Tesla Detailed Its Autonomous Strategy and Offered Autonomous Test Drives at Autonomy Day 

Follow Tasha on Twitter @TashaARK

At its Autonomy Day in Palo Alto this week, Tesla offered test drives featuring some of the Autopilot’s soon-to-be released capabilities. The car merged on and off of the highway, changed lanes, and turned left at a four-way-stop, all without the engineer taking the wheel. It did wobble on two occasions, once at an intersection and then in response to cones around a service truck, but the engineer never needed to take the wheel. All in all, it was an impressive test drive.

For the first time, Tesla disclosed the economics it anticipates from its autonomous taxi platform. According to its forecast, while the autonomous taxi service will cost 18 cents per mile to launch, it could be priced at $1 per mile, both estimates in line with those in ARK’s model for 2023.  Before 2023, Tesla could charge even higher rates if it is first to market with an autonomous service, undercutting Uber and Lyft prices in the $2-3 range per mile.   

Autonomy Day highlighted Tesla’s data advantage over competitors. With nearly 400,000 cars on the road, Tesla is collecting billions of miles of driving data and identifying corner cases and other conditions that its artificial intelligence (AI) experts are harnessing to train its fleet. With only a few hundred cars, other companies pursuing autonomous strategies are years away from collecting enough data to train their cars.

Venmo’s User Base is Overtaking That of Banks

Follow Max on Twitter @mfriedrichARK

Venmo is the second largest digital finance platform in the US, based on new user numbers released by PayPal this week, as shown in the chart below. As of the first quarter, the peer-to-peer Digital Wallet had 40 million active users, second only to JP Morgan Chase’s 49 million digital customers. Square’s Cash App ranks number six with 15 million active users.

If measured by daily active users, Venmo probably would top the list. A 2017 survey showed that 43% of all Venmo users aged 18 to 29 use the app daily.

Thanks to strong network effects and low customer acquisition costs, Digital Wallets like Venmo and Cash App have been growing at an exponential rate. Digital Wallets can acquire customers for as little as $20, whereas traditional retail banks spend between $350 and $1500 per customer, as ARK’s research has illustrated. Given high engagement, low costs, and superior data collection and insights, Digital Wallets could displace banks in consumer finance during the next five to ten years.

Some Crypto Exchanges Are Under Scrutiny 

Follow Yassine on Twitter @yassineARK

Hong Kong based cryptocurrency exchange Bitfinex has come under scrutiny after The New York Attorney General’s (NYAGs) office released a report alleging that it had lost $850 million and had covered the shortfall by shifting capital from its affiliated stablecoin operator, Tether.

The NYAG obtained a court order against iFinex, the parent company of both Bitfinex and Tether, to investigate the “fraud being carried out by Bitfinex and Tether.” The document revealed that Bitfinex covered its losses by transferring money from Tether’s reserve funds. Specifically, Bitfinex drained $700 million worth of Tether reserves and commingled client funds with those of Panama-based payment processor, Crypto Capital. Reportedly, Poland, Portugal, and the US seized $851 million from Crypto Capital. 

In response to the allegations, Bitfinex released a statement that “both Bitfinex and Tether are financially strong – full stop. Bitfinex and Tether will vigorously challenge this, and any and all other actions, by the New York Attorney General’s office.”

Crypto traders beware.

High Systematic Errors Continue to Handicap Nanopore Sequencers

Follow Simon on Twitter @sbarnettARK

Researchers analyze the genome to connect mutations to the diseases they cause through a process called variant classification. Classifying mutations is a necessary step in creating accurate diagnostics for oncology and rare diseases.

Unfortunately, stretches of DNA that contain repetitive base sequences are difficult to study with contemporary, short-read instruments. Long-read instruments are better suited to analyze mutations that are near or in repetitive regions of the genome.

While Pacific Biosciences’ (PACB’s) long-read technology currently is the frontrunner, Oxford Nanopore could be a disruptor in the space. Nanopore sequencers use a different paradigm to interrogate DNA sequences. While PacBio’s instruments match colored light flashes to base pairs of DNA, nanopore instruments measure changes in electric voltage across a microscopic channel as DNA strands pass through.

While all long-read machines have relatively high error rates, PacBio and nanopore sequencers produce different types of error. PACB’s instruments generate random errors, meaning that incorrect base calls are distributed evenly throughout the genome. The solution to this kind of error is running the DNA through the machine a multiple of times to reach a consensus. Conversely, nanopore instruments have systematic error, meaning that incorrect base calls are concentrated in specific genomic regions. Unfortunately, more DNA reads or error correction software cannot fix systematic errors on the base layer.

Interestingly, nanopore errors occur most often in regions of DNA that are crucial for variant classification. Repetitive areas of the genome create subtle signals, making them difficult for nanopore machines to flag correctly. Consequently, Oxford Nanopore’s sequencing solutions seem ill-suited for variant classification.