got only drunk twice outside my home. It is funny how these moments make a good memory.

The first time, Barcelona (2013) Bar and club crawl. Lost count of the shots. My bestie sensed I was slightly unstable, grabbed my last shot glass from my hand, spilled it to the ground in the bar, and told me, "Enough" (He was drinking as much as me - but he had a higher tolerance then).  I was waiting outside for a few minutes after I went out of the bar to vomit in the street corner. I couldn't bother to find the bathroom in the bar. My friend came outside to ask whether I am doing ok. I told him I vomited, and now I am ok. Then we (I guess, maybe 6 out of 17 - 20 of the original group that we were with) decided to go back to the hotel.

I went out and vomited instantly.

The second time was in San Francisco (2016) on a cruise in a conference dinner. We were student volunteers. The professor in charge had given us the drink tickets and said, "you can have as many you want." I drank maybe 8 or 9 drinks. Somehow my friends (the other student volunteers) lost me when the boat landed (I thought so - they later told me I left before them). I walked alone for 30 - 40 mins straight to the hotel using my memory - because I know I was too drunk to call an Uber and interact with a human. I judged that walking drunk is safer than an Uber (and a potential to vomit inside the Uber). I safely arrived at the hotel and fell asleep instantly after removing my shoes.

I don't get drunk as often. So when I do, my memories are vivid. Other times I drank outside my home, never reached that level at all. First, drinks outside are expensive (unless I buy bottles - which I sometimes did), unlike me stocking hard liquor at home and drinking. Second, we never had alcoholics in our bunch (usually international students). Most of us were doing one or more of the below: partying, trying to find relationships/romance/..., and exploring the new countries and cultures. Getting hard-core drunk was not anyone's goal. We were mild at our parties.

During the pandemic era, I got drunk home alone twice (and perhaps drunk more than what I am used to when I go out - as usually it is just one or two beer or cocktail - the cocktails outside are watered down with too much ice too). In any case, I know when I am getting drunk. So I stop before that.


Next weekend is a long weekend. I am looking forward to it.

SigKDD2020 and EuroPar2020

This week went with 2 virtual conferences - SigKDD and EuroPar. Due to their timezone differences, I was able to attend them both. EuroPar was in the morning, whereas SigKDD was in the afternoon. Consequently, I spent 7 am - 9 pm with these conferences for most of Sunday to Friday. I skipped most of the EuroPar workshops as they were 3 am - 7 am my time. However, all these videos are now in YouTube. Most of the conference videos were initially unlisted. But they will eventually be properly listed.

EuroPar used Slack and Zoom Webinar. SigKDD used Zoom webinar, Zoom meetings, Whova, and VFairs. There were 'introduction" threads on Slack and Whova where we all introduced ourselves. I was attending as an audience. I did not present a paper. Especially those who presented papers received a considerable interaction (Q&A and suggestions) through these channels.

Online conferences have a long way to go before they can reach their expected heights though. For instance, many times there were problems with the audio or screensharing during the conference. There was a racist sexist zoombomber once during the KDD session. There was a token registration fee from KDD and the event was free to attend for EuroPar. I believe, ideally the virtual conferences should be free for audience. You may charge the authors for the publication fee, of course. Also, I think, the zoom meetings should be configured *not* to enable everyone's cam and mic by default, especially in a large-scale event such as SigKDD. Every time someone joined we heard some disturbances in zoom meetings.

SIGKDD 2021 Goes Hybrid!

SIGKDD2021 aims to be a hybrid conference, with on-site event in Singapore with online streaming with the same or similar apps for those who cannot travel to Singapore. These are all assuming everything goes well with the COVID-19 situation, of course. Indeed, a virtual conference will never come close to the feeling of an in-person conference. But in 2015, when I had a workshop paper in SIGKDD. However, my Australian visa got delayed and I couldn't participate - my friend who lives in Sydney presented the paper on behalf of me. As such, a hybrid conference can bring the best of the both worlds. Who can afford to and like to present their paper in-person can fly to the conference on-site, whereas others can present online. This is better than a random colleague presenting the paper on-behalf of the authors and suffering to answer the questions. This will also encourage researchers to conferences regardless of the location. Often, we are forced to choose conferences in proximity, due to visa, funding limitations, and travel restrictions.

EuroPar 2021 Goes Lisboa!

