Video - The End of Embedded Processing and the Right Architecture for IoT

It’s easy to imagine that every connected device will have its intelligence in the cloud, but that’s a costly model for many connected devices that are expected to live for decades and cost little. Add in security and privacy fears and the model seems broken. Paul Brody from IBM has a new idea. Using block chain to manage some of this cost and complexity.


Caroline:  Hi, everyone, thank you for joining us today.  My name is Caroline McCrory and I am in Product Development for Gigaom Research.  And part of the focus that we are doing now at Gigaom Research is providing more actionable and practical content.  So that, when you read them you actually have an idea what you want to do next.  We published a sector road map yesterday around the smart home and the connected home and so now, I have Paul with me to actually talk about what does it actually mean and what are the real benefits of actually having a smart home.  So Paul, I understand that you are a very big smart connected home enthusiast personally as well?

Paul:  Yes, I am.  I'm torturing the rest of my family right now.  I've replaced every light switch and electrical socket in my house with the product from smart things.  And after linking everything up to add this and that and making my home really, really smart, I pretty much still most of the time get up and reach the light switch.

Caroline:  Tell us about your journey, to how you actually got to that point.  What was the first thing that you acquired that was a smart and connected home device and when?

Paul:  So I have a little bit of a kick-starter problem.  Actually, the first smart home device we had was that was a nest and we got that very early on installed that.  That was a really great experience.  At IBM, many of you know the center of the tech world is in New York and they think that 7 AM Eastern is a great time for conference calls.  So we would have these 7 AM Eastern calls and I would get up every morning and switch on the heat and the nest did a wonderful job.  It learned my habits after a week and then I went on a business trip.  I came home and I discovered that every programming and learning feature of the nest had been switched off by my husband who actually had gone to the extreme length of actually reading the manual, and he did that because he got tired to being woken up at 4 AM every morning with the heating going on.  And the problem fundamentally was our smart learning thermostat, just not smart enough yet.

Caroline:  Interesting.  So now we talk a lot about IOT being emended into our phones and into the sensors when you walk in through a store and all that sort of stuff.   Like us as consumers, how could a smart home system be better?  How could it be more autonomous where we wouldn't have to remember to flick that light switch or know that oh, he's a way, he is out of town, the settings will be for the people that are still at home?

Paul:  So I think that the problem right now is, first of all, these devices are not smart enough.  They need to be order of magnitude smarter and I love if this and that.  I'm the hugest fanboy of that service.  But I could never program it with enough sophistication to really make all of those things work.  I think the biggest issue with the Smart home is it's not a better home, right.  Lack of functional value is the biggest issue.  Inside of IBM, we talk about the smart connected toaster of the future and it's a running joke because nobody can figure out we'd actually want one.  Until I can have a smart connected toaster that makes better toast.  Yeah, I mean, the idea of, like, I don't want advertising burned into my toasts, right?  I don't think that Kaiser Permanente should know how many pop tarts I ate last week.  I want better product and right now, that's not happening.  That's the number one obstacle I think to the smart home or even the smarter industry is how are we making this better.  Just connecting it, that's not a feature.

Caroline:  So do you think it's a problem of a lack of an ecosystem or openness in the space where people will actually want to share and build something towards a greater goal rather than their own revenues which we all understand?

Paul:  No, actually, I think it's almost a little bit reverse.  We've been too much thinking about ecosystem.  Everyone's trying to fight the last battle, right.  Everyone wants to be the next Apple and to create the ecosystem and to tax the ecosystem.  Well, I'm not going to change my TV so that I can have it from the same manufacturer as my washing machine so I can get a text message when the laundry is done.  That's not -- right, there's no eco -- the Internet of things is too big to be taxed as an ecosystem and I think the opportunity is around very specific concrete ROI.  Nest is great because it's a great thermostat and I don't need fifty other items.  I have an apartment in San Francisco which I rent out on Airbnb to the obviously distressed of many activists in the city, but the one thing I've invested in which I never regretted at all was a remote connected lock and so I save myself a ton of time.  That's ROI.  I never regretted spending one penny on that equipment.

Caroline:  And it also is, you know, peace of mind for you that you could be anywhere and if someone checks out that you can be sure that it's at least safe or is ready for the next person to come in.

Paul:  That's right.

Caroline:  And you can reset the security settings and everything else.  So my husband and I use Sleep Cycle and instead of having to go and do a sleep study and things, we use this app called Sleep Cycle.  We leave it on our bed and it tracks the level of deep sleep we get, and this and the other.  And so we know why we're cranky the next day.  And not just think that we need coffee.  Now we are actually analyzing our actual patterns.  One of the things that we really like about it is that it also will connect with the Philips Hue.  Now, I'm very aware that there are so many companies out there now they're saying that they have these connected lights and we chose the Philips Hue because it's part of the app and they could then change the light in our bedroom to actually wake us up so that we're not cranky and don't necessarily need as much coffee.  What do you think of that space?  Do you think it's becoming so saturated because everybody is trying to jump on the bandwagon or do you think they are actually thinking about the actual uses and the value that is actually bringing to us as people with the times zone changes and all that sort of stuff?

