All posts in “Science”

The five technical challenges Cerebras overcame in building the first trillion transistor chip

Superlatives abound at Cerebras, the until-today stealthy next-generation silicon chip company looking to make training a deep learning model as quick as buying toothpaste from Amazon. Launching after almost three years of quiet development, Cerebras introduced its new chip today — and it is a doozy. The “Wafer Scale Engine” is 1.2 trillion transistors (the most ever), 46,225 square millimeters (the largest ever), and includes 18 gigabytes of on-chip memory (the most of any chip on the market today) and 400,000 processing cores (guess the superlative).

CS Wafer Keyboard Comparison

Cerebras’ Wafer Scale Engine is larger than a typical Mac keyboard (via Cerebras Systems)

It’s made a big splash here at Stanford University at the Hot Chips conference, one of the silicon industry’s big confabs for product introductions and roadmaps, with various levels of oohs and aahs among attendees. You can read more about the chip from Tiernan Ray at Fortune and read the white paper from Cerebras itself.

Superlatives aside though, the technical challenges that Cerebras had to overcome to reach this milestone I think is the more interesting story here. I sat down with founder and CEO Andrew Feldman this afternoon to discuss what his 173 engineers have been building quietly just down the street here these past few years with $112 million in venture capital funding from Benchmark and others.

Going big means nothing but challenges

First, a quick background on how the chips that power your phones and computers get made. Fabs like TSMC take standard-sized silicon wafers and divide them into individual chips by using light to etch the transistors into the chip. Wafers are circles and chips are squares, and so there is some basic geometry involved in subdividing that circle into a clear array of individual chips.

One big challenge in this lithography process is that errors can creep into the manufacturing process, requiring extensive testing to verify quality and forcing fabs to throw away poorly performing chips. The smaller and more compact the chip, the less likely any individual chip will be inoperative, and the higher the yield for the fab. Higher yield equals higher profits.

Cerebras throws out the idea of etching a bunch of individual chips onto a single wafer in lieu of just using the whole wafer itself as one gigantic chip. That allows all of those individual cores to connect with one another directly — vastly speeding up the critical feedback loops used in deep learning algorithms — but comes at the cost of huge manufacturing and design challenges to create and manage these chips.

CS Wafer Sean

Cerebras’ technical architecture and design was led by co-founder Sean Lie. Feldman and Lie worked together on a previous startup called SeaMicro, which sold to AMD in 2012 for $334 million. (Via Cerebras Systems)

The first challenge the team ran into according to Feldman was handling communication across the “scribe lines.” While Cerebras chip encompasses a full wafer, today’s lithography equipment still has to act like there are individual chips being etched into the silicon wafer. So the company had to invent new techniques to allow each of those individual chips to communicate with each other across the whole wafer. Working with TSMC, they not only invented new channels for communication, but also had to write new software to handle chips with trillion plus transistors.

The second challenge was yield. With a chip covering an entire silicon wafer, a single imperfection in the etching of that wafer could render the entire chip inoperative. This has been the block for decades on whole wafer technology: due to the laws of physics, it is essentially impossible to etch a trillion transistors with perfect accuracy repeatedly.

Cerebras approached the problem using redundancy by adding extra cores throughout the chip that would be used as backup in the event that an error appeared in that core’s neighborhood on the wafer. “You have to hold only 1%, 1.5% of these guys aside,” Feldman explained to me. Leaving extra cores allows the chip to essentially self-heal, routing around the lithography error and making a whole wafer silicon chip viable.

Entering uncharted territory in chip design

Those first two challenges — communicating across the scribe lines between chips and handling yield — have flummoxed chip designers studying whole wafer chips for decades. But they were known problems, and Feldman said that they were actually easier to solve that expected by re-approaching them using modern tools.

He likens the challenge though to climbing Mount Everest. “It’s like the first set of guys failed to climb Mount Everest, they said, ‘Shit, that first part is really hard.’ And then the next set came along and said ‘That shit was nothing. That last hundred yards, that’s a problem.’”

And indeed, the toughest challenges according to Feldman for Cerebras were the next three, since no other chip designer had gotten past the scribe line communication and yield challenges to actually find what happened next.

The third challenge Cerebras confronted was handling thermal expansion. Chips get extremely hot in operation, but different materials expand at different rates. That means the connectors tethering a chip to its motherboard also need to thermally expand at precisely the same rate lest cracks develop between the two.

Feldman said that “How do you get a connector that can withstand [that]? Nobody had ever done that before, [and so] we had to invent a material. So we have PhDs in material science, [and] we had to invent a material that could absorb some of that difference.”

Once a chip is manufactured, it needs to be tested and packaged for shipment to original equipment manufacturers (OEMs) who add the chips into the products used by end customers (whether data centers or consumer laptops). There is a challenge though: absolutely nothing on the market is designed to handle a whole-wafer chip.

