ReactOS Is Finally Able To Build Itself

[phoronix]

Phoronix: ReactOS Is Finally Able To Build Itself

ReactOS, the "open-source Windows" operating system re-implementation, is now able to finally self-host itself in fully compile ReactOS from ReactOS...

http://www.phoronix.com/scan.php?page=news_item&px=ReactOS-Full-Self-Hosting

[GraysonPeddie]

ReactOS building ReactOS! Later on, we could have robots building robots!

[tildearrow]

ReactOS building ReactOS! Later on, we could have robots building robots!

https://reprap.org/wiki/RepRap

[lolren]

Phoronix: ReactOS Is Finally Able To Build Itself

ReactOS, the "open-source Windows" operating system re-implementation, is now able to finally self-host itself in fully compile ReactOS from ReactOS...

http://www.phoronix.com/scan.php?pag...l-Self-Hosting

What would really be cool, is to try a demo directly on their website, in a browser! If you can do it with win98, you can do it with Reactos!....

[anarki2]

ReactOS had great potential... around 15 years ago. Since then, everyone moved on. Too bad.

[Guest]

That's a very basic milestone. Nevertheless, excellent job. Good for them!

[starshipeleven]

Detroit: Become Human

I'm so annoyed by "AI" like that, apart from being a not really novel concept, they aren't realistic.

That's a "tin man" (a human disguised as a robot, which is what most "self aware robots" are in fiction), not anything remotely resembling what an actual AGI product would be.

Most of the stuff that would require "high intelligence" for a human would be actually dealt with by specialized modules (text and speech on hundreds of languages, or even complex math or whatever else), making a single AI software capable of all things combined is going to be ridiculously more expensive and complex, for no real reason.
You would have a "decision-making" or "analysis" core that would actually do the "high level decisions", and then add modules to have additional functions on top of that, be it image recognition, navigation, motor control, speech/text, math, and so on and so forth.
The "high level decision" part can actually not even be a true AI at all and just have a bunch of pre-programmed "high level" reactions and it would work perfectly fine for most jobs, as the complex part of actually identifying what situation it is in and actually execute the "high level" reaction would be the job of other modules (each running its own specialist AI software).
For example, a nanny-bot hears a child crying.
The audio processing module figures out that:
1. it is the voice of a children it is taking care of
2. it is a cry, not a shout, not a whisper.
This is fed to the "decision making" unit, that has a pre-programmed reaction of "navigate to his location" to the above inputs, so it fires up again the audio processing module to try to get a bearing or an actual position, and then sends this data to the navigation module (and updates the bearing or position as it moves) that will then move the unit over to the crying baby, and it will also start to interrogate its visual recognition module to actively look for the child, and be fed answers like "no blood, normal posture, object on the floor recognized as toy at X distance from child, child face shows anger emotion" and so on and so forth.
Let's say that the inputs would cause a pre-programmed reaction of "talking to the child to get why he is angry and calm him down".
This means it will fire up the text/speech module with orders to ask why the child is angry and to calm him down, and this module would actually be a full-blown chatbot capable of dealing with humans (something like Cortana or Siri or Ok Google that runs locally and does not suck), that would then deal with that stuff on its own and relay back answers or the child's requests in the dumbed-down and standardized form that the decision-making unit can actually deal with.

The hard part in such a system (and in actual biological brains too, if you look at the areas dedicated to fuctions) is actually the categorization of the sensor input, not the decision making.

Which is why each module is actually running its own AI system. Sound analysis, navigation in a 3D world with crappy sensors like a human, and visual analysis to recognize objects and faces and stuff are pretty complex tasks, but they are just dumb categorizing systems, providing canned answers for a relatively dumb decision-making unit over some sort of standardized protocol.

With such a system you aren't going to require a full-blown AI unless the situation is very out of the ordinary, if such a situation is encountered the android would fall back to asking to his owner/employer for orders, which is also done by human workers in many fields and not particularly strange.

This way you can both keep costs down (mass-producing the same modules running the same software), which is very important for these high-complexity things, AND keep the complexity of each AI software down. The offshoot of this is that you also negate pretty much negate any "robot becoming human" scenario as the "decision making" part isn't going to run particularly complex software to begin with, let alone be an AI at all.

Also the ability to learn new things might or might not possible depending on each module's capabilities (most likely limited or even not possible at all because of hardware or commercial limitations).

[Kendji]

A great milestone, though still long-way from being a viable option to windows

[Kendji]

ReactOS had great potential... around 15 years ago. Since then, everyone moved on. Too bad.

From my understanding, there's only one or a handful of devs working on this while other OS's have bigger resources

[waxhead]

People seem to think that software that is not maintained anymore is obsolete and not useful. What a load of crap. ReacOS has it's uses and I am very happy it exists. It allows me to run Windows programs on a free Windows re-implementation. Very useful for businesses as well , no need to let Microsoft shove stuff down your throat and obsolete it just a few years later. Thumbs up for ReactOS by all means!