On the Future of Digital Storage and Synthetic DNA—Thomas Coughlin—Coughlin Associates

Richard Jacobs: Hello! This is Richard Jacobs with the future tech and Future Tech Health Podcast. I have Thomas Coughlin President of Coughlin Associates, all the widely respected digital storage analyst and a business and taking firms in over 35 years in the data storage industry various engineering and management positions. He has done a bunch of publications and the six patterns and credit. Those are the author of digital storage and consumer electronics, the central guide, and we’re gonna talk about something unusual DNA and Bingo storage so Tom Thanks for coming.

Thomas Coughlin: Thanks very much, Richard. Looking forward to talking to you.

Richard Jacobs: Yeah! So tell me what’s the premise of being able to store information using DNA?

Thomas Coughlin: Well DNA is, of course, our own genetic information in that of our whole life on earth is stored on, and people have been able to recover DNA from old species and basically potentially reconstruct and these learn a lot about those creatures.

There’s even talk about using old DNA, say from a woolly mammoth, the refrozen in the Tundra to eventually recreate such animals. So a DNA seems like it’s a means of the natural molecule that can store information for a fairly long period of time as long as it’s kept in the right condition. So I think that’s the kind of initial inspiration. The other part of it is the tools for being able to right creates a DNA, right it as well as to read the information on the DNA, read the sequences of DNA have improved enormously. A sense of first a genomic, a full genomic analysis was done in the early part of the century. And so the prices and the time it takes to write and read information and the case that you’re doing digital storage, it’s going to be a synthetic DNA, not natural DNA.

Those costs have gone down to the point where people are starting to consider that as an option for it, keeping Information and especially for keeping digital information.

Richard Jacobs: Can you give me a comparison between what it cost, the first sequence, the human genome and storage capacity and the cost and the speed now?

Thomas Coughlin: Yes! The first ones were enormous, huge government projects and also private projects costing many millions of dollars in order to get the first genomic sequencing. And now I’ve, I believe the Cost of a genomic sequence is around a thousand dollars, and there are projections and that would get in the range of $100. It’s up to a too long period of time which also, of course, will have great benefits for personalized medicine that’s based upon the person’s actual, genomic makeup. And so the thing that’s really driving a lot of the development of that technology in fact is the medical profession trying to create personalized medicines based upon utilizing your own genomic information to tailor drugs and processes that may benefit you individually, maybe have fewer side effects as well.

Richard Jacobs: Yeah! So if you were to compare storage on current, just don’t mean yet. How long does it last before it gets corrupted and what’s the density of the storage capacity versus a DNA? A long thing, the storage density.

Thomas Coughlin: A DNA potentially can store very high densities of information. There’s a lot of stuff, a lot of things need to be done in order to enable the commercial DNA storage and we’re still in the early days and such things you know, the potential capacity is maybe a thousand times that what’s currently achievable with modern storage technologies, probably hard disk or maybe the example particular for that.

Richard Jacobs: So it has a lot of a thousand times in dense?

Thomas Coughlin: Approximately a thousand times dense. Yeah, somewhere in that range I think.

Richard Jacobs: How would that be possible? What allows it to do that now? What is it?

Thomas Coughlin: Well, we’re talking about volume, metric density. So if I have strands of DNA, I could pack them into space and have a much synthetic DNA and pinch. They have a much higher bid density per volume per cubic centimeter than I could with say, hard drives or with flat flash memory for that matter, but magnetic tape, part of this drives for instance so that’s the potential advantage is that I could have a high volume metric density some proven under the right conditions, proven longevity of information. Now it’s not to say that there won’t be probably unique, a means by which bits get corrupted and that you won’t need to put in some kind of parody and some error correction in recording the information that’s highly likely that you’d have to do that and it’s going to have its own. I’m unique that rock characteristics and things of that sort. So interesting thing about DNA, one of the really interesting things about DNA storage is when I write a DNA sequence, I can write like a million copies. It’s about as fast as I can write a single copy. So if you want a lot of copies of a piece of information, this would be a way of making that everyone could have their own pieces of this information on a piece of DNA if they wanted.

Richard Jacobs: Well why is that possible? What is it about that allows you to make so many copies so quickly? No more time than one copy?

