Happy Tuesday.

Yesterday’s essay is freely available. I use some new data around “employee wellness programs” to talk about selection effects – including branching off into loyalty programs, employee recruitment and the NBA.

Take a read (and sign-up to get the essays every week!)

Edward, Startup Curated

This week’s sponsor:


The easiest way to collect and manage website feedback

Turn email chains and spreadsheets of vague feedback into actionable tasks. Pin feedback to elements on a website and capture the technical information to help replicate bugs and solve issues. Track feedback tasks to completion.

Try BugHerd for free today.

Edward, Startup Curated


  • Microsoft vs Google: Google is holding strong and refusing to pay large Australian media companies (i.e. Rupert Murdoch) 10% of their Australian revenue PLUS 30-days notice of any algorithm changes. Microsoft, sensing weakness, has told Australia they would be happy to become the new search engine for the country. This is a pretty evil play - Microsoft knows what is happening in Australia is wrong, but they see weakness. And this version is being evil doesn’t give them any bad PR. Very gross.
  • Clubhouse clones: Clubhouse has officially blown up - its latest funding round, less than a year after founding values the company at $1B - and it is still in private beta. Now Twitter and Facebook are both launching clones. I have tried Clubhouse and I am not impressed so far, but it has clearly found a niche. And the opportunity for live audio is only beginning. Clubhouse is to podcasts as Twitter is to blogs. It is possible to killer use case is listening to a smart commentator during live events. Its coming (If any Marketing BS subscriber would like an invite to Clubhouse beta, just respond to this email). Related: Marc Cuban is also launching audio app - Fireside Chat. The app is designed for both live and recorded audio (podcasts), but focused on ways that the producers can monetize that audience.


Edward, Startup Curated

Marketing to Employees

  • Rebranding: Aunt Jemima re-brand has been announced. The new name will be “Pearl Milling”. It harkens back to the origins at least, but expect this to hurt their market share a lot. I hope the employee retention was worth it. Meanwhile, the San Fransisco Symphony is also rebranding - to appeal to a younger audience. One has to ask, how much of this is for the (younger) employees rather than the (older) audience?
  • What Disney Employees care about: Disney has defended their choice to film Mulan with the support of the Xingjiang Chinese province (where Uighur interment camps are located). But they were unwilling to tolerate (far) right wing tweets from Gina Carano (Cara Dune on the Mandalorian). Many people seem to be surprised by this, but it comes down to what Disney employees care about - and it is domestic politics, not international ones that matter (at least for now).
Edward, Startup Curated



COVID and the New World Order

Edward, Startup Curated

AI/Machine Learning/GPT-3

  • Metahumans: Unreal engine is a platform for building games. They just released their latest “character generator” and it is… unreal? [Worth stopping what you are doing right now and watching the video](https://www.unrealengine.com/en-US/blog/a-sneak-peek-at-metahuman-creator-high-f
Edward, Startup Curated

  • Bullying: Facebook invests more in moderation than anyone. Anytime they can find a way to improve automated detection they save millions on human beings. For most types of banned content they were able to catch 95% automatically. For “bullying and harassment” the algorithm only caught 13% of infractions (the rest were identified by humans). But their new AI is much better and are catching harassment almost 50% of the time. This is an arms race and AI is winning.
  • Hiring bias: MIT’s technology review has a piece on auditing AI-hiring algorithms. The conclusion is it impossible to know if the hiring tool is really doing a good job or not (evaluating employees is ALSO hard - so what are you measuring against?). What is not asked - and should be - is “are algorithms MORE biased than the humans they are replacing”. There is a fear that algorithms will systematize prejudices, but at least we can audit algorythms and improve them. That is much harder to do with individuals who may be prejudiced and we never really know.
  • AI Incident database: Sometimes airlines crash. When they do they get added to an aviation accident report. This allows researchers to figure out patterns for why accidents happen and eliminate the root causes. Now the same is being done for AI-errors.
  • Talking to dead people: AI Dungeon already does a great job with simulating discussions with fictional characters - as long as there is enough information about the character in the data set (i.e., it worked great for my 5-year old to talk to Elsa, but not so good to chat with Rascal from the Dory Fantasmagory books). Microsoft has just patented an AI chatbot for talking to dead people. In theory if you could get enough of a person’s content creation (tweets, texts, emails) you could simulate them pretty well. So creepy there was already a Black Mirror episode on it…



Keep it simple,


Edward, Startup Curated