View profile

🏢 🚙 🤖 Issue #33; waymo v tesla; scooter hacking and uber's instagram moment

🏢 🚙 🤖 Issue #33; waymo v tesla; scooter hacking and uber's instagram moment
Welcome to my newsletter, where I discuss thoughts and news on the intersection of the built world and technology #retail #mobility #realestate #tech
Please contribute to the community by forwarding this to someone who might enjoy it also.

PSA: I’m writing an article looking at US venture investors’ participation in European rounds - from Seed to Growth stage. For any US investors, would love to get your take on:
  • What attracts you to investing in European companies?
  • What is your primary sell / value-add to those companies?
One Big Thought
🚕🤖 Tesla v Waymo the race for data
This week Waymo decided to put safety “drivers” back into the drivers seat of their autonomous vehicles, after nearly a year of them being in the vehicle but unable to interveen in an emergency. This is bad timing as the company just launched the first commercial ride-hailing service for unmanned vehicles, Waymo One in Phoenix, Arizona.
This got me thinking about the different ways in which autonomous vehicle companies are training their autonomous systems. In large part the race to Level 4/5 automation is a data problem, gathering sufficient, variable data and parsing it through autonomous systems. This is a gross over simplification of a highly complex problem - variables such as data sources, quality and location means this problem can be attacked in a number of different ways.
Whilst a full analysis of methods may be overbearing for this newsletter, let’s have a quick look at how the two most prevalent companies in this space, Tesla and Waymo, are doing this.
The two biggest and most obvious differences is the way their vehicles have been deployed on roads. Waymo’s retrofitted vehicles have been predominantly trained on suburban environments, whilst Tesla’s “full-stack” cars fitted with Autopilot gathers training data generated from their customer journeys. Let’s take a look at both these tech stacks:
Tesla notably doesn’t use LIDARs, which Elon has described as a short term ‘crutch’ which OEMs will need to move away from in the future due to their high cost. Though most would disagree with Musk on LIDARs long term viability, citing likely cost decreases and outsized precision. Beyond this, the company is generating training data through its human-driven Autopilot system - which it has recently been using in “shadow mode” to effectively A/B test decisions that its autonomous system would make versus a human in a real life scenario. It’s worth noting that Tesla is able to ship Autopilot software updates over-the-air (OTA) to its vehicles. Tesla is primed to benefit from a virtious cycle of -> more production vehicles -> more data gathered -> improvemenents to software delivered OTA -> repeat.
A Waymo vehicle in action
A Waymo vehicle in action
Waymo to date has generated large datasets from real-world testing in defined geographic areas (Austin, Mountain View and Phoenix) and the use extensive of computer simulations in a program called Carcraft (if you just laughed you’re a gaming nerd!). Waymo is currently undertaking the majority of it’s testing miles through simulations - which allow it to subject the AV to a wide number of scenarios whilst in the safety of a virtual environment.
“Without a robust simulation infrastructure, there is no way you can build [higher levels of autonomy into vehicles]. And I would not engage in conversation with anyone who thinks otherwise.”
It carefully combines simulation testing with real world autonomous testing. Adding real world testing is crucial in developing statistical models which can be further refined and tested inside a simulation. Conversely, simulations might generate thin-tail events which require testing in controlled real-world scenarios. Waymo currently has a considerable advantage in terms of Level 4 (fully-autonomous on roads) miles logged as the extent of their testing program has been considerable.
The challenge of AVs remains the same; to prove that they are considerably safer than human-driven vehicles which currently see 1 death per 1oo million miles travelled. Comparing this with the current number of vehicle miles tested, its clear how far AVs have got to go.
Lastly, I’d say whilst most believe that Waymo has a data advantage over Tesla, many, including public market investors may be underappreciating Tesla’s data gathering capabilities.
Interesting Product
Yelling Across Cubicles (YAC) - is the distributed workforce + desktop version of push to talk cell phones from the early 2000s (remember those!?). The product, is fresh out of beta, and allows remote co-workers to quickly communicate with one another via voice
Vague Scientist
Distributed teams: Founder Collective looks at the five types of distributed team communication systems that companies can use to empower their workforce
Scooter hacking - this informative and potentially incriminating piece looks at the various ways to f*ck with a scooter, and shows just how much the hardware needs to evolve. My personal favourite is the bike lock over the wheel so no one else can use it. So analog.
Scooter consolidation (paywall) - one of the questions I asked in last weeks newsletter was whether scooter consolidation was afoot? According to several sources, Uber may be looking to acquire Bird. Two notable points: a) Uber is an investor in rival Lime b) This will be a great sweetener to Uber’s IPO story. Will this be Uber’s acquisition equivalent of Instagram or Youtube?
Spain to ban cars in cities - Spain is quickly positioning itself as one of the most forward thinking in terms of urban policy and mobility. The country is passing law to ban everything but no-emission vehicles from major cities
New York’s MTA really outdid itself with this tweet:
🏙 Real Estate
Is Shared Living the New Luxury? An interested piece by Space10 - who explore whether overpopulation and increasing disconnection will trigger a new age of collective living. Interestingly, this piece looks at co-living through the lens of community instead of purely being cost and space effective housing
Grab in talks to invest in OYO - The Indian asset-light budget hotel chain which has standardised, branded and listed over 8,500 hotel rooms is courting Grab (Taxis) for a $100m investment in their latest $1bn investment round. There are also rumours that OYO is going to be entering Europe soon 😗
Airbnb building homes - Airbnb’s future division, Samara, announced a new initiative called Backyard which will be launching new single and multi-unit properties in 2019. The company states that new manufacturing processes and the connected home will be key parts of the build. What will be interesting to see is how they will use their data advantage and community to predict demand and sell supply of units - also whether they will lease or sell the units remains unclear
Listeners Corner
Ark Invest - Micromobility and the Future of Transportation with Horace Dediu: Horace and Ark’s Sam Korus discuss the evolution of micro-mobility, manufacturing and infrastructure
a16z: When software eats the Real (Estate) World: One for the eyes as well as the ears. Alex Rampell talks through the transaction layer of Real Estate and how software is removing friction
Thanks for reading!
Did you enjoy this issue?
Sam Cash // Physical World Technologies Newsletter

The intersection of the physical world and technology; with a focus on future mobility,real estate, retail and cities.

In order to unsubscribe, click here.
If you were forwarded this newsletter and you like it, you can subscribe here.
Powered by Revue