#253: The OCC Has Given Banks the Green Light to Use Public Blockchain Networks and Stablecoins, & More

On Monday, the Office of the Comptroller of the Currency (OCC) published an interpretive letter addressing INVNs – independent node verification networks, or public blockchains – and stablecoins. In the letter, the OCC announced that financial institutions will be able to use INVNs to streamline payment functions like processing, validation, and settlement.

As a result, national banks and federal savings associations can participate as nodes on public blockchains. As highlighted in the letter, “The OCC has recognized that bank-permissible activities may be conducted with new and evolving technologies. INVNs and related stablecoins represent new technological means of carrying out bank-permissible payment activities and therefore conclude that a bank may validate, store, and record payment transactions by serving as a node on an INVN.”

In our view, Bitcoin could become the default settlement system for banks and businesses. Unlike traditional settlement systems, Bitcoin’s network is global and cannot censor transactions. Perhaps more important, unlike central banks, Bitcoin’s network cannot inflate the number of bitcoin outstanding beyond the mathematically metered increases put in place at its inception. If it were to capture 10% of financial intermediary settlement volumes just in the United States, our analysis suggests that Bitcoin’s network value would increase by an additional $1.5 trillion and the bitcoin price would rise by roughly $75,000.

OpenAI has demonstrated a new generative network called DALL•E that creates realistic images based on text descriptions. Trained from a dataset of text-image pairs, according to OpenAI, DALL•E can “creat[e] anthropomorphized versions of animals and objects, combin[e] unrelated concepts in plausible ways, render text, and apply transformations to existing images.”

While text-to-image networks are not new, OpenAI’s implementation shows a remarkable understanding of nuances in language and composition. Not only can it draw avocados, dogs, and radishes, but it can establish relationships and convey “emotions”. At 12 billion parameters, the DALL•E model is only one tenth the size of GPT-3, an autoregressive language model, making training and deployment that much easier.

While automation discussion and debate today focus primarily on repetitive physical and mental work, generative networks will be the first tool to automate creative work in the future. “Creative” work can be painstaking, with much trial and error. In our view, generative AI will enable many more creative careers that will be easier and perhaps more enjoyable in the future.

Quibi, the failed mobile-only streaming service which hit a $1.75 billion valuation at its peak, sold its 75-show content library to Roku for less than $100 million. Leading the US smart TV operating system (OS) market with more than 51 million households, Roku plans to add Quibi’s content to its growing ad-supported streaming service – the Roku Channel.

Offering 40,000 on-demand titles and 150 free analog television channels, the Roku Channel reaches more than 61 million households today. Missing from the Roku Channel has been made-for-mobile content. Quibi’s mobile-friendly content should round out the Roku Channel’s content library and attract mobile viewers while offering leads for conversion to Roku TV. In our view, the Quibi acquisition seems to be much more than a content deal: It also is a savvy customer acquisition strategy.

This week, after news broke that Hyundai was in talks with Apple to build a car for Project Titan, Hyundai quickly backed away from its original statement. Apple’s decision to build both hardware and software for autonomous electric vehicles is a pivot back to its original strategy after its first pivot to a software-only strategy in 2016. Recognizing that autonomous driving is an extremely complex artificial intelligence (AI) problem, Apple recently put Project Titan under its top AI expert, John Giannandrea.

Although Apple is dedicating its top talent now and plans to launch an autonomous car in 2024, Project Titan has been laboring for six years with little success apparent relative to the competition. Meanwhile, Tesla seems poised to launch a fully autonomous driving service within the next two years, and Waymo already has commercialized its robotaxi pilot in Phoenix.

Apple could partner with a company like Hyundai, but traditional auto manufacturers have struggled to build electric vehicles that can compete with Tesla on performance and price and Waymo has demonstrated the painstaking difficulty associated with fully autonomous driving. Moreover, based on the number of real-world driving miles necessary to inform AI engines, we believe the robotaxi market is likely to submit to natural geographic monopolies, lowering further the odds of success for late entrants.

Pfizer/BioNTech’s and Moderna’s mRNA vaccines against COVID-19 showed approximately 95% efficacy, raising the bar for companies still developing vaccines.

Recently, Arcturus Therapeutics (ARCT) announced Phase I/II data for its single and prime boost COVID-19 vaccines that showed a weaker humoral response but a stronger cellular response than those for the two-dose Pfizer/BioNTech and Moderna regimens. The level of neutralizing antibodies (NAbs) measures the humoral response while the level of T-cells measures the cellular response. Pfizer/BioNTech’s and Moderna’s vaccines generated NAbs two to three times that of convalescent sera and have ~95% efficacy with little evidence of reinfection thus far. Arcturus’s data showed a strong T cell response but moderate NAbs.

Now the debate will shift to the relative importance of humoral and cellular responses in producing strong and durable SARS-CoV-2 protection. We believe that if Arcturus can show comparable efficacy in its differentiated self-amplifying mRNA vaccine, its one-shot, low dose vaccine should enjoy considerable success.

In South Africa and the UK, scientists documented two SARS-CoV-2 variants that are highly contagious, though perhaps not as deadly. The variants have multiple mutations in their spike or S glycoproteins. One mutation they have in common is N501Y, which is located where receptor binding takes place, potentially threatening the efficacy of vaccines.

Based on data from a Pfizer and University of Texas Medical Branch in vitro SARS-CoV-2 mutation study, Pfizer/BioNTech have allayed some of the fears associated with these mutations. Although the study did not include all the spike mutations found in the United Kingdom and South African variants, it did include N501Y which did not resist the Pfizer/BioNTech vaccine.