Re: [Iotsi] interactive vs. programmatic IoT

[Michel Kohanim <[e-mail]>]

Hi Ravi,

Thank you for the details alas I cannot discern anything practical in what you are suggesting (i.e. how would I implement it?).


1.     The boundary may not be a distinct and stark transition – it may be more diffuse. We are already seeing evidence of serious and useful technology where this boundary has been blurred.
[MK] If you don’t mind, please provide links to some of this useful technology.

2.    I am not sure that the machine “must know everything a priori” – much of what is needed can be late-binding and discovered.

[MK] How much does “much” refer to? Late binding and discovered does not mean that the machine “understands” them. In short, either the machine has to know about what’s being discovered OR have some mechanism to discern the high level natures thereof.  I propose the latter. Although I would love nothing more than having philosophical and theoretical discussions about IoT, however – and from a manufacturer of automation devices perspective – words such as “blurred” and “diffused” do not easily translate to requirements (use cases), design (classes, interactions, sequences), and ultimately code.

 But it does make it necessary for a lot of information and data to be explicit and not implicitly encoded in some code/SW implementation.
       [MK] Explicit as in coded in the machine? If not, then where? Who are the consumers of this information and data?

 This has always been true in adaptive systems.

[MK] Would you please give me a link to one of these adaptive systems?

3.    Yes, learning is useful but there are very sophisticated behaviors in the aggregate that are possible with simple rules of behavior and dynamic application based on environment – we have analogs in the insect world (bees, ants and other social creatures) and fractals. So if IOT were to be viewed as many such simple elements manifesting complex and sophisticated behavior in the aggregate then we need some well-defined patterns of behavior and explicit definitions and we can see some very useful and complex results.

[MK] I am no longer certain what we are trying to solve here!

4.    Sure we would also like to add machine learning to the mix but that would be another element higher in the “logical” system hierarchy very similar to the architecture of the human brain. The systems that are designed for 2 and 3 will make it easier for the learning to be “integrated” into or be “part” of the system. There are many techniques for learnings and from my knowledge of this area, I cannot discern one paradigm that is universal though there are many that make a good starting point.

[MK] I agree and thus my point: we are not trying to solve AI and learning. We are trying to solve how things can understand each other at the highest level (late binding/discovery) . My question is: are there any boundaries with regards to decision making? i.e. if X::A == n then Y::B() && Z::C(w , y1)


With kind regards,

******************************
  Michel Kohanim
  CEO

  (p) 818.631.0333
  (f)  818.436.0702
  http://www.universal-devices.com<http://www.universal-devices.com/>
******************************

From: Subramaniam, Ravi [mailto:[e-mail]]
Sent: Thursday, March 24, 2016 9:38 AM
To: Michel Kohanim <[e-mail]>; Kreuzer, Kai <[e-mail]>; Andy Bierman <[e-mail]>; [e-mail]
Subject: RE: [Iotsi] interactive vs. programmatic IoT

Hi Michel,

I do agree that there may be a boundary between user and machine but I find it hard to follow the conclusions that you have derived …


1.     The boundary may not be a distinct and stark transition – it may be more diffuse. We are already seeing evidence of serious and useful technology where this boundary has been blurred.

2.    I am not sure that the machine “must know everything a priori” – much of what is needed can be late-binding and discovered. But it does make it necessary for a lot of information and data to be explicit and not implicitly encoded in some code/SW implementation. This has always been true in adaptive systems.

3.    Yes, learning is useful but there are very sophisticated behaviors in the aggregate that are possible with simple rules of behavior and dynamic application based on environment – we have analogs in the insect world (bees, ants and other social creatures) and fractals. So if IOT were to be viewed as many such simple elements manifesting complex and sophisticated behavior in the aggregate then we need some well-defined patterns of behavior and explicit definitions and we can see some very useful and complex results.

4.    Sure we would also like to add machine learning to the mix but that would be another element higher in the “logical” system hierarchy very similar to the architecture of the human brain. The systems that are designed for 2 and 3 will make it easier for the learning to be “integrated” into or be “part” of the system. There are many techniques for learnings and from my knowledge of this area, I cannot discern one paradigm that is universal though there are many that make a good starting point.

Thanks!

Ravi

From: Michel Kohanim [mailto:[e-mail]]
Sent: Thursday, March 24, 2016 9:01 AM
To: Kreuzer, Kai <[e-mail]<mailto:[e-mail]>>; Subramaniam, Ravi <[e-mail]<mailto:[e-mail]>>; Andy Bierman <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Subject: RE: [Iotsi] interactive vs. programmatic IoT

Ummm …. Where does one draw the boundary between what must be relegated to the user vs. the machine?  If we expect the machine to make all the decisions, then:

1.       The machine must know everything a priori … this means, for every change in the semantic world, there needs to be a change in the machine world

2.       Or, the machine has to learn … this means that we would have to use some AI techniques. Are we up for it? If so, which AI paradigm are we going to use?


With kind regards,

******************************
  Michel Kohanim
  CEO

  (p) 818.631.0333
  (f)  818.436.0702
  http://www.universal-devices.com<http://www.universal-devices.com/>
******************************

From: Iotsi [mailto:[e-mail]] On Behalf Of Kreuzer, Kai
Sent: Thursday, March 24, 2016 8:37 AM
To: Subramaniam, Ravi <[e-mail]<mailto:[e-mail]>>; Andy Bierman <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Subject: Re: [Iotsi] interactive vs. programmatic IoT

Hi,

I fully second Ravi on this – the ultimate goal is that algorithms can make “sense” out of the services that they find. Being able to only operate through UIs pushes the semantic knowledge” onto the user, so I would not even talk about semantic interoperability, but rather “only” about technical interoperability (see also the second half of my blog post here<http://kaikreuzer.blogspot.de/2016/03/semantic-interoperability-in-internet.html>).

> HATEOAS doesn't really help here

I do not agree on this. The idea of having the links on the resources is to make them self-descriptive and navigatable. So algorithms should be able to dynamically “discover” what the service is about and use it adequately without any prior knowledge. But Matthias is probably the best person to comment on this.

Best regards,
Kai

Von: Iotsi [mailto:[e-mail]] Im Auftrag von Subramaniam, Ravi
Gesendet: Dienstag, 22. März 2016 18:49
An: Andy Bierman; [e-mail]<mailto:[e-mail]>
Betreff: Re: [Iotsi] interactive vs. programmatic IoT

Hi Andy,

IMHO, for IOT to be successful it would be primarily (what you are calling) “programmatic clients” – most *useful* IOT systems would tend to be relatively or fully autonomous with human interactions at the “periphery” of such systems.

OCF has called out that it is “declarative and late binding” because the expectation is that a “human” would declare “what” they wanted the system to be and its objectives and the rest is done by autonomous interactions of “programmatic participants” (well… at this stage, OCF spec is in the “crawl” stage so this ‘vision’ is hard to see but the basic concepts support this direction ☺ )

Ravi

From: Iotsi [mailto:[e-mail]] On Behalf Of Andy Bierman
Sent: Tuesday, March 22, 2016 10:29 AM
To: [e-mail]<mailto:[e-mail]>
Subject: [Iotsi] interactive vs. programmatic IoT

Hi,

It seems to me that the use-cases discussed at the meeting
assume there is a human with access to UI driving the client.

Does this mean use-cases which do not assume any human interrupts
are possible are not IoT, but something else? In this environment
everything is usually programmed in advance. HATEOAS doesn't
really help here.  It is not likely the client can make decisions
about code-points it has never seen before.

Are there any expectations that IoT includes programmatic clients
or is it just for interactive clients?


Andy