#49 | Fail Fast, Home Gym, Hacking Covid, Vaccine Stats
|Benjamin Joffe||Nov 12, 2020|
Ben’s irregular newsletter mixes work and play.
Unsubscribe at the bottom / Sign up here
Find me: Twitter - LinkedIn - Medium - Website - Podcast
France went back from curfew to lockdown. Fortunately I’m better prepared now. I hope you are too!
1. WORK: Fail Fast
2. EXPERIENCES: Home Gym, Online Heckle, Hacking Covid
3. CULTURE: ‘French for X’, Queen’s Gambit, Fighters, Crowd Science
4. THOUGHTS: Plastic, Risk Cats
5. COVID LIFE: Shopping, Stats, Vaccine
As I mentioned in the previous newsletter, I took a leave of absence from SOSV to explore new things. I am currently part of the Entrepreneur First (EF) startup program in Paris. It’s been 6 weeks so far, and quite a journey.
EF calls itself a ‘talent investor’:
They select 50-60 participants with strong science, tech or business background.
For 8 weeks, they encourage them to form teams of 2 and explore opportunities. If the market is not promising or the team not productive, teams break up. To find a good idea and co-founder, most need a couple of tries.
At the end, EF funds the most promising teams (about 1/3 of participants get funded).
To me, the main advantage of EF is the selection of skilled, motivated and available people.
The challenge in teaming up is to find someone with complementary skills, and think of a project combining skills, interests, with market demand. It’s not easy.
My initial interests were in education and ‘future of work’, but I expanded my scope based on the group.
My third team just broke up. Here are some quick descriptions of the projects:
(1) Compact sensors
I teamed up with a Post-doc in optics.
He created a process to curve CMOS sensors, making the systems more compact and higher quality (our eyes are curved and compact optical systems).
We looked into AR/VR, but volumes are still low, and good consumer AR glasses are still far away.
Smartphones are the Holy Grail (1.5 billion units/year with 2 to 6 cameras in each) but qualifying a new process for smartphones looked like a (very) long journey, so I decided to explore other ideas.
(2) Informal Networking solutions
How to create higher value from connecting with strangers? That was the second project, where I teamed up with an NLP/ML/Big Data expert.
We looked into how to apply Machine Learning to match people better based on their interests. We also found some existing solutions.
The B2C ones (e.g. Lunchclub, Shapr) were struggling with monetization and/or engagement; the B2B ones (e.g. Donut, Lead.app) couldn’t do very good matching as the pool of users and data was limited.
After a series of customer discovery interviews with two dozen people, including in HR, our conclusion was that the need was not so high, except maybe with the ‘social onboarding’ of new hires. We also figured our passion didn’t lie in HR solutions, and decided to let it go.
(3) [New technology]
One way to find opportunities is to look at technologies recently introduced, when applications are still lacking.
With this in mind, I teamed up with a Post-doc in sensor fusion to realize quickly that the skills needed for this project were rather… mobile app development and AR/VR or 3D gaming skills. It wasn’t a good match so we dissolved the team.
I am still keen on this one — even as an experiment.
(4) Next project
Back to the drawing board with only 10 days left!
I am now looking into a ‘micro-coaching’ concept. It could start fairly low tech. Otherwise maybe I’ll find a good idea here.
If, like me, you don’t like running, and prefer to practice a sport with others, lockdown might also be wearing you down.
During the previous lockdown, a friend at my jiu-jitsu dojo had converted his bedroom into a mini-gym by putting mats on the floor (3x3m). A few of us trained there. Well, I just did the same in my living room, and restarted training. It’s likely to be a booster for my well-being. I hope I don’t break anything in the house!
I also use some online yoga and ‘Tabata’ workouts (named after Dr. Izumi Tabata and his 1996 HIIT research, apparently for Japan’s olympic speed skating team). The latter are 4min exercise sets. It’s quite clever as the time commitment is so low that’s it’s hard to not do it, and it’s often tempting to repeat a set or add a different one (or two).
I was invited to speak at an online event discussing the ‘third wave of deep tech innovation’ for a generalist audience (= not my usual startup/VC crowd). I had prepared all sorts of examples to surprise and delight.
Previous online events I spoke at were often ‘too peaceful’ and lacking in audience interaction but this session turned out the opposite: some commenters in the Zoom chat were quite negative from start and seemed only concerned about ethics, control and impact rather than discussing innovations.
The positive of this format is that unfiltered live comments help read a room and respond better to their interests. The negative is that a few vocal people might derail a panel.
I guess that’s where the art & science of moderation comes in — ideally with clear rules and technical tools (think about the difference between the first and second US presidential debate).
France is getting a second wave of Covid cases and has entered lockdown again. We have to produce an authorization each time we go out.
Normally you need to fill your details (including time and date), and either print and sign a document (pdf or docx), or do it online and download a file that includes a QR code.