CS Wafer Inspection

Cerebras designed its own testing and packaging system to handle its chip (Via Cerebras Systems)

“How on earth do you package it? Well, the answer is you invent a lot of shit. That is the truth. Nobody had a printed circuit board this size. Nobody had connectors. Nobody had a cold plate. Nobody had tools. Nobody had tools to align them. Nobody had tools to handle them. Nobody had any software to test,” Feldman explained. “And so we have designed this whole manufacturing flow, because nobody has ever done it.” Cerebras’ technology is much more than just the chip it sells — it also includes all of the associated machinery required to actually manufacture and package those chips.

Finally, all that processing power in one chip requires immense power and cooling. Cerebras’ chip uses 15 kilowatts of power to operate — a prodigious amount of power for an individual chip, although relatively comparable to a modern-sized AI cluster. All that power also needs to be cooled, and Cerebras had to design a new way to deliver both for such a large chip.

It essentially approached the problem by turning the chip on its side, in what Feldman called “using the Z-dimension.” The idea was that rather than trying to move power and cooling horizontally across the chip as is traditional, power and cooling are delivered vertically at all points across the chip, ensuring even and consistent access to both.

And so, those were the next three challenges — thermal expansion, packaging, and power/cooling — that the company has worked around-the-clock to deliver these past few years.

From theory to reality

Cerebras has a demo chip (I saw one, and yes, it is roughly the size of my head), and it has started to deliver prototypes to customers according to reports. The big challenge though as with all new chips is scaling production to meet customer demand.

For Cerebras, the situation is a bit unusual. Since it places so much computing power on one wafer, customers don’t necessarily need to buy dozens or hundreds of chips and stitch them together to create a compute cluster. Instead, they may only need a handful of Cerebras chips for their deep-learning needs. The company’s next major phase is to reach scale and ensure a steady delivery of its chips, which it packages as a whole system “appliance” that also includes its proprietary cooling technology.

Expect to hear more details of Cerebras technology in the coming months, particularly as the fight over the future of deep learning processing workflows continues to heat up.

Flexible stick-on sensors could wirelessly monitor your sweat and pulse

As people strive ever harder to minutely quantify every action they do, the sensors that monitor those actions are growing lighter and less invasive. Two prototype sensors from crosstown rivals Stanford and Berkeley stick right to the skin and provide a wealth of phsyiological data.

Stanford’s stretchy wireless “BodyNet” isn’t just flexible in order to survive being worn on the shifting surface of the body; that flexing is where its data comes from.

The sensor is made of metallic ink laid on top of a flexible material like that in an adhesive bandage. But unlike phones and smart watches, which use tiny accelerometers or optical tricks to track the body, this system relies on how it is itself stretched and compressed. These movements cause tiny changes in how electricity passes through the ink, changes that are relayed to a processor nearby.

Naturally if one is placed on a joint, as some of these electronic stickers were, it can report back whether and how much that joint has been flexed. But the system is sensitive enough that it can also detect the slight changes the skin experiences during each heartbeat, or the broader changes that accompany breathing.

The problem comes when you have to get that signal off the skin. Using a wire is annoying and definitely very ’90s. But antennas don’t work well when they’re flexed in weird directions — efficiency drops off a cliff, and there’s very little power to begin with — the skin sensor is powered by harvesting RFID signals, a technique that renders very little in the way of voltage.

bodynet sticker and receiver

The second part of their work, then, and the part that is clearly most in need of further improvement and miniaturization, is the receiver, which collects and re-transmits the sensor’s signal to a phone or other device. Although they managed to create a unit that’s light enough to be clipped to clothes, it’s still not the kind of thing you’d want to wear to the gym.

The good news is that’s an engineering and design limitation, not a theoretical one — so a couple years of work and progress on the electronics front and they could have a much more attractive system.

“We think one day it will be possible to create a full-body skin-sensor array to collect physiological data without interfering with a person’s normal behavior,” Stanford professor Zhenan Bao in a news release.

Over at Cal is a project in a similar domain that’s working to get from prototype to production. Researchers there have been working on a sweat monitor for a few years that could detect a number of physiological factors.

SensorOnForehead BN

Normally you’d just collect sweat every 15 minutes or so and analyze each batch separately. But that doesn’t really give you very good temporal resolution — what if you want to know how the sweat changes minute by minute or less? By putting the sweat collection and analysis systems together right on the skin, you can do just that.

While the sensor has  been in the works for a while, it’s only recently that the team has started moving towards user testing at scale to see what exactly sweat measurements have to offer.

RollToRoll BN 768x960“The goal of the project is not just to make the sensors but start to do many subject studies and see what sweat tells us — I always say ‘decoding’ sweat composition. For that we need sensors that are reliable, reproducible, and that we can fabricate to scale so that we can put multiple sensors in different spots of the body and put them on many subjects,” explained Ali Javey, Berkeley professor and head of the project.

