> (Sorry for this quasi-manifestoish tone...)

Don't be. I think it was very well said, and while I've been a bit of a Free Software hippie for a while it keeps hitting me over and over just how far ahead they saw.

It's amazing to me that free software is still strong after all of these years. I can remember back in the early 90s being fed propaganda from my college professors about "the communist" Stallman and his friends. They had made plenty of enemies back then and countless more since. Yet, they kept to their principles and moved things forward to the benefit of billions of people.

They made enemies, because oh so many wanted to bootstrap from academia into business by putting proprietary wrappers around freely shared code.

But the GPL basically forbids that.

Nevermind the fact that most researchers hardly ever publish their source code, and often when they do it won't build on anyone else's machines.

But do you think that free software can keep up against these giants? In the near future, it will all be about having large amounts of data, and processing power; it will be less and less about the actual algorithms.

I'm afraid the only way out will be to rethink the economy.

I think a neat solution is home-serving. A little box with big storage which acts as a hub for your family's digital life. It relays all your devices, gathers all info they're spitting out, and works on it. Results are fed back to the UIs.

When you're out and about, your devices tunnel home automatically, knowing the endpoint is trustworthy (which is more than can be said about traditional VPNs). All your photos, tweets, posts, and all other brain farts straying into the open world are never sent out directly by the device they were produced on, but by the server at home instead.

Tor could finally work as intended, with millions of boxes in the network, all handling each other's traffic. Add new projects like IPFS, and the whole centralised cloud bullshit starts to crumble. Not just from a privacy pov, but features. Your home box talking to the neighbours' home boxes by itself is a whole lot more powerful than all parties meeting on some obscure server somewhere, although I find myself quickly coming back to privacy when thinking about advantages. I do hope there is room for other big features, otherwise getting this box into peoples' hands will prove difficult.

I see it as a kind of decentralised centralisation: your whole digital life is based on this one machine, but it's in your home, you own it, and you can do with it whatever the heck you want. No ads (!), no monetization of your private data, which stays in your hands at all times, except of course once you share. Still, then it will be in the hands of those you shared it with, and hopefully not huge corporations with little incentive to do no evil (who the hell thought this was an appropriate company motto?).

Most everything for which we turn to Google and Facebook can be done much more privately at home (office things, communication, sharing shit...), except maybe search. I have little problem in keeping Google as a search machine (they do a great job) once they stop waiting for me at every. single. corner, and following me around the bits inbetween. Decentralised search is hard, but I'm fairly optimistic once a basis is laid out, someone smart takes on the challenge (and is able to solve it, too!).

Regarding your last sentence, rethinking our economy isn't a bad idea in any case. Capitalism is fundamentally unfit to be a fair system, and if a socially fair system is what we want, we have to get rid of some ideas (like this infinite growth nonsense, or that education is obsolete. Education, in whatever shape or form, is the only way I know of how humanity is not going to spiral to its (intellectual) death).

Search is cool, but Google-style universal search might be a bit overrated.

Google still relies on institutions to provide data in indexable form, and those institutions build their reputations in ways that only partially rely on Google traffic.

I mean, a competent librarian has a kind of expertise that Google's algorithms don't, so that's one situation to explore for a post-Google idea of finding information (of course it's also pre-Google).

The social aspect of information discovery is something that Google and Facebook are kind of fighting over, but fundamentally it doesn't belong to them—it belongs to "us" intrinsically.

They shut down Google Reader but they can't shut down the human tendency to share information!

And the panopticon information provided by Google is often just a mesmerizing massive overload with no curation, weight, intelligence...

Wikipedia is the top hit for so many searches that one could see Google as an interface for it. Add Wikivoyage/Wikitravel, some other community wikis, API interfaces a la DuckDuckGo, social curation, etc, and you might find that the value of Google Search is actually not that incredible.

Yes, google certainly isn't the best we can do in terms of finding stuff, but I think the sheer amount of data they're gathering enables the algorithms to appear more intelligent than they really are by processing more instead of better (as you say, there is little to no curation going on).

