This is also interesting and related. http://www.forbes.com/sites/theopriestley/2015/12/08/a-series-of-unfortunate-tech-predictions-artificial-intelligence-and-iot-are-inseparable/#5905c19523ad
It suggests the battleground is not over APIs and who controls them, but that IoT devices and processes in the cloud and on the edges will not, and can not, wait for standardization or deterministic control and stablization over APIs. APIs in this world are only a convenience for devices that will try to acquire "environmental data" in any fashion possible using APIs, or not, and will try to make sense of it and use it in appropriate ways to produce a successful outcome. It's still another set of criteria for us to be looking at and knowing up front for each of the connected device projects that we set in motion, each of the devices may have one or more of these characteristics: -> Is this intended to be a "controlling device" where its acting on its own or maybe on behalf a one or more users. -> Is it a "controlled device" mostly receiving instruction from some other source(s), -> or will it try to just acquire as much data from it's environmental surroundings as possible, and then figure out what to do with it to produce successful outcomes using AI often acting on its own. Advanced algorithms that produce simulation data, or even generates code or instructions on the fly could be the the required future for almost all connected devices. The way they acquire the data is just part of the details. that last seems scary, but think about the places where this happens already. Also think about this as a more logical path for those people that will want and need these devices. Many will not want to, or will simply get tired of, programming and/or instructing every single IoT device in their life and every kind of setting in every kind of situation. Also the reason that last one is scary is, because it is. The biggest security bugs and catastrophic technology failure often come from situations where the sending or data generating side of the API, and the receiving or consuming side of the API, are not quite in sync about what the data is or the conditions on which it is sent. -chofmann On Thursday, January 21, 2016 at 10:21:33 PM UTC-8, David Ascher wrote: > Agreed, OpenAI is interesting, and worth keeping an eye on. > > > More immediately, one of the most useful techniques in "weak AI" is so-called > "deep neural networks" (cf. Hinton et al for the last 15 years), which can > now be deployed quite effectively (see e..g "Smart Replies" in mobile > gmail/inbox apps). Google's recent code release in this area is quite > interesting: https://www.tensorflow.org/ > > > The challenge with a bunch of these approaches from a Mozilla perspective is > that like a lot of machine learning approaches, they tend to work better with > more data, which provides structural advantages to highly centralized data > silos. Still, it's technology worth getting familiar with, especially as > there are likely many problems where "enough data" doesn't have to be "scary > data". TensorFlow looks intriguing to me, although I haven't dug in (and > haven't kept up with neural net research in a very long time). > > > -david > > > > > On Thu, Jan 21, 2016 at 9:23 PM Sandip Kamat <[email protected]> wrote: > > A couple thoughts on AI and us -- > > Thinking about the connected devices world in general (and personally, with > my current focus on Vaani in particular), it is hard not to think about AI. > In Vaani team, we are (on purpose) not looking into AI in that context for > various reasons including - we need to focus for voice enablement first and > AI is a vast area in itself. (In fact, AI is playing a big role in natural > language understanding). However, it is an area that we should definitely > keep an eye on for the sheer impact it could have on the tech world in > general and users + our values in particular. As the big players collect > insane amounts of data from users and combined with the power of AI, start > predicting your needs before you think of them, it could be a slippery slope > for privacy, user choice and dubious practices around user consent. > > On that note, the openAI announcement in December is interesting. It is > backed by tech billionaires but they say their goal is "to advance digital > intelligence in the way that is most likely to benefit humanity as a whole, > unconstrained by a need to generate financial return.". This is great. Sounds > a lot like how Mozilla would think. However, there are not many code snippets > or plans/roadmaps available yet. They say they will "collaborate with others > across many institutions and expect to work with companies to research and > deploy new technologies." I am unsure if anyone has any insights on their > plans yet? Sounds worthwhile to get some insights. > > Secondly, if some day Mozilla does get involved, what would be the ways our > connected products could benefit from something like OpenAI? thoughts? > > (For me, I would like my own API to that intelligence that I can control my > data access with, train/untrain my models and you know...make it work for me, > not the other way around). > > -Sandip > _______________________________________________ > > dev-fxos mailing list > > [email protected] > > https://lists.mozilla.org/listinfo/dev-fxos _______________________________________________ dev-fxos mailing list [email protected] https://lists.mozilla.org/listinfo/dev-fxos