Thomas Coughlin: Well, in fact, that’s how they purify a DNA as they get the strands of DNA as they want and then they create multiple copies of it. So a lot of the technologies they developed is to create large amounts of a DNA sequence in order to what they magnify that for other words that they want to do on it. So it’s just one of the characteristics of how people work with DNA that they can do that. So essentially it has those capabilities. I can make an awful lot of copies of a piece of data on DNA. I’m now writing it, it’s not necessarily fast, but I can make a lot of copies. And likewise, reading is not very fast. And part of that whole getting the costs down both for potentially for digital storage applications and also for these medical applications is increasing the throughput of DNA, both writing at the creating DNA, writing it and also reading the DNA to decode the sequences and those speeds.

People are putting a lot of technological development work in RND into improving those speeds. A lot of being driven by this med, by the medical applications, potential medical applications and uh, but it also allows for the possibility of lowering the cost of writing and reading data, synthetic DNA and therefore being viable storage mean. Like I said we’re still in the early days but in 10 years’ time there are some possibilities that particularly for a passive archiving type application that DNA could be an option. Synthetic DNA writing and reading information could be an option for long-term passes story that, information that you don’t access all the time that you could keep that on synthetic DNA sequence.

Richard Jacobs: What happens in a typical, hard drive? How’s information stored and how does it become corrupted?

Thomas Coughlin: So we all face the laws of thermodynamics and tends to increase, so there’s certainly activated so weighted the hard disk drives work and there are similarities in the way that almost any kind of digital storage technology is going to work in case of hard disk drives for using magnetic fields who read and write information. So you have something called a HIB which generates a magnetic field and also we’ll be able to read back magnetic information. Often the magnetic media and the heart, the magnetic media rotates underneath these magnetic heads to treat right information, concentric tracks on a disk surface. And that’s how you read and write the information. Then there’s also circuitry for doing signal processing, extracting the data. There are special coding is done in writing the data that gives you some amount of repetition of data.

That if there is some either noise during the read or write a process or if there is some minor corruption of the data that you can still recover. They’re also our processes for going through data when it’s not being used and making sure it’s okay and rewriting it if it’s having trouble. So, what happens is and this is true of any means of storing information is that a thermally activated process. In the case of the magnetics, it’s the random reversal of magnetism of the very fine grains of material. In the case of a hard disc drive on the disc to meet this media surface over a period of time, there are some random reversals that happen to a period of time. If enough of that happens it would tend to cause problems in some areas in the media and therefore your signal will degrade and eventually if that continues, you won’t be able to read information or you, you’ll lose pieces of information. So when a key element, any kind of a memory storage technology is to build in redundancy, to build an error checks and occasionally to go through and make sure that the data is okay and to rewrite it if it needs to be rewritten in order to recover the data or make it so that it will last longer and often requires some kind of active management to do that.

Now hard disk drives is true of magnetic tape. It actually is true some excreta based technologies like flash memory that over time data has a tendency to degrade and so you need to actively Rewrite it and moving on that sort of the natural phenomena.

The other thing that happens versus what it has called Technology apps, the technologies that we use to read and to record information have been changing over time. For instance, how many people nowadays have the tools to be able to read a tape, a cassette tape for example back in the eighties, nineties, It will be a fair amount of technology to do that. A lot of people don’t have that anymore, the small cassette tapes and so as technology changes over time and that can be the recording technology itself, but also could be even more important, the interfaces and even the software drivers that enable the devices that can read write then information on those technologies may change. So another factor in terms of trying to keep information for the long term is in addition to natural degradation of information that can if you don’t intervene and try to go up to catch it before it happens.

The other thing is the technology that you’re using can become obsolete and that can even be the software itself. So actually carrying information into the long future, in particular, has a lot of requirements behind it. And so there’re people that are working on interesting technologies where, for instance, they will essentially archive a package that contains, for instance, the information about the original software that created that one could even make, creates an emulation of that software in order to be able to recover the data, the sort of metadata constraints but the other thing of course it’s physical interfaces, things like that. For instance, a hundred years from now, will I have, will there be such a thing as the USB interface, would my old USB be able to plug into that work? So these are in terms of long term storage of information, whatever the technology there are definitely things that need to be considered and trying to bring data in the future, and there’s definitely challenges in doing so.

Richard Jacobs: What are some of the current roadblocks with storing information on DNA and you know, where would the DNA be housed and how can it get information when needed?