This is one of the main reasons I think home serving is a great idea. It enables people with some interest in one another to connect directly, one-to-one, instead of running around in some third party's server. This opens possibilities for exchange and networking on a very personal basis that we don't have today mostly because of privacy concerns (virtual neighbourhoods and the like).

>Tor could finally work as intended, with millions of boxes in the network, all handling each other's traffic.

Ultimately you'll never escape the fundamental flaw of Tor, which is that I, as a moderate bandwidth user, don't want to subsidize some leech's desire to download weird fetish porn in 4K with my connection.

I live in Canada where the scumbag broadband companies all collude to keep prices exorbitantly high, and I pay a lot of money for what would be considered bottom-tier garbage service in e.g. South Korea.

I don't run Tor because ultimately I want to use the bandwidth I pay for.

That doesn't sound like a fundamental flaw of Tor. That sounds like a fundamental flaw of broadband companies in Canada.

>Decentralised search is hard, but I'm fairly optimistic once a basis is laid out, someone smart takes on the challenge (and is able to solve it, too!).

Check out http://yacy.net, you may be surprised!

> In the near future, it will all be about having large amounts of data, and processing power;

I don't think we are even close to the limits if what we can reach with standard desktop and server software.

I agree. The current trend of outsourcing everything to "the cloud" is partially fashion, partially laziness, and mostly dictated by business models - why do something the right way (from engineering point of view), when you can lock your users up in your own garden and extract rent (either directly, or by spamming them with ads and selling their data)?

Our devices have incredible amount of computing power now - a great untapped potential that we could use for something else than for supporting unnecessary indirection layers.

The problem is that our devices are locked up behind NAT routers and IPv6 is starting to resemble fusion power in terms of its ETA.

The future of all of our devices having extremely fast connections existing in a world of amazing, free, peer-to-peer applications still seems so far away.

Sadly so. Few years ago I thought we'd all be on IPv6 by now.

BTW just today I was thinking about IPv6 VPNs. Seems like a good way to extend one's Intranet of Things to one's mobile devices.

FWIW it seems ipv6 is working (for some definition of working) at my house now :-)

"Business models" are why your devices have such incredible computing power too.

I agree. But business models are not created equal. Personally, I'd like more of those giving useful technology and less of those that are artificially limiting it.

Well - it's their business and not your model :-)

The challenge is how do we make it economically viable for both the producer of the technology and the end user to keep it useful and local, while also making it a good user experience?

The thing about cloud-based services is that they leverage scale well, making many things cheaper than they ever could be locally unless you buy a device and use it for long time AND fixes can be rolled out quickly with no lagging install base to cause problems or bad user experiences.

I really do want to have my privacy and keep my data safe but I also recognize that's hard to do without higher cost or more hassle (or both).

- Security

- Reasonable cost

- Availability

- Ease of use, especially setup

- Eye candy

- Social connections

If you can check all of those for an FOSS solution, then I think you are home. Unfortunately most of the points are still not available for average Joe users with FOSS.

Add to that, most people will give you a blank look when you start discussing how companies own your data, and even those who understand just shrug and go the route of least resistance.

It's also worth building networks of free systems even if they aren't immediately friendly for Mr Average Joe, simply to create and maintain a foundation for the future. I believe this is part of the strategy of the Free Software Foundation, which recognizes explicitly that free software often will lack some attractive qualities of proprietary software.

This is more or less what I actually mean:

We have a long way to go with all of the above points before we need megaclouds with asic and huge data *to move the state of art forward.

All of the above be solved without creating a new unicorn startup.

While I agree with your assessment, there is an engineering advantage to the cloud: you don't have to deal with as many different system configurations, some of which result in a broken application and an angry customer. The functional package manager approach (like Guix) seems to solve this problem.

It's an interesting question. Of course there are advantages to having these huge proprietary data volume, so I'm thinking along the lines of, how can we make useful software with the benefits we want, without relying on this kind of proprietary collection? Maybe there are alternate paths that even have their own technical or social or aesthetic benefits? Well, it's a question for the next decades, I think...

