The Other Side of The Deep Tech Coin/ Implantable Neurograins as BMI/ Waymo Is 99% There, The Last 1% is The Hardest/ The Metaverse Value-Chain
Massimo Portincaso
Founder & CEO at Arsenale Bioyards, Industrial Romantic and Antidisciplinarian Stoic
The other side of the deep tech coin. I have been listing through this newsletter the many opportunities arising from the dawn of deep tech as a new approach to innovation, and I have been pointing, every week, to new applications and possibilities. Below you have this week’s “dose” with flexible fabric enabling bridges, lab-grown everything, synthetic biology enabled indigo dying, the multiverse and neurograins as the next brain-machine interface. There is a lot going on. This is in fact the essence of the deep tech approach to innovation: an exploded option space, with collapsing timelines.
At the same time, I reported a few weeks ago about the implosion of Zymergen, one of the poster children of synthetic biology, this week a fairly bearish article on Ginkgo (the other poster child of synthetic biology) was published in MIT Technology Review, the autonomous driving company Waymo is struggling to get the last 1% for self-driving vehicles right (see below) and taking much longer than expected.
Under normal circumstances, I would say that we are heading toward a “productive bubble”, something intrinsic to innovation in a capitalist environment, as Bill Janeway has thought us. The volume of investment in deep tech is soaring (often “dressed” as climate investment), in a market that is flooded with capital, and there is a race to find the best investments, with valuations that are not always justified using normal parameters. The perfect bubble storm brewing, so to say.
But the problem is that we are not under normal circumstances, we have a desperate necessity to redo the economic and industrial tissue of modern civilization to have any chance to avoid making the planet inhabitable. And we need to do this quickly, very quickly. We cannot afford to have a bubble burst and then go through a deep-tech winter. Unfortunately, we don’t have that time anymore.
Is there an alternative? Yes. One of the core characteristics of this wave of innovation is that it is driven by the powerful combination of bits and atoms. And after more than two decades of innovation driven by bits, we need to realize that atoms do not scale the same way of bits and that scaling atoms must comply with the laws of physics and nature. To avoid that the deep tech bubble is triggered, all companies (and their investors) need to recognize the difficulty of scaling up production, and with it the engineering and cost challenge ahead of them.
Differently than with digital, where the disruption was very often based on a marginal cost close to zero, thus providing an intrinsic cost advantage to the disruptors, with deep tech companies are very often challenging incumbents that have a cost structure optimized over decades, at times even more than a century, and who can rely on a set physical infrastructure with a certain scale, which makes the cost comparison even more problematic.
The only chance to make it is to create value for consumers, not going for mere substitution, and at the same time take the engineering challenge extremely seriously, so to make sure that the added value is delivered, at a cost that is competitive.
I have been personally investing a lot of time and effort, over the past years, in making people aware of the possibilities that come with a deep tech approach to innovation. It is time now to emphasize the other side of the deep tech coin. The one linked to scaling up, to engineering, and driving down the cost curves, it is now time to start focusing on execution. As we will not be able to redo the economic and industrial tissue of modern civilization with “proof of concepts” or failed scaleups.
If you send rice or sand through a funnel, the grains jam up if you pour too quickly. Researchers recently leveraged this "jamming" process to create a fabric that could stiffen to 25x its original state. They connected together flexible octahedronal subunits that were able to become immobile once pressure was applied, similar to how medieval chain mail was flexible enough to wear but strong enough to deflect sword jabs and arrows. Such a material could be employed in a number of technical applications, such as easy-to-transport construction material for temporary buildings and bridges.
"Using computer simulations, the group examined how jamming takes place at the level of individual components in the fabric and which factors influence these mechanical properties. In addition, the researchers constructed similar fabrics from differently shaped building blocks. These materials contained the rings of classic chain mail but also octahedrons with different angles or fewer cross struts. The team came across a consistent property in such fabrics: how much they stiffen under stress depends solely on how many contacts the multi shaped elements make with one another on average."
News items:
Wood, human blood, meat, and dairy can all be potentially grown in a lab. What else is in our lab-grown future?
A company using a bio-based approach to making more environmentally friendly indigo dye is the first small startup to sign up with Ginkgo Biowork's automated synthetic biology platform.