Paul:  I'm a big believer in the Internet of Things will grow one very compelling use case at a time.  My father is a doctor.  He says there are no good days if you haven't had a good night of sleep.  There is a reason why sleep, exercise, cracking one use case at a time is going to be the most compelling way to grow this.  And then things will connect later on, but first and foremost, you have to -- I mean, Peter Thiel talks about it right create little monopolies, right.

Caroline:  Yes.

Paul:  Solve one little problem at a time is how we're seeing this market grow and when I talk to IBM clients who have successful thing related businesses, They are all remarkably -- they're not visionary, they're tactical.  They're useful.  They're compelling in that respect.

Caroline:  But don't you think with -- I agree that sort of approach because it's use case focused and it's relevant to people.  We all have problem statements and we all want certain things to do and solve that problem.  How do you deal with when it grows so in order of magnitude?  How do you -- what you think is the best approach to bring everything together?

Paul:  So this is the project that we've been working on right now which is, how do we scale from tens of millions of devices to billions to hundreds of billions of devices and for this we are actually working on using the technology from Bitcoin, the blockchain, to connect and manage the Internet of Things.

Caroline:  So I think a lot of people who have heard about blockchain seem to assume that it's only relevant to crypto currency.  Could you give us an example of how that type of methodology, technique, you know, technology, can be applied to, say, something like your smart car knowing that it needs to be charged when you're home?

Paul:  Yes, so from our perspective the blockchain, right, Bitcoin is a banking system, transaction processing system without a central bank.  The blockchain strip away the financial component and what you have is a distributed transaction processing engine that allows for contracts and transactions between parties.  Well, your smart watch, unlocking your front door that's a transaction.  The contract that enables those things those can all be set up and run.  So at CES in January, IBM and Samsung are actually going to show a proof of concept Internet of Things network running entirely on the blockchain with no centralized infrastructure doing very ordinary Internet of Things kinds of interactions.

Caroline:  Wow!  So from a processing perspective, let's say somebody here decides to try to use the same sort of technique that you're doing, how much of a backend heavy system is this?  Is that something a developer could test on their own or is it something where they have to buy heavy backend services to be able to build upon?

Paul:  No, this is -- so what's amazing about this is we are headed towards a world where -- and we have the blockchain.  We are running on a Raspberry Pi, I have on Arduino, we have running on Tizen.  So it's a big compared to some devices today but we are moving away -- I mean the biggest thing that's going on right now is we're moving away from a world where you have customized embedded chips.  Your refrigerator has like $5 worth of chips in it and that $5 today is not a very bright $5.  A couple of years from now that $5 of computing power on your refrigerator is basically going to be the original iPhone, dual core system on chip.  And for that you can run the blockchain, you're out of embedded and you into basically general purpose computing.  And when that happens, you can really change how you think about running this network and you can also change the economics of running a network.  We actually got interested in this at the beginning because several of our clients came to us and said they felt like the problem they were having is they were financially under pressure because their Internet of Things networks weren't generating as much revenue as they hoped they were.  But they still had users and they couldn't figure out how to manage the cost, right, and if you think about something like an LED light bulb, a Philips Hue or LED bulb it's going to be out there for 20 years.

Caroline:  And so there would be sending traffic, they'll be sending data and all of that and it will last a lot longer.  So what kind of embedded systems do you think would need to be -- like, how should embedded systems be redesigned to be able to cope with that because like you said these things are now lasting a lot longer as well.  How do you foresee not necessarily an ecosystem but anyone who wants to build an embedded system to deal with that?

Paul:  So we talk a lot about -- a couple of things number, number one design for longevity, right.  The Internet of Things, we're not going to replace our door locks, our light bulbs or industrial equipment every two years, right, much as the manufacturer would like us to.  The Internet of Things is going to be around for a long time.  What lasts for a long time?  So first of all, you have to keep it simple.  Secondly, open source.  Open source communities built on things like Bitcoin and other really well established tools, that foundation is going to last a lot longer than kind of any proprietary thing that requires maintenance.  So think carefully before saying well, we can just do it ourselves.  And then I think the third thing is we are busy getting all of this data, tons of data, all of this connectivity, why?  99% of all the data that I see our clients (inaudible 00:11:19) what are you doing with it?  They're, like, nothing, right.  I mean really who cares how much toast you eat and this is a really good example because all of our clients are, like, we should we should keep that data because we might need to sell your toast consumption data.  Or what if the government comes with a national security letter to know how much toast you have.

Caroline:  Subpoena that.

Paul:  Exactly, but the problem is actually the market for information is such that most of what we're storing is worthless and what people are willing to pay for it is zero.  But we're spending money to store it.  We're spending money to analyze it.  We're spending money to transmit it.  It doesn't make a lot of sense.

Caroline:  And I think it's like the whole big data thing that published earlier.  Everybody thinks that everybody is doing it, but I don't think anybody really understands the value of the data as a byproduct can actually do.  And I met another company Saffron downstairs who actually talks about applying cognitive, you know, the way brain functions and in cognitive sciences to applying context to big data and data analysis.  Do you think that IOT as it is today is set up to be able to allow that sort of contextual thinking into processing data?