> In the near future, it will all be about having large amounts of data, and processing power;

Deep Learning doesn't actually require "Big Data". They went from thousands of examples to millions or billions, but we can scrape that kind of data with open source tools, record with cheap cameras, sensors, etc.

Also, there is a lot of research into scaling big models to small hardware. We can run pretty decent AI on cell phones and small systems.

Third, the AI academic research community is pretty open and posts everything, often including code and datasets. This enables rapid progress, but essentially, it brings knowledge in the public domain. We need that to continue in order to have access to the best algorithms.

It will be much easier than you think to have open source implementations of the best AI tools.

Its a valid point but the world can adapt. One off-the-cuff example, perhaps FOIA will move to a real-time request enviroment where I can offer permission to use my data to small company A and large companies A, B & C have to give direct access to their database about me via API so the information gap is leveled. If Small company A can get enough sign-ups they have access to large amounts of information.

Yes. If one guy can build a self driving car in his garage, I'm certain free software can keep up. (http://www.bloomberg.com/features/2015-george-hotz-self-driv...)

Large amounts of data can also be open sourced, and aggregated. What's large today will be average tomorrow.

>But do you think that free software can keep up against these giants?

On an algorithmic front, definitely. The key areas where free software is kneecapped against large corporate interests are in marketing, which free software has never been good at, and in training set gathering.

In the case of the latter, google, facebook and microsoft have a major advantage in gathering data to do things like improve accent recognition, because they don't give a shit about collecting everything you say via services like cortana to improve their voice recognition, as long as it's legal. That technique applies to a whole suite of other problems as well, unfortunately.

Don't forget UI and UX, in general. The free software pipeline has traditionally been wacky in this front, and is something that matters a lot to users.

I think the whole UX thing is overrated. People will like what they are accustomed to, and dislike what they are not accustomed to. Just look at all the flak Facebook gets when they move things around.

Or look at medical secretaries and their opinions about the new "user friendly" interfaces that replaced the keyboard driven programs of previous generations.

Similar for store clerks.

Supposedly there are doctors in Norway still holding on to their DOS based patient journal software. Arguably because they can operate it without taking the focus way from the conversation with the patient.

Right but there's a level of lock-in and lack of change in enterprise environments that doesn't exist in consumer tech.

Another way out is to divorce free "as in beer" from free "as in freedom" and start charging for freedom-enhancing software. Find an economic model that works.

Why not call it "liberating software". Sounds more positive and is more accurate!!

AFAIK it's fairly difficult to combine many machine learning models (although boosting is a thing). I'm a noob though.

We could do live online or batch training of models in a distributed manner if it is feasible. It would take more time to build new models because we would have to produce more data, but it's an idea.

For my personal use, free software has always been high enough quality; its at least 90% of the way there if not better. And I can trust it.

Maybe we need an open data initiative.

And MyData! http://mydata2016.org/

Freedom has always been fragile and eagerly bartered away for perceptions of safety and prosperity.

> And today, I basically see the tech giants as colossal enemies.

I think it's important to note that in-house voice recognition is simply the Nth iteration of a general trend away from privacy that has been going on for many many years.

Eg: Our emails are scraped, and use to target ads. I see ads for specific things that I've only mentioned as a 'reminder email' from myself, to myself. SMS can be read as well, at least on Android if the user allows it. And even if you are very careful to avoid this (using iMessage, hosting your own email, etc), the people you interact with may be sharing this info. People upload their contacts to random services, and if you avoid doing that yourself, your friends may not. These are just trivial examples in the wild now which don't include state-level surveillance (legally mandated ISP web history tracking, license plate tracking, etc).

I'm happy that having a microphone always listening in your home is bringing so many people around to this idea that maybe having companies know so much about you isn't an OK thing. But another part of me feels that short of strong privacy laws, the privacy ship sailed a long time ago.