As anyone who’s working in hardware will tell you, going from a hand-built prototype to a mass-produced model is a huge challenge. So the Berkeley team tapped their Finnish friends at VTT Technical Research Center, who make a specialty of roll-to-roll printing.

For flat, relatively simple electronics, roll-to-roll is a great technique, essentially printing the sensors right onto a flexible plastic substrate that can then simply be cut to size. This way they can make hundreds or thousands of the sensors quickly and cheaply, making them much simpler to deploy at arbitrary scales.

These are far from the only flexible or skin-mounted electronics projects out there, but it’s clear that we’re approaching the point when they begin to leave the lab and head out to hospitals, gyms, and homes.

The paper describing Stanford’s flexible sensor appeared this week in the journal Nature Electronics, while Berkeley’s sweat tracker was in Science Advances.

These robo-shorts are the precursor to a true soft exoskeleton

When someone says “robotic exoskeleton,” the power loaders from Aliens are what come to mind for most people (or at least me), but the real things will be much different: softer, smarter, and used for much more ordinary tasks. The latest such exo from Harvard is so low-profile you could wear it around the house.

Designed by researchers at Harvard’s Wyss Institute (in collaboration with several other institutions), which focuses on soft robotics and bio-inspired mechanisms, the exosuit isn’t for heavy lifting or combating xenomorphs but simply walking and running a little bit more easily.

The suit, which is really more of a pair of shorts with a mechanism attached at the lower back and cables going to straps on the legs, is intended to simply assist the leg in its hip-extension movement, common to most forms of locomotion.

An onboard computer (and neural network, naturally) detects the movements of the wearer’s body and determines both the type of gait (walking or running) and what phase of that gait the leg is currently in. It gives the leg making the movement a little boost, making it just that much easier to do it.

[embedded content]

In testing, the suit reduced the metabolic load of walking by 9.3 percent and running by 4 percent. That might not sound like much, but they weren’t looking to create an Olympic-quality cyborg — just show reliable gains from a soft, portable exosuit.

“While the metabolic reductions we found are modest, our study demonstrates that it is possible to have a portable wearable robot assist more than just a single activity, helping to pave the way for these systems to become ubiquitous in our lives,” said lead study author Conor Walsh in a news release.

The whole idea, then, is to leave behind the idea of an exosuit as a big mechanical thing for heavy industry or work, and bring in the idea that one could help an elderly person stand up from a chair, or someone recovering from an accident walk farther without fatigue.

rt scitoc aug16 r1

The whole device, shorts and all, weighs about 5 kilograms, or 11 pounds. Most of that is in the little battery and motor pack stashed at the top of the shorts, near the body’s center of mass, helping it feel lighter than it is.

Of course this is the kind of thing the military is very interested in — not just for active duty (a soldier who can run twice as far or fast) but for treatment of the wounded. So it shouldn’t be a surprise that this came out of a DARPA project initiated years ago (and ongoing in other forms).

But by far the more promising applications are civilian, in the medical field and beyond. “We are excited to continue to apply it to a range of applications, including assisting those with gait impairments, industry workers at risk of injury performing physically strenuous tasks, or recreational weekend warriors,” said Walsh.

Currently the team is hard at work improving the robo-shorts, reducing the weight, making the assistance more powerful and more intuitive, and so on. The paper describing their system was the cover story of this week’s edition of the journal Science.

Elysium and the quest to bottle the fountain of youth

In the latest episode of Flux podcast, I sit down with Eric Marcotulli, the co-founder of Elysium, a life sciences company developing consumer-facing health products based on aging research. The company’s first product is Basis, a supplement that combines compounds designed to increase NAD levels and activate sirtuins, boosting cellular health and longevity.

In this conversation we discuss why precursor companies failed, including Cambridge-based Sirtris Pharmaceuticals, which was bought for $720 million in 2008. Eric explains how Elysium is a platform-based company that will sell a host of products and diagnostics, why he believes direct to consumer is the best market strategy, and what the current user base looks like. The company just announced a new clinical trial this week. Eric gets into the importance of bringing academic rigor and peer review to the supplement category, how he plans to build consumer trust and ultimately pull it into the mainstream. He shares why he believes in open source research, how cellular senescence is a particular area of interest right now, what his personal health routine is and how he thinks about the singularity.

An excerpt of our conversation is published below. Full transcript on Medium.

Eric Marcotulli—Bottling the Fountain of Youth

ALG: Welcome everyone to the latest episode of Flux. I’m excited to have Eric Marcotulli here today. He is the co-founder and CEO of Elysium Health a company that is rethinking healthcare whose first product is a science based supplement that promotes cellular health. Welcome.