EuroPar2021 will be in Lisboa, organized by Tecnico-ULisboa (my university) and INESC-ID Lisboa (my research lab). I am so tempted to submit my paper, hoping the travel restrictions and COVID-19 will be history by then.

EuroPar 2020

Joining the EuroPar sessions remotely
EuroPar is a top conference in parallel and distributed processing. As such, it contains topics that are directly relevant to me from my MSc/PhD days. There were so many interesting papers. 

I attended the below sessions in full or in part.
1) Cluster, Cloud and Edge Computing
2) Scheduling and Load Balancing
3) Best Paper and Best Artifact
4) Data Management, Analytics and Machine Learning
5) Parallel and Distributed Programming, Interfaces, and Languag
6) Theory and Algorithms for Parallel and Distributed Processing
7) Keynotes


KDD 2020

The papers in KDD used to focus on data mining and patter recognition more years ago. Now, everything has turned to deep learning. Especially, this year, with the COVID19, there were more sessions focused on COVID-19. Even some presentations that are not really related to COVID-19, made some reference to COVID-19. That is the impact of COVID-19 on the world.
 
Interestingly, one conference organizer mentioned how there were more participants in all the sessions that had "deep learning" in their names, whether workshops, tutorials, or the main sessions. Listed below, are a subset of the sessions that I attended:

1) Learning with Small Data
2) Fairness in ML for Healthcare
3) KDD 2020 Opening Ceremony and Keynote Address
4) Hands On Tutorials: Put Deep Learning to work: Accelerate Deep Learning through AWS EC2 and ML Services
5) KDD 2020 Opening Ceremony and Keynote Address
6) Research Track Oral Presentations: Parallel and Distributed Learning and System
7) Late-breaking Session: Emerging Data Science Problems in the Age of COVID-19
8) [DSHealth] 2020 KDD workshop on Applied data science in Healthcare: Trustable and Actionable AI for Healthcare
9) Research Track Oral Presentations: Big Data and Large Scale Methods
10) Plenary Session: Diversity & Inclusion Closing Remarks by Latifa Jackson
11) KDD 2020 Closing Ceremony - Keynote Address by Allesssandro Vespignani and Closing Remarks

Tuesday, August 25, 2020

Why do I see East Asian music videos popping up randomly on Twitter?

TWICE, a group commonly found in Twitter fan cams

The Internet is filled with stars and their stans (basically, fan accounts that dedicate the entire account to post videos and photos of the star/idol). Often they just post the videos and photos in their own profile. Sometimes, they post them as a reply under popular posts, hoping to get more views.

There was a time when K-Pop stans were sharing fan cams to completely random Twitter threads that have no relevance to them. Their only aim was to increase the views of their idol's videos and the reach of the idol. These stans are not necessarily from South Korea. Their love for K-Pop music unites them. Their posts' had a short lifetime since others find it quite annoying and report them instantly for spam every time they post.

 

Then came 2020

Like everything else of 2020, K-Pop fan cams unexpectedly faced a paradigm shift. They formed as a left-leaning political force in the USA. From what I know, this change happened in a decentralized manner. There was no leadership or careful planning. What they did was simple. Many random stans independently targeted hateful hashtags, and then post a cute fan cam of their idol with the respective hashtag of their idols -- for example,  #TWICE. Then everyone else of the idol's hashtag did the same. #[IDOL-NAME] #[HATEFUL-TAG] and boom, the hateful hashtags became music libraries! K-Pop won, hateful hashtags lost.


Interestingly, this pattern also extended to J-Pop and C-Pop to a smaller extent. For example, I noticed a few fan cams of #SNH48 C-Pop girl group popping up to dilute the racist hashtags. Eventually, K-Pop stans also conquered TikTok. People started noticing only after the stans had already made a difference in the USA political scene.

 

 

Consequently, people's perceptions changed. Most tweeps (i.e., Twitter users) started to see the K-Pop stans more and more positively. People stopped marking fan cams as spam. I certainly stopped reporting them altogether!



I decided to test this. I tagged #Twice on a hateful post from a "verified" account. Yep, I got the response back from some stans as a video reply.

 

Another time, I had an interaction similar to this:

 

Stan: People should be kind to each other #[HATEFUL-HASHTAG] [A random video from TWICE].

Me: That's a beautiful video, thanks for your service.

Stans: Thanks for your kind words. #[HATEFUL-HASHTAG] [Another random video from TWICE]. 