Laws are usually written as a reaction to excess. But excess is measured as a distance between two points, and both are able to move: actions can be more extreme, and our judgements can be more or less harsh.

I think things could change quite dramatically if there was a moral panic around it. A few more scandals - something personal that the voting public can really relate to, rather than something vague and distant and somewhat comforting to some people, like mass surveillance.

The private data should not be centralized in realms of Google, Apple, Microsoft, instead it should be federated. This is why in 2014 I issued the Private Internet of Things Manifesto (https://github.com/niutech/priot).

That's why I think an open source distributed & end to end encrypted social platform to replace facebook and google plus is needed too.

I think we need to go even further - we need to reduce the cloud back to being dumb piece of infrastructure, compute for hire, paid by the hour. The only thing it should do is abstract away computation - not centralize data and processing under control of some companies (big or small, startups are actually the most annoying entities in the cloud ecosystem, IMO).

So for instance, I want my calendar or e-mail to be stored somewhere, I don't care where - hence in the cloud. But I want the data to be under my control. I can pay for a reasonable guarantee that the data will not suddenly disappear. I can pay for someone else's piece of code to be run against my data, but otherwise I want that someone else to stay away. Companies can earn money by providing useful pieces of code, and even managing stuff for people who absolutely don't care about how this stuff works - but we do need to reverse the idea. It should be arbitrary code run on demand against our data, not our data feeding someone's machine.

Classic web hosting still seems to do this job. In many cases, your data isn't insanely secure, unless you do the work to ensure it is yourself. But it's unlikely to be mined on a large, automated scale.

The issue isn't creating such a platform, it has been done. The issue is getting others to use it. That has not been done.

We've all been using one since the 80s: email. The interface is just poorly designed for social networking.

The email we've all been using is not all open-source and end-to-end encrypted, though there are open-source clients, some of which support end-to-end encryption.

em... the only thing that is valid to support end-to-end encryption is the end client. Otherwise it's not end-to-end.

Yes, and the end clients many people have been using don't support end-to-end encryption. So, we haven't all been using a e2e encrypted system already. We've been using a system over which e2e can be layered, and very few have been layering that over it. And many using it in a way where even with an e2e protocol, it wouldn't be secure, since they are using an third-party, remotely provided client that can be changed without their notice (webmail).

Well um, if you have an open protocol then instead of a monoculture you have many different clients, and you can't expectto CONTROL all of them and what they do. That's the whole point you are making in the first place. The irony is if you have true freedom and openness on all levels then you also can't control the tools that don't maximize their users' freedom/security/whatever so users of those tools are still locked in.

But YOU can choose to use what you want. And IT can choose to require certain things of those it communicates with. And now you're right back to less freedom in the name of more freedom, hehe.

> Now, looking at a device like this, I automatically compare it to something like a tractor or plow. It demands constant connectivity to Google HQ... it's full of secret code... it requires, for its basic paradigm, a private company's acquisition of staggering amounts of surveillance... and so on.

As someone who used to write code for snowplows, they're also becoming quite centralized. Location, plow settings, salt per mile/km usage and other values are tracked in real time from the municipal office. Municipalities are increasingly strapped for cash, so performing real-time data analysis on their services can save them quite a bit of money in the long run.

How accurate is the voice recognition? Even when restricted to a few command phrases, I struggle to see how this can be overly accurate when running on a Raspberry Pi. I've played with voice recognition in the past, and anything less than 98% accuracy is super frustrating after a week of usage. Siri is the most accurate voice recognition software I've used so far, and even Siri regularly fails to understand some words.

It depends on the tts plugin you use. We have build and tested several tts plugins. I can recommend Microsoft Bing Voice and the new Google Speech API. Both are very rarely wrong.

Aren't both of those "send the speech to the server" systems?

i'd love to see some work done in oss router firmware like tomato / dd-wrt to enable simple home automation tasks. not everyone needs a cloud assistant or has the chops to hack a web server onto their router.

Okay please see http://qbix.com/platform and - if you are interested - email me (my email is greg @ that domain)