I have a friend with 2 kids who needs at least 3 documents a day, and the police doesn’t like documents filled with pencil so you can’t re-use them.
So I found myself a fun little project with my newfound Python skills: automating authorizations. This is trivial for an experienced developer, but doing this myself was quite fun and empowering.
To make it short:
Editing a .pdf is not so easy, but editing a .docx is, and so is recreating a QR code for the digital version.
To cover the month, I now have an authorization for every hour of every day, and each of the 9 possible reasons (shopping, exercising, etc.). That’s about 6,500 versions.
Next step: putting it online with Heroku, Flask and Github (thanks to B. for his support and guidance!).
MOVIES / DOCUMENTARIES
David Attenborough: Our Planet***
If you like spectacular images to take your mind away from lockdown, you’re in for a treat. I didn’t watch all the episodes but was awed by the Forest one (with the hornbills, caterpillars), and the mating dances of various birds.
Something’s Gotta Give****
I re-watched this very cute romantic comedy starring two single seniors — and one of Diane Keaton’s best performances.
Merci patron ! (Thanks, Boss!)***
Francois Ruffin is France’s Michael Moore: an investigative journalist, activist and now even a member of parliament. This movie shows how he helped a couple, both Kenzo/LVMH factory workers, who were laid off when their jobs were relocated to Poland. Ruffin buys shares in LVMH to disrupt the shareholders meeting, then obtains a settlement and a job for the couple.
Adieu Les Cons (Farewell, Assholes) [unrated]
I like the director Albert Dupontel a lot, and watched interviews for twice the length of the film … then France went into lockdown before I could watch the actual movie! (apparently, it’s a story of a burn-out gone wild)
Dupontel is a bit unusual as he studied medicine, then went into standup comedy before switching to movies. He is not well-known outside France but worth watching. His stories remind me of Terry Gilliam for the surrealism (Brazil / Meaning of Life) and Ken Loach for the realism and caring about the working class (e.g. Sorry we missed you). I also recommend Dupontel’s older Bernie and See You Up There.
The Social Dilemma**(*)
This documentary talks about the risks of social media echo chambers, algorithms, manipulation and more. Due to my high exposure to the topic in recent years I didn’t feel I learned a whole lot but it’s probably worth watching. I also found ironical that Netflix (the producer) is not even mentioned in it…
The core of the problem is that tens (hundreds?) of billions are spent every year on very capable people to create and improve those ‘addictive technologies’, while our psychology and biology hasn’t evolved an immune system for this. And regulation is always a few steps behind (like with novel financial products — remember 2008).
Since the forces at play overwhelm us most of the time, the only refuge might be to use our rare moments of clarity to distantiate ourselves from the sources: modify our personal environment to make the undesirable behaviors difficult. Until a better solution arises?
A well-produced show about the rise of a young female chess champion during the cold war. I used to play chess and had some interest in the topic, but the show focuses on the mindsets of the characters rather than the game itself. It also delivers its feminist message efficiently.
Out of curiosity, I looked a bit into the history of women in chess. The Guardian reported an analysis of how statistical distribution could explain the prevalence of men in competitive chess.
There are about 800 million players worldwide of all genders, and only about 1,300 grandmasters. One study reported a 16:1 male to female player ratio in one country as an example.
The most recent figure I could find is that women represented about 2% of grandmasters in 2010 (all-gender GMs, not ‘Women Grand Masters’) and the top female player today is a Chinese ranked #88.
The Guardian talks about promoting more female role models — no doubt this show will help! (apparently chess streaming activity on Twitch already grew).
The original actors from the 1984 Karate Kid movie have grown up. One is successful, the other not so much. I thought it would be far worse than it was — the fighting skills are not impressive but the authors found interesting story angles. The Jackie Chan Karate Kid remake remains my favorite.
Dix Pour Cent (I'm the agent) — Season 4**
The final season of the show about a Paris-based talent agency, where some friendships miraculously survive the many betrayal. The most interesting in this season is how guest stars play out of character.
Emily in Paris*
A case of hate binge. A 20-something American moves to Paris to provide a US perspective to a local marketing agency. Imagine Amelie and The Devil Wears Prada had a baby.
I rarely watch MMA but one guys stands out: Khabib Nurmagomedov.
He’s a lightweight fighter who just decided to retire, undefeated at age 32 after 29 wins (including against Conor McGregor).
He is also apparently the most followed Russian on Instagram with 25 million followers.
His decision to retire seemed largely motivated by the passing of his father. His dad was a wrestling champion and trained him in wrestling (a popular sport in Dagestan) from age 8.
For more, check out this free 30min documentary where you’ll see kid Khabib wrestle baby bears.
Saladhine Parnasse, the Mbappe of MMA***
‘GregMMA’ is a former heavyweight MMA fighter, who now works as a YouTube channel host for the martial arts magazine Karate Bushido.
