Hey, I'm Maurice Weber. I craft this newsletter to provide you with the latest innovation news from the current week.
📯 This particular newsletter will be the first from InnoLetter. I've never written one before, so please shoot me an email if you have feedback!
If you want to know what innovation actually means, check out the latest blog post. From now on, the blog posts will go online each Wednesday while the newsletter will be sent on Sunday. Wish me luck for that new schedule!
🛳 The cargo ship Ever Given has been refloated, finally freeing up the Suez Canal (CNBC)
🔨 Hedge fund Archegos Capital, run by ex-Tiger Asia founder Bill Hwang, triggered $30B in large stock sales, facing liquidation (WSJ)
🎭 What is Bitclout? The social media experiment sparking controversy on Twitter (CD)
📈 Chinese ride-hailing company DiDi Chuxing expands to South Africa, will take on Uber and Bolt (TC)
🔭 ARK Invest launches a space exploration ETF, to began trading on Tuesday and saw $294M in shares traded (CNBC, Yahoo)
💶 PayPal launched a new crypto checkout service, allowing U.S. consumers to use crypto to pay online (RT)
💰 Goldman Sachs is close to offering investment vehicles for Bitcoin and other digital assets to its wealth management clients (BBG)
👓 Microsoft wins US Army contract for augmented reality headsets, worth up to $21.9B over 10 years (RT)
This Week's Selected Topics
AI In Boat Races
End of March, the Emirates Team New Zealand won the America's Cup, the most prestigious sailing competition. With the help of reinforced learning, an advanced AI technique, the team was able to improve their boat design process by an order of magnitude. Instead of relying on a simulator which needed to be operated by human sailors, this year, the team simulated new designs with an AI. So how does reinforced learning work? Essentially, an AI agent performs actions within an environment and receives rewards when it takes the “right” actions. For that to work, engineers have to set up three important steps.
1) There has to be a thought-through learning algorithm and a corresponding reward function.
2) The algorithm then needs a learning environment to take actions and learn based on the reward function. Often, this environment is a simulation of real-world conditions. However, it could also be a digital platform, like an e-commerce website, with a high number of identical actions.
3) All this entails a need for high computational power which should not be underestimated. Depending on the actions to be trained, even the use of advanced infrastructure can require thousands of hours.
In the original article of McKinsey, the authors feature a wide range of industry applications. It will be interesting to see how fast businesses will adapt and how sustainable new competitive advantages will be.
Facebook Data Leak Resurfaces
On Saturday, a dataset of 533 million Facebook users' personal data has been published in a hacker forum. According to Alon Gal, CTO of cybercrime intelligence firm Hudson Rock, the dataset includes the user's phone number, Facebook ID, full name, location, past location, birthdate, (sometimes) email address, account creation date, relationship status and bio. A Facebook spokesperson told the Insider that the data was scraped in 2019. This is in line with Gal, reporting that the data was offered online as a paid service before this weekend.
Thank you for reading the first newsletter! If you have any feedback, feel free to send me an email, LinkedIn message or whatever is most convenient to you.
The next blog post (hopefully online on Wednesday) will deal with the basics of Neural Networks. This will be helpful to better understand future (AI) projects.