Thomas Coughlin: Sure! So there’s a number of companies that are a number of startups that I’ve run into. I had done some work with some of them that are looking at using DNA for storage information. Some of them basically gonna create DNA strands, you dry them out essentially, and then when they’re dry they can remain stable for a fairly long period of time under controlled conditions and then you reconstitute them and put them in a reader and you can read back the information. Not a very fast process, but you know that’s one way in which people are looking at doing it.

They’re also, and I can’t go into too much detail on it, but folks that are working on actually using some attempt your technologies and there’s a lot of work that is done, for instance, called IMac which is in Belgium which is a nanotechnology and electronics research center moving in Belgium, other people have been doing a lot of work on building, for instance, microfluidic and electronic devices on silicon. So sort of nanotechnology Nam type devices microfluidic devices, building sensors believe electronics and actually be able to, for instance, move fluids around inside of a silicon matrix. And there are people that are looking at building, using that kind of technology in building read and write capabilities essentially on a chip basis and saying if they can achieve very high densities of information storage with some high certainly a higher data rates, both in reading and writing you could get from that, a more passive approach where I just dry it and then I have to reconstitute it.

And read it, read the DNA again. So the DNA essentially stays inside of some cells on a chip from which they can be a written to or read from overtime. So there are some fascinating ideas that people are working on, which again hadn’t learned the early days. But they offer the promise that we may add DNA to some of the other digital storage technologies that we are used today. In fact, in many regards, I think if you consider all the new solid-state memory technologies as well as a flash memory, which has been around for a while hard disc drives tape, we can add DNA to the mix. And essentially we’re just getting the whole a storage memory hierarchy is becoming even more complex and more specialized in terms of how we may use these technologies together or singly for accomplishing different though different tasks.

Richard Jacobs: So what’s going to be your specific roles? Are you investing in startups or what are you specifically doing?

Thomas Coughlin: So what I’ve done in this area right now is that, so first of all, I’m a story journalist. I’m always interested in what people are doing with any kind of memory or stories, technology. So I love this stuff. This is my bread and butter. And I write on technologies, for instance, I just finished a report recently with a colleague on that emerging memories for artificial intelligence applications, but I also do consult and I worked at startups and other people in terms of technologies and how it might be implemented. So I’ve done a bit of that with some of a DNA storage folks, and I love doing this kind of stuff because it’s so fascinating. First of all, digital storage has allowed essentially it’s the enabler of modern civilization though, processing, storage and networking and everything we got nowadays digital and that all has to be kept somewhere.

And so all applications, since they are dependent upon various types of digital storage and memory technologies. So that’s kind of stuff I deal with. That’s what my interests are. I also put on some events in this area. In fact, I’ve got a workshop with a professor at Stanford. I’m doing on August 29th on emerging memories in artificial intelligence. So I’m sorry, the report that I did recently with a colleague was on emerging memories of their development that cap the market and also capital equipment in order to be able to manufacture manufacturing. But I’m also doing this workshop on August 29th at Stanford with a professor there at the Bechtel center, which is going to be looking at various types of emerging memories and artificial intelligence, that includes, there’s some interesting AI app AI Interactions with these emerging memories like examples or phase change memory or resisted memory where they’re actually using them as emulating. To some extent the way that nerves in the brain work and so they call it neuromorphic computing, just sort of spike-based process and analog computing processes.

They’re using these memories cells in order to feel enabling. It emulates the brain. So there are really fascinating things people are doing, applying technologies like that as well as other technologies for deep learning, machine learning, convolutional networks all of these related applications.

Richard Jacobs: How are the set of stored data indifferent if an AI was emulated more how people stored data in their brains, like would that data set change over time and prescribed ways or certain elements diminished?

Thomas Coughlin: Yes. So, so what they’re doing is they’re actually building memory technologies. The particular examples that are being used are what’s called resisted memory and phase-change memory. And they’re used, but they’re using them as sort of analog processing units, and so they process information in a way that’s related to the way the brain does it. And so that’s essentially what they do. Now there are other things with these emerging members. People are doing that’s called neuromorphic computing and maybe very useful, especially for training applications and artificial intelligence. In the case of actually using that stuff in the field, there’s a process called inference, which is done where they take a training model, it’s already been created. Now I’m trying to apply that to say images captured in the field or what voices I can do. The image recognition, voice recognition thing, that sort of, and the technology they used for that. It’s called inference. So they have these inference engines which need weighting functions that represent, for instance, characteristics or sound or characteristics of an image that it can then process using the algorithm developed during the learning process in order to say in this camera and seeing a dog and seeing a car, things like that. Say if an autonomous vehicle recognizing the voice of somebody or something of that sort. So a lot of IoT type applications, which include artificial intelligence. What they’re doing is they’re either in the endpoint device or in the edge close to it. They’re going to have inference engines. And these inference engines have to have wedding functions that are stored in memory, and what they’re looking at, battery-powered device that is non-volatile emerging solid-state memory is like magnetic random access memory may be an option to include in these, devices because I can save power because they don’t have to refresh the memory all the time as I do and it’s all-time memory.