EM: Thank you.

ALG: Appreciate it. I’ve been excited about your company for a long time. It’s nice to meet in person.

EM: Likewise.

ALG: As a New York based VC it’s also great to meet New York based companies, especially science focused companies. I’d love to start by hearing the beginnings of Elysium. You started the company in 2014. I talked to one of your investors last summer, he said to ask you the story of how you met your co-founder at Equinox — is that true?

EM: So there’s two co-founders and they both have their own stories. I’ll start with my scientific co-founder, Leonard Guarente. Leonard’s run the biology of aging Lab at MIT for the last 25 to 30 years. I didn’t set out to become an entrepreneur. It was a confluence of events.

If you go back to 2011, 2012 I was in business school and in one of my classes we were studying a company that in the field of aging is well known but outside is not. It’s called Sirtris Pharmaceuticals and it was a Boston based biotech company going back to the mid 2000s. What’s interesting about this company is they were studying processes of aging and they identified one in particular. It was a class of genes that we now affectionately refer to as longevity genes. They’re called sirtuins. What they identified was that these genes are found in every living thing and that the activity level of these genes decreases through the normal course of aging. And when they reactivated these genes they saw amazing benefits, regardless of which model or organism you were looking at. They would live to the human equivalent of 120 years old. They don’t get cancer. They don’t gain weight.

ALG: You mean the mice?

EM: That’s right. Life is pretty good at this point if you’re a mouse. It was a monumental discovery in terms of aging research and it got the researchers nominated for the Nobel Prize. One of the researchers involved in that series of discoveries in the late 90s early 2000s went off to Harvard to open his own lab. And being more risk-seeking he was screening for natural molecules that could potentially activate these genes. The hypothesis there, which has since carried over to Elysium, is that aging itself isn’t a disease. It’s about the interconnected degradation or failure of our own biological processes and metabolisms. There is a prevailing hypothesis we are seeing develop that natural compounds will be the most effective interventions. That was the approach taken there, and the researcher’s name is David Sinclair. He screened natural product libraries for potential hits that could activate these genes.

And he found one, a derivative of red wine called resveratrol. Some people have heard of this. If you look back at the ‘04 ‘05 timeframe there was a massive spike in red wine sales due to all the media coverage around it. So they started a company, Sirtris. So they make you the protagonist in this case study, and you have to make a decision as the management of the company. What was interesting was that you have a natural product and that aging isn’t a disease. To try and create a traditional pharmaceutical company and go after diseases you’d be trying to fit a square peg in a round hole. You’d likely have to modify the molecule, you would have to start looking at disease models. On the other hand, you could build a direct to consumer company, where you don’t have to modify the molecule, and aging isn’t a disease so you don’t have to go through a laborious long-term and huge cost effort from an approval standpoint. We debated the merits of both of these business models. I was firmly in the camp of the consumer facing effort, because I was reading this research and saying, how could anybody sitting in this room not want this for themselves or their parents or their friends?

ALG: Right.

EM: It ends up not mattering which position you take. [In 2008] GlaxoSmithKline stepped in and bought the company for almost three quarters of a billion in cash, before they read any human data. I was fascinated at this point with the research. If you had played word association with me going into that class and you said “anti-aging” and “longevity” I would have just rattled off “late-night infomercials” and “snake oil.”

ALG: So that class awakened you to the industry and got you interested?

EM: That’s right. I didn’t know this was something you could study. Aging — most people think it’s an unstoppable ambiguous force. But it’s not. It’s something that we can now quantify and measure and potentially intervene. That was new to me, the fact that people at MIT and Harvard were studying this and making progress. So I left that fascinated. Shortly thereafter I reached out directly to the the the MIT professor who was the original discoverer of these genes, the sirtuins. I reached out to the scientific co-founder at Harvard. The question I had was, whatever happened to this? Because now it’s almost a decade since the acquisition. There had been little news on it. If you fast forward to today the MIT professor, the one where they made the original discoveries of the longevity genes, is now the co-founder of Elysium.

Elysium Health co-founders Eric Marcotulli, Dan Alminana, Leonard Guarente.

ALG: Leonard?

EM: Yes. Dr. Guarente. Or Lenny as we call him. Lenny and I just started off with conversations around how the research had progressed. At one point Lenny called me and said, “I’ve been approached by a Japanese venture capitalist who has invested in a company in Taiwan. They believe they have a potent sirtuin activating compound that’s very different from the resveratrol molecule.” He said, “I know nothing about the business side of things. We’re dealing with a venture capitalist and you’re a venture capitalist.” I was at the time — before business school I was at Bain Capital Ventures, after I was at Sequoia. He said, “would you want to go with me to look at this potential molecule?”