Waymo is the leader in the autonomous vehicle space, but it's number one in a sluggish and increasingly more difficult race. Recently, the company saw an outflow of talent, including its CEO, CFO, and heads of its trucking, manufacturing, and partnerships teams. After a decade of R&D, testing, and billions poured in from investors and Alphabet, the company has given tens of thousands of driverless rides in Arizona.
By all accounts, self-driving taxis are 99% of the way there, but that remaining 1% has been found to be a frustrating impasse: "Small disturbances like construction crews, bicyclists, left turns, and pedestrians remain headaches for computer drivers. Each city poses new, unique challenges, and right now, no driverless car from any company can gracefully handle rain, sleet, or snow. Until these last few details are worked out, widespread commercialization of fully autonomous vehicles is all but impossible."
Former CEO John Krafcik once remarked that there's not a whole lot involved in fitting existing vehicles with self-driving tech. In fact, the process isn't simple at all: "Engineers must take apart the cars and put them back together by hand. One misplaced wire can leave engineers puzzling for days over where the problem is, according to a person familiar with the operations." Further, the process has "left Waymo without a viable path to mass production."
News items:
Google's new search language model called MUM may eventually behave as a virtual research assistant.
Recommended by LinkedIn
In machine learning, more (clean) data usually means better accuracy during inference. The same can be said for BCIs - the more neurons you can measure, the better you can translate thoughts into actions. Neuralink's wireless1,024-channel electrode arrays are a step change, but then again, the brain has something like 86B neurons. The more points in the brain you can probe, the better the data. In a recent Nature study, researchers used 48 "neurograins" to record activity in a rat brain. Eventually, they say the tech could scale to thousands of units in a single brain. These independent, wireless sensors - each the size of a grain of salt - can record and stimulate brain activity, a change of pace from the "little beds of needles" of current and past devices:
"The second challenge was developing the body-external communications hub that receives signals from those tiny chips. The device is a thin patch, about the size of a thumbprint, that attaches to the scalp outside the skull. It works like a miniature cellular phone tower, employing a network protocol to coordinate the signals from the neurograins, each of which has its own network address. The patch also supplies power wirelessly to the neurograins, which are designed to operate using a minimal amount of electricity."
News items:
A team of Indian researchers has developed a new teeth authentication app for mobile devices. The DeepTeeth Android app operates like a more standard facial recognition system but focuses on the user’s mouth rather than the entire face.
If you haven't read Epic CEO Tim Sweeney's interview on the economy of the metaverse, it's a great preamble to this piece, which explores the concept's wide value chain. Beamable founder Jon Radoff describes the seven layers of the metaverse - experience, discovery, creator economy, spatial community, decentralization, human interface, and infrastructure. This stack has enormous breadth, each layer an industry worth billions on its own. We've covered many of these over the last months, though we haven't touched on the infrastructure part much.
"Enabling the untethered functionality, high performance, and miniaturization required by the next generation of mobile devices, smart glasses, and wearables will require increasingly powerful and tinier hardware: semiconductors that are imminently dropping to 3nm processes and beyond; microelectromechanical systems (MEMS) that enable tiny sensors; and compact, long-lasting batteries."
News items:
Facebook's new virtual reality app brings Oculus users into the "infinite office."
Most researchers saw nothing wrong in scraping the internet for images and text as a way to compile training datasets. That sentiment has changed as of late. Microsoft's MS-Celeb-1M dataset, for example, contained 10M images of 100k celebrity faces, gathered without consent. After facing some intense criticism, Microsoft removed MS-Celeb-1M from its websites, but - along with other notorious datasets - it continues to live in perpetuity online or as part of the AI models it helped train.
According to Kenny Peng, author of a recently released study on the subject, deleted datasets linger on a list of websites "more expansive than we would've initially thought." According to Margaret Mitchell, who was recently fired from her role as the head of AI ethics at Google: "Data set collection and monitoring isn't a one-off task for one or two people. If you're doing this responsibly, it breaks down into a ton of different tasks that require deep thinking, deep expertise, and a variety of different people."
News items:
New research by Northeastern neuroscientists Lisa Feldman Barrett shows that interpreting a person’s facial expression can’t be done in a vacuum; it depends on the context.
Program Executive at QurAlis | Solvandria Foundation Co-Founder | MIT Research Affiliate (PhD) | Caltech Alum (BS)
3yAgree with your insights on Deep Tech's "next wave" - we need more focus on execution.