But the other thing is I can turn them off and not using it. So I have a lot more ways of reducing my power usage and therefore increasing battery life.

Richard Jacobs: What about on the initial stories? I mean there are data streams that are so large that it just wouldn’t be practical those to where all the information coming in. So what about a discriminatory function that’s on the front end of the storage unit that it’s using AI, that story that builds from that and deprecates this is more important, kind of work with the data, and then storage change they can take it.

Thomas Coughlin: Yeah! I think maybe the prime example that would be, let’s say I have an autonomous car that’s in a connected network environment in the city.

So I have something called a smart city and I have a lot of the states 5g network or something like that. I’ve got intelligence, actual processing power at the edge. They have cell towers or even in even more distributed to that. And the car itself takes in a lot of Betas they’re talking about these things generating maybe a terabyte of data a day, but most of that data is meant for immediate processing and analysis because the car, if it’s going to be able to help you to drive or drive itself, it’s got to make split-second decisions. I can’t wait for the latency of going over the network. So it does a lot of stuff locally and that might be with an inference engine, but then there are results of that analysis for the commands that the car gives that it wants to share with the outside network.

So, for instance, that car and all the other cars in the area, I may have at work that keeps track of what cars are doing or Warne’s other car. There’s a car that you can’t see right now. It’s around the corner that he’s going to be driving through the intersection, right? So it can warn another car and help to keep accidents down and that sort of thing. So it’s gonna require both intelligence capabilities and some learning capabilities of course at the edge or in the device themselves as well as in the data centers. So the big data centers, the cloud. So what we’re seeing actually, it’s kind of interesting, there’s like this eternal back and forth between centralized processing and distributed processing. And I think with the growth in AI and requirements that we’re moving back into a distributed computing environment where there’s going to be essentially processing all over the place, and that’s what enables a lot of these Iot and autonomy. The autonomous driving or any kind of a, or even in factories enabling, factory 4.0 or I’ve got robots and make of decisions themselves. They’re working directly with human beings rather than restricted areas. They’ve gotta be friendly to humans and that this new distributed computing environment that’s using these artificial intelligence algorithms and various sorts of doing both learnings and in training and also application of those models, the inference engines, that kind of thing all of these things are going to be very distributed, and if we’re going to generate more data than ever before and some of that data is going to have to be kept and that’s where the storage comes in and the memory comes in for the short term stuff.

Richard Jacobs: Any other storage seems that it really was surprising to you when you have a lot of open or you think they’re going to be like really instrumental in the future?

Thomas Coughlin: Oh, there’s, there’s a lot if anything we’re getting, it’s a storage, and options are more complex. I mean they’re greater than ever before, and it’s an interesting thing. I think it has to do with the fact that it is as human beings, we’re natural stores and information and our brains work by getting layers and layers of information and that’s what’s created culture is the handing down of information. So we’re naturally information-processing creatures by our nature and the options and means by which we do this our CB is increasing over time, so I keep track of things are going on, always looking for notices of this or that.

And there’s always seems to be about every two or three weeks some new storage technology that often times will announce. It says it’s going take over everything and you know, it’s done in the glass with holograms or it’s done with some kind of weird semiconductor technology or other. So there’s enormous creativity that goes into trying to create new ways in which information processing creatures. New ways in which we can understand it, can capture and understand the world around us and ourselves. And so I think it’s their enormous creativity, there enormous demand and I don’t see it ever diminishing. I think it’s just the nature of who we are.

Richard Jacobs: Trying to date a mind, literally someone’s brain there are neurons in your network. I don’t know what you’d get from it, but anyone that’s trying to hit some link that to emulate it and, or download it somehow.