I said, “I don’t know how much help I can be but I’m happy to go with you.” So Lenny and I met for the first time six months after our first phone call. This was late 2012. We met in San Francisco International Airport and went to Taiwan for a few days. It was a fascinating experience. We ultimately passed on that molecule despite some interesting research. But it was through that that Lenny and I came up with this idea that you could build a direct to consumer facing effort and that there would be more of these types of products, that it wasn’t limited to a single product idea. So this was the vision for creating a platform-based company.

ALG: So that’s how you met him. And it sounds like you explored a couple of different routes in terms of what molecules could stimulate sirtuins right?

EM: That’s right.

ALG: And the one you ultimately went for first is NAD?

EM: So there’s two components to the product we have today, which is called Basis. One of the things that Lenny and his constituencies in the research community had identified was that sirtuins are dependent on a coenzyme called NAD. We didn’t know that at the time that Sirtris was founded. It’s the production of NAD — a coenzyme, a fuel that’s used in a variety of reactions at the metabolic level — it was actually the production of this coenzyme that was decreasing in all of these living things. So NAD itself is not new. We’ve known about it for a hundred years. Two Nobel prizes have been awarded for elucidating its function. It’s important for things like DNA maintenance and repair, the creation of energy, the way the cells communicate both internally and with one another. Without NAD you’d be dead in under a minute. It’s very important. So this idea that it was decreasing, which we didn’t know until 2012, was a monumental discovery.

ALG: Decreasing over a mouse or human lifespan?

EM: Universally. Whether you’re a plant, animal, bacteria — doesn’t matter. You have NAD, you use it for these critical functions and it declines in its production in everything that ages. But not every living thing ages. Jellyfish don’t age for example.

ALG: Oh wow. What’s going on with their NAD?

EM: Well we don’t know yet. But it’s a small number [of organisms]. In everything else you see this decrease [in NAD] when the organism ages. Since we didn’t know this, Lenny said trying to activate these sirtuins would have been a failure regardless. So what we first need to do is restore levels of NAD. Then we need to activate these sirtuins. And we know that resveratrol does not work in humans. So that was another discovery that happened in the subsequent time period after the acquisition.

ALG: So all the red wine articles are baloney?

EM: Well interestingly if you drink red wine you do get the benefit of sirtuin activation. You just have to drink quite a bit of it.

ALG: I can do that.

EM: Ha most people say that. So that’s just an example of removing a high purity molecule from its natural carrier state in the wine to a pill as an example, which was a failure in humans. So we identified a cousin of resveratrol called pterostilbene [an antioxidant], which from a structural standpoint, at the molecular level, is more stable.

ALG: The stuff found in blueberries?

EM: That’s right. If you could choose one food for the rest of your life my recommendation would be blueberries. Some people would disagree with me. Maybe it’s wine. But that was the idea behind Basis. First we need to restore levels of NAD. Secondly we can then go in and activate these longevity genes. And that there would be a synergy associated with that. The best way to think about it would be sports cars. If you’re activating sirtuins it’s like putting a turbo in the engine, but the car still requires some energy source. So if there’s no gasoline, or if you own a Tesla and there’s no battery power, it’s not going to work. But once you have the two of them together there’s a supercar.

ALG: It’s actually an analogy a lot of people use for aging. I don’t know if you know Aubrey de Grey and the SENS foundation, but he talks about aging as a disease. That it’s just like a car and we need to figure out how to repair the car and the many different things that go into that. I also want to ask you more about this reframing of how we think about living longer, and how the healthcare system doesn’t consider aging a problem so far or something to tackle. How does the shift happen?

EM: Part of it is, we are going to have to deal with it regardless. Everything we’re seeing now given the advancements in medicine is somehow related to aging. If you survive cancer you are unfortunately going to die from Alzheimer’s or cardiovascular disease or Type 2 diabetes. I won’t get the number exactly right, but if you cured every single form of cancer it would only add about three years of lifespan collectively, on average.

ALG: Because there’s going to be another disease that kills you.

EM: Correct. So for example, one of the areas we’re interested in is something called acute kidney injury. Thirty percent of people who go in for cardiovascular surgery will develop acute kidney injury, and with too much of it you’ll get kidney failure and dialysis. In 2004 there were just shy of a million cases in the United States of acute kidney injury. In 2014 just 10 years later there were four million cases. It’s not that our surgical techniques changed, it’s that older people are going in for these surgeries more often because our healthcare system is actually getting better. So we’re going to have to deal with these things. Two, from a diagnostic standpoint we are moving — out of necessity, at the research level — into an area where we’re now able to quantify aging. As an example there is an epigenetic test, a cheek swab or a spit tube type test, developed at UCLA in conjunction with the National Institutes of Health. It can basically tell you your biological age. The age on your passport or your driver’s license is your chronological age. But there’s something that the gene activity expression data we collect can tell you about how you are aging.