Paul:  Not yet, but it absolutely needs to.  Because I don't think we will ever be able to develop enough if, then, else, kinds of statements to adequately automate our lives.  I thought I was very clever.  I had our house set up.  When I arrived home, unlocked the door and switched on the lights.  That is awesome the first time you do it.  The second time you do this kind of neat.  The third time you do it and you're home late and all of your family has gone to bed, you're very unpopular.  They are like why all the lights on?  Who unlocked the door?  That's, you know, it's a good way to make yourself popular at home.

Caroline:  So I think that's sort of a contextual analysis really would be, okay, it's 11 pm probably not and but that's missing, right.  Because a lot of the systems just they pick up your presence and then they activate.  They don't quite have the context of time and everything else.  So we have a few minutes left.  Are there any questions from the audience?  Please, walk up to the microphone if you don't mind or yell it and I will answer.

Audience:  You say that (Inaudible 00:13:42) vision for IOT is basically a peer to peer vision as (Inaudible 00:13:47)?

Paul:  So our vision is it's context to where.  But our thinking is that really if you were to make that two by two and I'm ex McKinsey guy.  I heard trash talking about that earlier.  We are not all bad, but if you were to make it a two by two I would say there is device value and then there is device longevity.  So if you think about low value devices that are going to be around for a long time, that's perfect for peer to peer.  You don't want 25 years of OpEx with your smart lightbulb.  On the other hand, if you think about something that is really valuable and it's going to be around, think about like a massive piece of industry equipment like mining equipment.  So you want centralized services to do analytics and intelligence, but you also want decentralized.  You want this devise to be smart enough to look after itself and interact with the other systems in its environment.  So our vision is it's a multiple items, but my thinking right now is that we can't scale to hundreds of billions of devices if everything gets sent back to the data center.  The cloud needs to go all the way out to the edge.

Caroline:  And there was another question from a gentleman.  There you go.  Sorry.

Audience:  Just curious.  What you said is interesting that everybody is doing Big Data or IOT and there's not a lot of use.  Why are people doing with that?  (Inaudible00:15:08) humans or companies (Inaudible00:15:12)?

Caroline:  The question is because we are saying that there are not that many people doing it and he saying why is everybody spending all this money doing Big data and IOT, if there is no real point because it's not human nature to spend to do something if there's no benefit from it.

Paul:  We are going to like a classic kind of a bubble phase right now where people are trying to figure out the value proposition.  I'm an optimist.  I think we will get there.  We will figure this out one really compelling value proposition at a time.  But right now people are sort of rushing off with their business plan, like, oh my gosh, I'm going to collect data on usage and then I'm going to sell it and then we're going to get a subscription fee and then people are going to buy apps for their toaster and it's going to be like a $28 billion toaster market by the year 2000.

Caroline:  So I actually agree with Paul from this perspective.  It's very much use case centric. The toaster thing, some people would like it but I don't think it's really that practical.

Audience:  Let's go back to the toaster example. There may be no use to ever know how much toast got eaten or how much toast are being consumed in a year, but are there any examples out there that you've seen that are something date-backed not necessarily about the primary use of the toast, but maybe about the circuity in the toaster or the (Inaudible 00:16:32) that’s used to put that toaster together.

Caroline:  So the question was not related to the consumption of the toast or the type of bread that you put into the toaster, but more how the toaster is made, the circuitry, the temperature, the duration of each toasting cycle, how does that data actually matter in terms of an IOT system?

Paul:  It does matter and what we're seeing very early on is predictive maintenance and management and repair of devices is part of the use case, but what's really interesting is that people get started down that path and suddenly they realize asset utilization getting more toasting cycles or getting more use out of the device, the value proposition of that the worse maintenance.  Maintenance in repairs great, but it's the tip of the iceberg.  I think IOT is going to drive a complete shift in how we think about the asset utilization in our economy.

Caroline:  Thank you very much.  Oh sorry.  One last question.  We'll squeeze one in.

Audience: My name is Andrew.  On the session topic of a architecture and location of processing, in thinking about long term architectures of distributed systems can you compare the difference between putting intelligence in the device which means firmware selection, microprocessor selection, a lot of long term survival of that distributed device versus putting intelligence in a centralized place, is there a major trend there that you recommend?

Paul:  I think we're headed to a world of sort of Linux or Unix on ARM on everything.  And so in some sense like (inaudible 00:18:06) are doing it.  These things are going to be -- there is going to be a standard substrate of everything and it's all going to work in a relatively similar way.

Caroline:  I'm actually seeing more of the distributed sort of compute power.  Like, we were saying iPhones are now -- there was more power in the iPhone than there was what was used to send men to the moon, and we use it to play Angry Birds and so you're absolutely right.  I think the distributed system is probably going to be where it is for the processing of it.  The data I still think will probably be stored in the central way.  So, thank you very much, everybody, for your time.  We hope you enjoyed the session and have a good rest of the day.

Paul:  Thank you, Caroline.

Written by Melvin Draupnir on November 1, 2014.