He goes to visit various gyms and dojos to explore different martial arts.
After an introduction, he generally ‘gets to work’ and does two rounds of sparring: one standing round (boxing/muay-thai/karate), and one focused on groundwork (grappling/BJJ). With his experience and size (he’s about 190cm and likely 95kg), Greg generally dominates both rounds.
In this episode, he visited 22yo Saladhine Parnasse, 173cm and 74kg in the video — a sizeable difference with Greg (and he usually cuts weight below 66kg). Still, things didn’t go well for Greg this time. So far this 22yo is undefeated with 14 wins and 1 draw. My money’s on him!
How to Start?
This budding artist felt stuck and got this piece of advice: ‘draw the same thing every single day’. This is what happened, which might be a good indicator of ‘how to get started’ = decide on a simple repeat task, so as to minimize the use of willpower (like for the Tabata method).
Crowd Science (Fouloscopie)
The YouTube channel of a cognitive scientist looking into … crowds. His latest episode (in French) explores prediction models and sampling biases in crime.
Apparently crimes have ‘aftershocks’ like earthquakes (note: I studied earthquakes in engineering school many years ago). Monitoring the predicted ‘aftershock hotspots’ for a week or two after a crime could prevent further waves.
The sampling bias comes from this: with equal crime rates, deploying larger police forces in one area would lead to more arrests, then to more police deployment, thereby reinforcing the perception of a higher crime rate there. It is possible to correct some of the bias by indexing police deployment to population size, but it will also be less effective on the whole. Tough choice!
Not much reading these days but except this newly published collection of essays by French hit novelist/essayist Michel Houellebecq. He shares his views — often not mainstream — on various topics: Europe, Trump, protests, euthanasia.
If you’re wondering about whether recycling plastic is economically sound, this NPR article might be of interest. I also remember my surprise when I heard that people sorted trash worse than randomness (< 50% correct if 2 bins) because we are so confused about the rules, but also the reality: your Starbucks cup? It generally has a hard-to-recycle plastic lining inside, and ends up ‘contaminating’ the recycling stream.
More Games: Risk
A classic strategy board game I’m playing on mobile now. I play mostly against the computer as it’s faster than human players, but diplomacy doesn’t count much. My takeaway: expanding aggressively stretches resources thin, which often backfires, so consolidate patiently first, while letting other factions kill each other.
Rental cat number 6 arrived and he has some problems: apparently the father of the owner overfed him and he’s now a solid 8.5kg with diabetes, and requires two shots a day. He also has a special diet and is often thirsty. He might also be stressed and/or sad being in a new home, and expressed it by meowing at unholy hours and avoiding his litter a few times. This might be the last cat we host for a while.
France is back into lockdown, as are various other countries. Here are some stats and personal views.
I forgot my mask once when going to the supermarket and ended up covering my face with layers of clothing. I felt extremely self-conscious. It’s also interesting to realize that seeing people’s face has probably become a sign of intimacy.
In France, we’ve had 42,000 deaths so far attributed to Covid. From March it seems that the excess mortality (compared to 2018 and 2019) is about 25,000 people, out of a yearly total of about 600,000.
The news of a vaccine sent some stocks up (e.g. airlines) and some down (the ‘stay-at-home / lockdown stocks’ like Zoom, Peloton).
Despite this, I doubt vaccines will be ready at scale very quickly so lockdowns will probably persist throughout 2021. We’re already at the end of 2020, should we still talk about Covid-19 in 2021?
How to test a vaccine?
The Pfizer announcement looked promising and I got curious about the clinical trial numbers they mentioned, and the process to validate a new vaccine. Here is how it works:
Since it’s not good form to contaminate people intentionally with a disease, you run a trial with thousands of people and compare with the ‘natural’ infection rate. In the case of Covid-19 the incidence is assumed to be 1.3% per year for untreated people.
Pfizer gave its treatment to about 22,000 people above 12yo, and a placebo to the same number. Then they waited to compare what percentage of people in each group were infected over the following weeks and months.
The announcement mentions a 90% efficacy at 7 days after the second injection, which is 28 days after the first. Total = 5 weeks.
With a 1.3% incidence among 22,000 people, you would get 22,000*1.3%*5/52 = 17.5 people infected. I guess 90% efficacy means that the treated group only had 2 infected people? The announcement mentions that ‘evaluable case count [reached] 94’ — maybe by batching the treatment over time? Not sure how that works.
Note that the start date of the trial was April 29. However, only about 8,000 people were recruited by then, and the document indicates: Estimated Primary Completion Date: June 13, 2021 (Final data collection date for primary outcome measure) and Estimated Study Completion Date: December 11, 2022.
In conclusion: it’s still pretty small numbers but could be promising.
Let’s hope things improve before Christmas!