 

 

It works

The stans are diluting hateful hashtags, taking over the Internet with their cute fan cams. You may consider it a spam. They even risk an account suspension, as indeed their behavior fits the typical profile of a spammer. But they are always harmless.

 

Such replies annoy a racist or a troll.

It goes like this: 

 

Racist: [some nasty stuff]

Stan: [random-fan-cam]

Racist: [some more nasty stuff]

Stan: [another-random-fan-cam]

 

The racist gets all worked up and loses his/her energy. But a stan doesn't lose his/her cool. The racist says nasty things to them. He/she will get a reply, yet another fan cam. You may mistake the stans as bots. But they are not. They are cool people out there, conquering the Internet with fan cams.


K-Pop fan cams are the new cat videos of the Internet! Yes, 2020 y'all.

 

Now, if you see me replying on Twitter fights with K-Pop, C-Pop, or any random music videos, you know where I got the motivation from.

 

For actual Twitter wars, we shouldn't give too much variety in the videos in our replies. That derails the purpose. Best to stick to one artist or a girl group such as #TWICE. For added moral support, add their recent tags. #TWICE_Beyond_LIVE#트와이스#MOREandMORE

Sunday, August 23, 2020

Sockpuppets and where to find them

The Disguiser. An unrelated scene.
Internet Sockpuppets are an annoying online phenomenon. 

* A sockpuppet qualifies as a fake account, when it assumes a fake identity -- typically with a real-sounding name, a physical location that sounds real, a random photo scrapped from the Internet ideally not of a popular person, and some profile text/story to create a persona that looks real to a non-suspecting reader. 

* A sockpuppet qualifies as an anon account, when it comes with minimal identifying information -- no real name, no photo, nothing personal or identifiable information.

So what exactly is a sockpuppet?

There are several definitions. My definition goes like this:

"A sockpuppet is a fake or an anonymous profile created by someone with an existing established profile in the same platform, when the new profile aims to boost the popularity/prominence of his/her established profile (i.e., the "primary" account) or attack those presented opposing views of the existing account."

The distinction needs to be made because, sockpuppets are typically attached to the hand of the owner. They cannot stray away. That is the same for the Internet sockpuppets too. They stay close to the primary account.

Why do sockpuppets exist?

1) To manipulate the social media democracy + meritocracy

Social media and online platforms such as Twitter, Facebook, Wikimedia, and StackOverflow thrive by a combination of democracy and meritocracy. 

Each account usually has a single "vote". That means, you can typically "like" or "upvote" a comment/story only once. This is a democracy.

On the other hand, those who with most views/likes/shares/retweets receive larger reach and prominence in the discussions. As you receive larger reach, you are also rewarded by the platform. For example, you may get promoted to an "upper level" and receive additional capabilities such as moderating new posts (in case of platforms such as StackOverflow and Wikipedia). This is a meritocracy.

Then why doesn't this beautiful combination of democracy and meritocracy work flawlessly? This combination should work well in the physical world. But in social media, a vote typically means an account. Most platforms do not have proper means in place to ensure that a human creates only one account. In fact, several platforms allow and encourage owning multiple accounts for various purposes. I have maintained 3 bot accounts in my Twitter life so far (they are listed here), to automatically post updates from my blog and undergrad project. But of course, they are not sockpuppets. They are stand-alone bots. They are not fake, not really anon either as they have identifying information pointing to me. 

A sockpuppet, on the other hand, *will* interact with its primary in some way or another. Because that is their responsibility: making the primary appear stronger than it really is. With the "rich gets richer" scenario in social media, retweets by a large number of fake accounts (including sockpuppets) is expected to receive more legitimate retweets too. Sockpuppets, unlike typical fake or bot accounts, will organically interact with the primary account for the betterment of the primary account, through supportive comments, retweets, and likes.

2) To fight the opposing views without compromising the primary account's legacy

The sockpuppet, once created, will start traversing its primary account and respond accordingly. I have identified several sockpuppets in Twitter, from all sides of political spectrum.


Sockpuppet owners can be actually smart people. Don't underestimate them. That makes the sockpuppets smart too - but with no accountability (since they are a fake/anon). That is a dangerous combination.

Unlike the primary account, the sockpuppet will attack you vigorously, perhaps using racist, provocatory, or vulgar messages.

How to spot a sockpuppet?