ALG: Is that a hard test to do? I’m sure everyone would want to do that if they could.

EM: So we are commercializing this test. Think of it as rings of a tree. Over time you can get a pretty accurate understanding of a single tree. How old is it. Did it go through a bountiful spring or a terrible winter. Was there a forest fire. That’s at the individual level. Then when you look at the macro level, at the forest, you can learn a lot about that particular ecosystem. It’s an oversimplification of the idea but it’s the same thing every time, looking at something called methylation. Every time one of these sites is methylated it leaves a mark. So we can quantify that and say does this intervention or product reverse, slow, or stop the aging process.

ALG: That’s a game changer for you right. Because everyone thinks it’s a good idea to take [the product], why not. But without being able to measure a result, it is hard to say. People do report feeling better in the short term — less hangovers etc. I saw one of your advisers say something about his elbows getting softer—

EM: Oh Rich.

ALG: Yeah. A lot of great side effects that sound like they are worth having anyway. Like great energy peaks.

EM: Right. As we move towards aging, we would argue that it should be classified as a disease. But today it’s difficult because it’s not a moment in time diagnosis. It’s not like one day you wake up feeling symptoms and you go to the doctor and he says, “yes we ran the tests and you have aging.” It’s a decades long accumulation of mutations and failures and other things. So these types of diagnostics have been developed by the research community out of necessity. They need to understand does this intervention actually slow, stop, or reverse aging. That’s just one measure. We’re going to see other diagnostics that take shape and form over the next several years. So to your point, that’s important for us. Because one, the conversation today has been around, OK if you can reverse this fundamental process of aging what does it mean for me? It’s one thing to say, the models that exist today could show efficacy in cancer or neurodegenerative diseases et cetera. But there’s been nothing between showing the reversal of that process and the outcome of it from a disease standpoint. So this idea that there’s a middle ground, something where you can say, well the speed at which you’re aging has changed for the better. That is an important step in the entire process.

ALG: That’s super exciting. So what does the roadmap look like? Lets get into the product so that listeners who haven’t seen it know. Out the gate, your first product was Basis, the daily supplement which is $50 a month for a subscription, or $480 for the year. And it’s recommended that you take two pills in the morning. So that’s product one. It sounds like there’s a lot of other things in the works? How do the diagnostics fit in?

EM: Sure. Going back to the hypothesis for the company — we sought early on to commercialize these technologies on a platform basis. We knew the diagnostics were coming. We knew that there would be other interventions. The third leg of the stool would be things on the digital front or the wearable front. We’re still a bit aways away from seeing the commensurate rigor in that camp. One is — as you mentioned earlier — the diagnostics are important to show the efficacy broadly speaking for these products. The other thing that’s exciting is this idea of N of 1. We’re finally going to be able to move into the realm of personalization. First, is this product working for me? Second, what is it doing for me? Third, how would I have to modify my lifestyle or administration of the product?

That’s how we think about the world from a product development standpoint, through these systems. Apple is an exaggerated example but they’ve done a fantastic job from a platform basis — of providing the app store and cloud services to integrate all the devices you have. What’s interesting about these diagnostics that we’re developing is that they are very much a subscription in nature. What you’re doing today is going to be different to what you’re doing tomorrow. Your health status may change for a whole host of reasons, genetic or otherwise. Since these aren’t just genetic tests looking at your ancestry, our hope is that 5 or 10 years in the future this is part of your annual checkup.

ALG: And there’s no negative side effects that we know?

EM: No. By and large this is one of the safest products we’ve ever seen, with all of the data that we have and millions and millions of data points in ongoing safety testing. Moving beyond that requires us to prove more digestible and accessible points of understanding. This idea of rate of aging is a step in that direction. We’re also, as an example, doing a study on photo aging of the skin based on both the existing body of literature that’s out there as well as feedback we’ve had from customers. The conversation changes a lot if I can put you under a special camera that shows the UV damage to your skin and then shows you the before and after of someone who has taken the product for six months and how it changed them. Even just showing wrinkles and things like that. Things that people are used to, from a marketing standpoint, but they might not actually see the science in it today. So there is that evolution. The evolution from, yes we can reverse this fundamental process of aging to, well what does that mean for me? Well it means it’s actually going to change the rate at which you age, which is tied to your health and all these diseases. Well OK, now I can actually tell you that it’s going to do XYZ for your skin, brain or whatever it might be.

ALG: You say upfront that this is about improving cellular function, but there’s no guarantee of longevity, though of course the name has connotations there. I’d like to ask your thoughts on longevity. There’s now things like cryo and people are signing up for places like Alcor. Do you think if we all had the ability to live forever, that that would be a good thing?