Thomas Coughlin: So I’m actually a volunteer with the I Tripoli. I’m currently the president of I Tripoli USA, but I’ve been involved in an activity and I triply tent collectively activities that are called future directions, and there is an initiative in I triple future directions. It’s actually, it’s called the brain initiative that has been working and there’s a number of different groups and organizations working on that. The actual brain to electronics interface and I would say that we’re not too far off from the practical app. In fact, there are some practical applications and that kind of technology, particularly to help people with disabilities. Let’s say to allow somebody who’s communicating with an artificial hand using their muscles in their brain, their neurons but even more complex things are in the works.

And I think it is gonna drive some interesting consequences in terms of personal privacy and the demand for such and legal requirements. But, we’re probably within decades of being able to essentially read a person’s brain which may allow people, for instance, to create an artificial telepathy where one person can read the mind of another person or can one person can send his thoughts to another person or device through electronics without speaking, but there may be even further things and there’s all this discussion of deep faith technology right now where I can create things images and videos and stuff where I can’t tell what’s real or not by projecting directly into the brain, right?

So there’s going to be some interesting legal and ethical implications of these technologies in the future and how do we use these things and what’s the right way to use what are human rights? And in the age of such technologies, how do you protect freedom? And how do you know what’s true? And probably one of the things that may eventually happen is something along the lines of people has a right to know the truth if you will ground the truth that what is real, actually real, physically real versus what is represented them. So interesting implications of all these things. But there’s a lot of people working on technologies and which of course for good or for ill, you know, in terms of applications but hopefully that we’ll be using these in order to enhance life and enhance the way we work with each other and get along with each other and make progress and do new things. I have adventures and fun.

Richard Jacobs: Any milestones that you think are realistic that are coming the next one to two years, that convention or personal problem that you’re working on that you think will come to fruition, at least an early version the next year.

Thomas Coughlin: So one thing essentially I just actually asked the semi conference and it was an AI event put on by applied materials, applied materials, make semiconductor manufacturing equipment particularly used for a lot of things. But memory is one of the big drivers that semiconductor spending and one of the things they were seeing in 2018 was the first year that more data was generated by machines units.

And that trend is only going to continue. So I think our machines in our devices, all the IoT devices, etc. are users of data, as I said before, kind of by our nature. But our machines now are generating more data and will do so increasingly in the future than we have. So I think human beings have specialized oftentimes in creating information and I’m sort of looking at the future information. It’s so hard nowadays to, for instance, read all the literature that exists but I could see the data. We create machines that can gather all the information generated in a field or unrelated fields, help us to with human intervention and direction, synthesize that information in order to create new types of knowledge and new ways in which to create wisdom and understanding of things that probably will exceed anything we’ve ever had before.

Richard Jacobs: So imagine it would have at hand all the knowledge and the given fields and you see it dynamically updated in real-time and then you can make a difference and anything that you think of, you can check it against that home corpus of knowledge would be amazing.

Thomas Coughlin: It would be amazing and collaborators, both machine and human, you know, and others were still going to be a need for a technical community for people, livelihoods and their passion is to know things, but we will have new tools for which they can do it. And perhaps even who knows maybe some of these humans will not be organic form. We’ll have to wait and see what happens there. But I just see fascinating possibilities in the future.

And you know, one of the things that are really cool this year as well the last few years, for instance, I’ve got a bachelor’s degree, my original degree was in physics, but the black hole event, horizon imaging work. And the other thing was the gravity wave detection that has been done within the last three, four years. Fascinating ways to see new things about the universe around us and that’s been enabled by a technology to developments of technology, you know, in terms of process and power and turn to storage and memory and networking and our ability to build stuff that’s evermore a complex and interesting, so we know less than we will never know than we will ever know. It’s probably true whenever think a few things, stuff, but we don’t know everything and there’s a lot more that’s out there that will be fascinating and we’ll keep humans and our tools working to Understand you know, for I think centuries to come, it will come.

Richard Jacobs: What’s the best way for people to get in touch and to find out more?

Thomas Coughlin: So they can find out about me at tomcoughlin.com. I’ve got information there on who I am and reports I do if you look up artificial intelligence, you can find out about our August 29th a workshop at Stanford and my email is tom@tomcoughlin.com. If you’re interested in this and want to talk more on things, I’m a frequent speaker. I write a lot of stuff. I do a regular blog on forbes.com. I did that book on digital storage, consumer electronics and I also do a lot of stuff and I am involved in Tripoli you probably run into my name as well. We’re good.

Richard Jacobs: Tom, thanks for coming on the podcast. I appreciate it.

Thomas Coughlin: Thank you very much. I appreciate your inviting me to come.