In fact, it is easier to spot and confirm a sockpuppet than you expect. Sockpuppets are close to their primary. They respond positively to the views that support the primary. They will attack the opposing views with nastiest mean comments. They all follow a similar pattern, and leave obvious traces to connect with the primary. Typically, they will reply to the negative comments that haven't been replied by the primary adequately (if at all). Their replies will be uninhibited. This aims to maintain the neutral stance of the primary and to keeps the primary more professional. Sockpuppets will try to provoke you. They are a sockpuppet whereas you are in your main and perhaps the only account in the platform, which you may also be using for professional purposes and connect with your colleagues. Sockpuppets thus use your restrain as their strength.

However, sockpuppets fortunately have a big weakness unlike other fake accounts. That is, their proximity to the primary. They will first reply to the messages directed at the primary, or try to be smart and reply to the posts made by the opponents - but not directly on the same thread as the primary. Their language and timing usually will leave sufficient hints to tie the sockpuppet's identity to the primary. 

Some sockpuppets try to be too smart and alter their behavior. I once dealt with a sockpuppet (who is in real a software developer and a Twitter star) who pretended he did not know how to message via Twitter! This is a deliberate attempt to look different. Most sockpuppets will also casually talk with the primary (as in a dual acting by the same actor in a movie), to make it sound like they are different people. Sockpuppets also have a trait of deleting their old tweets to remove traces for future interactions. Because, unlike you, sockpuppets don't really have a reason to keep their tweets safe. Once they have served their purpose (for example, trolling their opponent), they can be deleted.

Every time I suspect a sockpuppet, I casually drop the name of the suspected primary, after sufficiently distracting the sockpuppet. The sockpuppet instantly panics and loses its cool and track/flow. Once, a sockpuppet stopped responding to my tweets for hours but instantly in 5 seconds replied with a nasty tweet once I mentioned the suspected primary's name as in "Hi [primary]!"Usually, if you publicly accuse the primary for being the same person as the sockpuppet, the primary will categorically reject the claim.

How do online platforms treat the sockpuppets?

Wikipedia bans sockpuppets outright. I love that. They trace the IP addresses and once they confirm the sockpuppets (ensuring that you are not merely two people from the same household), they suspend and ban the sockpuppet instantly. It ensures that Wikipedia is fair, especially when making important decisions: should we keep this new wikipedia page that is proposed for a speedy deletion, for one example. If a repeat offense is detected, Wikipedia will also ban the primary.

Twitter does the same too, but not that vigorously. Because, Twitter indeed allows multiple accounts. But if the sockpuppet is proven malicious, then Twitter bans it. Make sure to report the suspected sockpuppets to Twitter when they become disruptive. Twitter will check the pattern and suspend them if confirmed.

The first step in winning a sockpuppet is to identifying one, and then making sure the sockpuppet does not emotionally attack you to make you a weaker persona than the primary. Let's keep the social media clean and sane. Peace!

Saturday, August 22, 2020

Pandemic Atlanta Lockdown - Week 23

Nothing new to report. A regular week. I was working on my papers. Did not do anything creative. I have 2 virtual conferences coming next week. EuroPar and KDD. I am looking forward to them.

There are several things that annoy me about 2020. The few things that make me going forward: 

* Memories of the beautiful past. 

* The hope that humanity will overcome the pandemic and associated restrictions eventually. 

 * Weekends sedated by strong cocktails.

> "We want to make your life into a movie, with actors Dhanush, Yang Zishan, and several others" 

>> "What, why me? I am just a roadside man"

I woke up from a weird dream. What is worse? I am still in 2020 and I must go back to sleep.

During a recent flight, I watched the movies "Anegan" and "Till the End of the World." I liked both of them. So this cast doesn't appear random. Dreams are a strange combination of reality and imagination, as well as past and future. We need more research on what makes a dream.

Monday, August 17, 2020

Pandemic Atlanta Lockdown - Week 22

Coffee with Amarula
I submitted a journal paper this week, which was dragging for quite some time. This week, I also successfully managed to swap my Tuesday with Sunday, thus shifting my weekend to Saturday and Tuesday. I may try this again later. There are more deadlines coming my way in the upcoming weeks.

Some Teledrama during the weekend

During the weekend, I finished watching a Chinese teledrama series The Disguiser  伪装者. It is included in Amazon Prime with subtitles. However, the last episode (episode 41) had wrong subtitles. Therefore, I had to watch that episode in Viki, with ads. A Twitter friend recommended Nirvana in Fire 琅琊榜 since I liked this drama. Nirvana in Fire seems to have an overlapping cast with The Disguiser. I may watch that one some time.

