I have not yet commented on IBM’s announcement that they are partnering with Samsung to create a new infrastructure for the Internet of Things (IoT), called Adept (said to be demoed at CES in January). One of the pieces of this as-yet-to-be-seen solution is a blockchain and possibly cryptocurrency as well.
Since that time, they have released a white paper which explains their analysis of the problem and their approach.  This note examines this WP in some detail.
Brody and Pureswaran on “Device Democracy”
They are thinking about the problems presented when billions of devices, “from doorknobs to light bulbs will be able to carry as much compute power and connectivity as the first smartphones.” (May heaven preserve us, they are probably correct.) This is the Internet of Things, IoT.
To date, B & V say, IoT has had limited success, especially in consumer markets. Worse, expectations and investments have run far ahead of payback. Clearly, things must change if such a scale up is to succeed.
They diagnose five problem areas:
- Cost of connection and maintenance
- Trust is over, privacy must be built in
- Long product life (decades)
- “A lack of functional value“ (ouch!)
- “Broken business models” (double ouch!)
To approach this problem they have studied the problem with a very much end-to-end perspective: simultaneously examining technology, business models, and user experience.
Of course, IBM is still a “hammer maker”, so they will always find “nails” that need to be pounded. In this case, they work from the idea that “[t]he foundation of modern computing is […] transaction processing.” p. 6 Well, maybe, one of the foundations. But IBM surely, surely knows transaction processing, so let’s see what they make out of this.
From their analysis they lay out some basic propositions.
- IOT has to be peer-to-peer
There is no way in the world that your doorknob is going to talk to the cloud. Or rather, they say “It’s time for the cloud to move from the data center to your doorknob”. P. 7
- It has to be “trustless”
This is in the sense that the infrastructure is not dependent on trusted computer systems. (The point actually is to make the actual devices more trustworthy, because they reliably work the way the users want. The use of the word “trust” is very confusing in this arena.)
- Privacy must be built in
They summarize the key features for a decentralized IoT solution. It
“should support three foundational types of transactions:
- Trustless peer-to-peer messaging
- Secure distributed data sharing
- A robust and scalable form of device coordination.” p. 7
All this must scale to huge numbers of devices, all without system admins or central services.
The “breakthrough” idea for this is to use the blockchain distributed consensus mechanism to support these decentralized transactions. The blockchain provides a shared, permanent, “write once” medium for all the entities to exchange messages. These include all the functions familiar from the last twenty years of “smart device” developments: registration, publishing capabilities, discovery, negotiating “contracts”, and so on.
Aside: the blockchain classically depends on a completely flat network, i.e., every record is seen by and confirmed by every participating node. Correctness is assured by the pool of nodes being so large that a dishonest subset cannot “out compute” the whole world. (Note that you are “trusting” the whole Internet instead of any given subset of the Internet—an interesting use of the word “trustless”.) In the case of the IoT, it is less clear what the appropriate pool of node is.
Do all the doorknobs of the world check all the bazillions of transactions for everyone’s doorknob? How would you make that work? And if not, what is the pool that maintains consensus? They mention something called “a community of self-maintaining devices” (p. 8). Let’s just say, “I’ll be interested to see one of these”.
Assuming this technology can be built (and IBM isn’t likely to be blowing smoke here), they foresee important business models, centering on what they call “liquification of the physical world”. This involves lightweight digital payments, algorithmic contracts, and so on. Analogous to what the Internet did for digital content, physical objects and services will be open to search, use, and payment. (And, like the Internet, it will mean renting everything.)
This also leads to flexible supply chains and peer-to-peer marketing, which underpins the “sharing” economy. They also believe this will lead to new and better credit and risk profiles. (Here is another interesting use of the word “trust”: the “trustless” network is used to establish “trustworthy” credit ratings.) It isn’t completely clear how this credit profiling matches up with the supposed privacy of the technology. I suspect not.
Finally, these technologies could, in principle, improve old style industries such as agriculture and transportation by inserting corporate computing power into every little step. (These are not exactly “sharing” economies, and it remains to be seen how the interests of all stakeholders are met.)
The third perspective is an analysis of user experience. Here, the paper has less to say except that design is critical. I note that, unlike the other sections, there are no case studies or example. The reason is simple: no one knows how to do this well, and most current systems suck. (Bob’s rule for IOT: “your smart phone is not the interface to your world”.) They note one of the most difficult and interesting aspects of this arena: the majority of the action will be invisible, machine-to-machine transactions. Figuring out a “grammar” for smart devices, their composition into assemblages, and their (hidden) conversations is a grand challenge for the early 21st century, in my opinion.
IBM and Samsung have announced that they will demo ‘Adept’ at CES in a few weeks.  This demo is supposed to be a technical platform to satisfy these needs, including peer-to-peer messaging, shared storage, and blockchain-based transactions. It will be interesting to see what this implementation does.
Given the broad end-to-end treatment in this white paper, there may be more than just some software. It would be quite cool to see example user experiences, as well!
It will also be interesting to see exactly how they use blockchains. How many chains are they? What incentives / ‘coins’ do they use? How do they do transactions using the blockchain? What kind of latency is there in confirmations?
Despite the corporate tone, this is a very thoughtful and useful paper. As someone who has worked in this area (e.g., McGrath (2003)  and refs), I will say that the paper is extremely well informed, even though it does not cite the extensive literature underlying it. (I understand that this is a business / sales document (a “though leader”), not a dissertation.) For that reason, I expect great things from these guys, and I am looking forward to examining what they produce.
- Brody, Paul and Veena Pureswaran, Device Democracy: Saving the Future of the Internet of Things. IBM Institute for business Value, GBE03620USEN, 2014. http://public.dhe.ibm.com/common/ssi/ecm/en/gbe03620usen/GBE03620USEN.PDF
- Higgins, Stan, IBM Executive Highlights Block Chain’s Utility for Internet of Things, in CoinDesk. 2014. http://www.coindesk.com/ibm-executive-block-chain-internet-of-things/
- McGrath, Robert E., Semantic Infrastructure for a Ubiquitous Computing Environment, PhD Dissertation in Computer Science. 2005, University of Illinois, Urbana-Champaign: Urbana. http://hdl.handle.net/2142/11057