EM: So you skipped over the easier question. We’re focused on healthspan first and foremost. If you go into a room of people, and we do this all the time, and you say to the audience, “We’re going to take a quick poll. How many of you want to live forever? How many of you want to live to 120 or 150? How many of you want to live to 80?” It’s interesting to see the distribution. If you then say, “you are going to live to that age but you’re going to be as healthy as you are today, would you change your answer?” Most people do. So we are first and foremost focused on quality of life and healthspan. The belief is, ultimately if you improve every day, that you’ll have more days on the back end.

ALG: So it would be fine to go to 90 with a 20 year old’s full health, then kick the bucket. That’s more the goal.

EM: That’s exactly it. We’ve all dealt with it ourselves with loved ones. No one wants to live for another 10 or 20 years in a certain state. Usually the conversation is, what are the implications of that. From our standpoint, every time humanity has had an order of magnitude improvement in health — the introduction of antibiotics for example — I don’t think we’ve ever seen humanity broadly say, we don’t need this or we don’t want it. I do think the question changes with the singularity. Which is living forever.

ALG: We should probably do another 40 minutes on the singularity, it’s an important topic.

EM: It is. I always say without question in our lifetime we will see a merging with something digital. Musk has announced his Neurolink technology recently and is claiming to make progress on it, so it may happen. We’re not going to be able to predict when it happens. And when it happens it’s going to happen quickly. That’s dangerous for a whole host of reasons. But I’m not sure we have a good answer for, “should we?”

ALG: I guess the answer is in splitting the question into healthspan and lifespan. People are generally in agreement on healthspan. Lifespan is more of a question mark.

EM: Without question. If you talk to anybody in the aging community at the research level, we would be surprised to hear them say, let’s focus on longevity first. Everybody is actually focused on understanding its implications and its role in human health more broadly and how interventions might change that. Then of course the idea is, well if we can get rid of all these diseases of aging, you would think that you’re going to live longer too. In a higher quality state.

ALG: It does amaze me. From a personal perspective I am signed up at Alcor. Do I think it’s going to work? Not really. But I did it for other reasons—

EM: Yes. I think we as a group need to hold companies in this space to a higher standard than we have in the last 10 years in terms of these types of things. I always say when someone brings me a product — even the products that I’m interested in — What is the research behind it? Where are the studies published? What do they find? How are they designed? If you just look at the supplement portion of our business, the consumer facing interventions, it’s a $35 billion dollar market in the United States.

ALG: And that’s with the current low standards and general snake oil perceptions [in supplements.]

EM: Of course. The other thing is, if you stop someone on the street and ask, “what’s your favorite supplement company?” you’re going to get a puzzled look back. No one walks around with a hat or a bag that has one of these companies on it. That’s because they lack legitimacy. That’s a huge part of Elysium’s mission. Hiring with the rigor that you would see in life sciences on the pharma side, into the consumer market. This is part of the shift. We will see legitimate companies not just Elysium, but others in the consumer sphere, changing this conversation. The market will look like a lot of other markets as opposed to this fragmented, untrustworthy one that we see today. That evolution might take a little longer. But ultimately we’re going to end up in a place where people feel good about the products they’re buying, because only the products that work are going to survive.

The global supplements market was estimated at USD $115.06 billion in 2018. It is expected to grow at a CAGR of 7.8% in coming years. [Source]

ALG: It’s interesting because you’re a pioneer in this area of supplements. There are other supplements startups, such as prenatal which is also taking off. But there’s less controversy — people say, prenatal vitamins? Of course, why wouldn’t you take that. With yours there’s more questions. My point on Alcor and backlash was that people have strong opinions on human longevity.

EM: It’s interesting. Our category is hot right now. In a lot of these established categories — take prenatal— there’s great literature supporting the use of folic acid. There’s companies that sell products around that. But they’ll make unique claims or link to literature that’s been done by other companies on other formulations or other delivery methodologies. Those can be dangerous. The data might appear to be good but in fact their own product hasn’t been tested. We have to do it by virtue of what you highlighted, the fact that we are new. But the buyer should beware of whether this exact formulation or exact product was tested for what it’s claiming to accomplish for you.

ALG: So you’re trying to do as much as you can in-house, which includes all R&D at the moment?

EM: Yes. We have a very open source model. One of the things we did, going back two or three years, was we did a randomized, placebo-controlled, double blind study on Basis to show that it could actually restore levels of NAD. We had to show that it actually did what it did. Now that’s mechanism of action in terms of what we’re showing. We didn’t show any tangible health benefits in that particular study, it was just the reversal—

ALG: That was the 2017 study?

EM: That’s right. It was an important first step to show that.

ALG: NAD levels increased by an average of 40 percent in your users?