 

Parallels of COVID19 Pandemic and the HIV AIDS pandemic

The parallels between the COVID19 and HIV AIDS pandemic are becoming harder to ignore. Just like AIDS, COVID19 has also become a humanitarian crisis. No way we can solve this pandemic only by medicine. It is not going to go away. Due to AIDS pandemic, people got more used to safe sex with barriers, many minimized casual sex, others are like "fuck it, if I get an STD, I get an STD" and continue their risky sex patterns. Although there are some medicines that minimize the chances of getting HIV, no vaccine or no curative medicines.

Similarly, we will learn to live with COVID19. Follow some best practices (masks, avoid too much crowds, etc). Others will continue the life as they were before, like, "If I get COVID19, I get COVID19". There may be vaccines or even medicines. But nothing is going to prevent or cure 100%, I fear. So this is going to be a mixture of changes to society, economics, geopolitics, and international relationships, and not just medicine.

Somewhere next year, I think, the world leaders will realize this, and take mitigative actions to minimize COVID19 spread, while resuming life. We also come to the realization that this is the "new normal." A bit more thankful that we are not affected (yet). This is unprecedented (at least, in our recent generations), after all. Another interesting distinction to be made is whether this is to be called an on-going pandemic such as AIDS or an endemic disease such as yellow fever. The differentiation maybe a combination of science, politics, and several other factors.

For how long do we have to live with this new normal of fear is the question. Avoiding AIDS is easy. Avoiding COVID19 isn't. That makes COVID19 a more dangerous pandemic.

Saturday, August 8, 2020

Pandemic Atlanta Lockdown - Week 21

Our Niffler abstract is accepted for oral presentation at CMIMI, a conference gone virtual. Now we are working on its full paper. Those are good news, especially when nothing else seems good - the pandemic, othering of foreigners and minorities, and disasters across the world.

The USA is considering a ban on the China-based social media platforms Wechat and Tiktok. Tiktok is a useless app. But the Wechat ban is going to be troublesome to those who have familial and business ties to China. We are already battling a horrible pandemic and governments acting all petty during this time. Internet censorship and othering, ostensibly presented as national security interests.

I am considering to innovate on the workweek as we are continuing to work from home. I will update once I have tried that.

Usually, my year goes fast once we reach August. Now, we are in August, and I feel 2021 is near. Although I am happy that this nasty year will come to an end soon, I fear that 2021 may be no better, as the pandemic continues to thrive.

Innovating the work week?

An illustration of data flow. 😅

Who else thinks innovating the workweek is a good idea, during the pandemic work from home? I am seriously considering that. The approach is simple. Rather than Monday - Friday, you choose your own workweek, making sure you don't have meetings and responsibilities for the regular weekday that you select as your "weekend" day. It could be Monday, Tuesday, Thursday, Friday, and Sunday. This thought is not entirely new to me. That is precisely what I did in Portugal during my Ph.D., because movie theaters, parks, and beaches were emptier on Wednesday than on a Sunday. However, there are a few challenges to emulate this. First, unlike me during my Ph.D. time (my Ph.D. supervisor is the most understanding professor ever, and we knew each others' work styles very well), students may not have that luxury and freedom. Besides, collaborations and other responsibilities will most likely prevent the majority of employees from attempting this. But I still think this could be an interesting option if supported by the employer, especially during the pandemic WFH. Before COVID19 WFH, I also worked 7:25 a.m. - 4:45 p.m. every day. In Atlanta, I rarely did the 995 (9 a.m. - 5 p.m., 5-day workweek) anyway as I like to start and finish earlier. Now, my times are quite stretched as I work from home alone. I start around 8 - 9 a.m. and go until late. 


There are a few EU countries that have adopted or planning to adopt a 4-day workweek. Interestingly, in Asia, it is going in the other direction towards a 6-day work week, including weird systems such as 996 (9 a.m. - 9 p.m., 6 days). In the future, if I happen to create my own company, I will consider our own workweek, with seven days of work followed by three days of "weekend", making a long week of 10 days, innovating the calendar altogether. Then, every weekend will be a long weekend, while maintaining the same work hours per month. During one of the upcoming weeks, I will try a weekend workday and split my weekend into 2, with either Saturday and Tuesday off, or Sunday and Wednesday off. I will see how that goes.