ark disrupt banner

ARK Disrupt Issue 113: EVs, LiDAR, AI, & Gene Editing

Please enjoy ARK Disrupt Issue 113. This blog series is based on ARK Brainstorming, a weekly discussion between our CEO, Director of Research, thematic analysts, ARK’s theme developers, thought leaders, and investors. It is designed to present you with the most recent innovation takeaways and to keep you engaged in an ongoing discussion on investing in disruptive innovation. To read the previous issue, click here.

1. EV Forecasts Are Falling Behind Auto Manufacturers’ Capital Spending Commitments

Follow @skorusARK on Twitter

For the past three years, electric vehicle (EV) forecasts have increased significantly, reaching nearly 4 million units on average for 2022. According to ARK’s research, global EV sales will be 4 times higher, roughly 17 million units, in 2022.

ARK bases its forecast on the declining cost curve of lithium-ion battery pack systems, which should drive down the price of electric vehicles. While we first forecast 17 million units in 2016, auto manufacturers did not have plans at that time for such a burst of activity, something that started to change in the last year: in January, Reuters reported that at least $90 billion was slated to be invested in EVs.

What does $90 billion suggest for future EV production capacity? Based on historical data of U.S. capital efficiency in the auto industry, investment approximating $90 billion would produce more than 6 million EVs globally per year. Yet, as shown below FordF, TeslaTSLA, and others have asserted that EV factories could be twice as efficient as those producing autos powered by internal combustion engines. If so, then $90 billion should be enough to produce at least 12 million EVs.

In either case, consensus forecasts for EV sales are ignoring the investments to which auto manufacturers already have committed. ARK anticipates that they will continue to ramp capital spending as the magnitude of the EV opportunity comes into sharper focus during the next few years. As a result, EV forecasts for 2022 probably will quadruple during the next two to three years.

ARK Disrupt Issue 113 graph

Source: ARK Investment Management LLC, Ford’s President of Global Operations

2. Has Waymo Fallen Out of Love with LiDAR?

Follow @tashaARK on Twitter

Recently Sacha Arnoud, Director of Engineering at WaymoGOOG, lectured at MIT’s course on Deep Learning for Self Driving Cars. ARK found the slide below in his presentation particularly interesting:

ARK Disrupt Issue 113

Source: https://www.youtube.com/watch?v=LSX3qdy0dFg&index=6&list=PLrAXtmErZgOeiKm4sgNOknGvNjby9efdf

In that slide, radar features prominently while LiDAR, in smaller and lighter gray font, pales by comparison. For years, the Google car project heralded LiDAR as critical to fully autonomous vehicles, so much so that it developed its own LiDAR sensor, one important enough to spark a lawsuit when its inventor, Anthony Levandowski, defected to Uber.

Why, then, has such an important component lost its standing in presentations such as the one for MIT? Is Google now de-emphasizing LiDAR?

Previously asserting that it would be crucial, ARK was surprised when Tesla did not include LiDAR among its sensors for fully autonomous vehicles, favoring instead radar and cameras to feed its on-board computing system. On Tesla’s last earnings call, Elon Musk offered somewhat of an explanation:

In my view, it is a crutch that will drive companies to a local maximum that they will find very difficult to get out of.“

While LiDAR can detect the presence and location of an object, it does not recognize images and is much more expensive than cameras and radar. As important, perhaps, no automotive grade LiDAR system is ready for mass production. Startups developing solid state LiDAR systems like Quanergy are not ready for prime time.

Why is Waymo downplaying LiDAR? Perhaps it is preparing to phase LiDAR out of its autonomous sensor suite before launching a large scale autonomous taxi service in the next year or so. Perhaps Tesla has been right all along.

3. It’s Time to Prepare for Malicious AI Use Cases

Follow @jwangARK on Twitter

While threats of killer robots have been overblown, artificial intelligence (AI) could serve some serious nefarious activities in the coming years. A paper released this week details all the ways that AI could be misused today and offers some advice for researchers and public policy experts.

Malicious AI applications that are feasible today include:

  • using deep learning to create realistic impersonations of anyone, in videos or over the phone.
  • using commercial drones and computer-vision equipped ‘drone swarms’ to drop bombs.
  • using facial recognition for massive video surveillance of a country’s citizens, which already is happening in China.

The key takeaway from this paper is that the risks associated with AI are serious. In 2016 and 2017, we saw how social platforms built for sharing cat videos were hijacked to disseminate fake news and sow social discord. Like social networks, AI will be used for both good and evil purposes but, hopefully, half of the solution is understanding the problem.

4. A Chimeric DNA/RNA-Guided CRISPR-Cas9 Could Reduce Off-Target Activity

Follow @msamyARK on Twitter

Increasing the safety of CRISPR gene-editing is critical now that the first in vivo applications will be entering human trials in the US this year. In a study published in Nature this week, scientists at MIT claim that they have been able to solve safety issues by replacing elements of the CRISPR-Cas9 guide RNA with DNA. The first CRISPR-Cas9 synthetic complexes were prone to unintended genomic alterations, known as off-target effects, creating permanent DNA alterations in areas of the genome other than the targeted sequence.

In the new rendition of CRISPR-Cas9, scientists decrease the Cas9 binding affinity by changing the chemical structure of the guide. While the earlier guide sequence was composed of RNA, the new guide swaps out individual RNA nucleotides for DNA-nucleotides, leading to a more stable protein complex and a cheaper manufacturing process. In the study, off-target effects dropped meaningfully while the editing efficiency of the classic CRISPR-Cas9 system remained intact.

In a separate study in pre-print, a different group of MIT scientists used a chimeric guide-DNA/RNA to reduce off-target effects and achieved 74% of on-target activity. The chimeric guide-RNA lowered the tolerance of DNA mismatches, leading to improved performance.

The question for the foundational CRISPR patent holders – IntelliaNTLA, EditasEDIT, and CRISPR TherapeuticsCRSP – is whether or not their intellectual property covers these new renditions, as the original patents focused on a RNA-mediated guide. Blanket language such as “nucleic acid” could refer to either RNA or DNA.


ARK's statements are not an endorsement of any company or a recommendation to buy, sell or hold any security. For a list of all purchases and sales made by ARK for client accounts during the past year that could be considered by the SEC as recommendations, click here. It should not be assumed that recommendations made in the future will be profitable or will equal the performance of the securities in this list. For full disclosures, click here.