EM: Yes. In a one month span. Then that was sustained over a period of time after that. And it was done safely. But if a traditional pharmaceutical company had done that study they would have just internally validated that the product works, then continued their research. We chose to publish and announce it. What we found was an influx of research interest from MDs and PHDs all around the world who said, “I’m interested in NAD repletion or sirtuin activation, and I now know that your product can safely and sustainably reverse this decline. Would you be willing to work on this particular health problem with me?” So a lot of it we do internally. And a lot of it is also driven by the scientific advisory board or collaborators that approach us and say we’d love to do something with you. This idea of open source is something that’s important to us and we encourage others to pursue it as well.

ALG: A lot of exciting stuff going on there. To wrap up is there anything else you want to share about the company or what we should expect in the next six to eight months?

EM: In terms of 2019, it is our plan to launch new products in both of the categories we talked about. You’ll see new diagnostics and you’ll see new interventions from us.

ALG: Exciting. I can’t wait to see. Thanks Eric for coming on. It’s great to meet and I look forward to seeing the products when they come out.

EM: Great. Thank you.

ALG: Thank you.

$600M Cray supercomputer will tower above the rest — to build better nukes

Cray has been commissioned by Lawrence Livermore National Laboratory to create a supercomputer head and shoulders above all the rest, with the contract valued at some $600 million. Disappointingly, El Capitan, as the system will be called, will be more or less solely dedicated to redesigning our nuclear armament.

El Capitan will be the third “exascale” computer being built by Cray for the U.S. government, the other two being Aurora for Argonne National Lab and Frontier for Oak Ridge. These computers are built on a whole new architecture called Shasta, in which Cray intends to combine the speed and scale of high performance computing with the easy administration of cloud-based enterprise tools.

Due for delivery in 2022, El Capitan will be operating on the order of 1.5 exaflops, or floating point operations per second, a measure of calculation often used to track supercomputer performance. Exa denotes a quintillion of something.

Right now the top dog is already at Oak Ridge: an IBM-built system called Sierra. At about 1.5 petaflops, it’s about 1/10th the power of Aurora — of course, the former is operational and the latter is theoretical right now, but you get the idea.

One wonders exactly what all this computing power is needed for. There are in fact countless domains of science that could be advanced by access to a system like El Capitan — simulations of atmospheric and geological processes, for instance, could be simulated in 3D at a larger scale and higher fidelity than ever before.

So it was a bit disheartening to learn that El Capitan will, once fully operational, be dedicated almost solely to classified nuclear weaponry design.

To be clear, that doesn’t just mean bigger and more lethal bombs. The contract is being carried out with the collaboration of the National Nuclear Security Administration, which of course oversees the nuclear stockpile alongside the Department of Energy and military. It’s a big operation, as you might expect.

We have an aging nuclear weapons stockpile that was essentially designed and engineered over a period of decades ending in the ’90s. We may not need to build new ones, but we do actually have to keep our old ones in good shape, not just in case of war but to prevent them failing in their advancing age and decrepitude.

shasta

The components of Cray’s Shasta systems.

“We like to say that while the stockpile was designed in two dimensions, it’s actually aging in three,” said LLNL director Bill Goldstein in a teleconference call on Monday. “We’re currently redesigning both warhead and delivery system. This is the first time we’ve been doing done this for about 30 years now. This requires us to be able to simulate the interaction between the physics of the nuclear system and the engineering features of the delivery system. These are real engineering interactions and are truly 3D. This is an example of a new requirement that we have to meet, a new problem that we have to solve, and we simply can’t rely on two dimensional simulations to get at. And El Capitan is being delivered just in time to address this problem.”

Although in response to my question Goldstein declined to provide a concrete example of a 3D versus 2D research question or result, citing the classified nature of the work, it’s clear that his remarks are meant to be taken both literally and figuratively. The depth, so to speak, of factors affecting a nuclear weapons system may be said to have been much flatter in the ’90s, when we lacked the computing resources to do the complex physics simulations that might inform their design. So both conceptually and spatially the design process has expanded.

That said, let’s be clear: “warhead and delivery systems” means nukes, and that is what this $600 million supercomputer will be dedicated to.

There’s a silver lining there: Before being air-gapped and entering into its classified operations, El Capitan will have a “shakeout period” during which others will have access to it. So while for most of its life it will be hard at work on weapons systems, during its childhood it will be able to experience a wider breadth of scientific problems.

The exact period of time and who will have access to it is to be determined (this is still three years out), but it’s not an afterthought to quiet jealous researchers. The team needs to get used to the tools and work with Cray to refine the system before it moves on to the top secret stuff. And opening it up to a variety of research problems and methods is a great way to do it, while also providing a public good.

Yet Goldstein referred to the 3D simulations of nuclear weapons physics as the “killer app” of the new computer system. Perhaps not the phrase I would have chosen. But it’s hard to deny the importance of making sure the nuclear stockpile is functional and not leaking or falling apart — I just wish the most powerful computer ever planned had a bit more noble of a purpose.