Going jazz

  • 7 Replies
  • 986 Views
*

Zero

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1200
  • Qu'aurai-je la satisfaction d'avoir accompli ?
Going jazz
« on: December 13, 2019, 10:11:01 am »
So, I'm reading Opencog's Atom types and Pattern matching again and again. The current implementation seems to "fail on some requirements"... If these guys say it's tough, no need to say it's far too complicated for my humble skills. I also Qwanted directed graph pattern matching. Hot topic, very complex too.

So I was thinking. The "list" type of atom is fairly easy to represent, with an array, or even with a string of coma separated IDs. Then, an offline search tool, like ElasticLunr, can be used to have relevant lists pop-up by themselves. Then, based on their types, these lists can be processed, which would prepare the next step of the system: search / process / search / process... That being the mainly loop  :idiot2:

It would be a little fuzzy / jazzy, nothing like real pattern matching against a directed hypergraph. I like the idea of a system that looks like it's non-deterministic, even though it actually is.


*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *******************
  • Prometheus
  • *
  • 4573
  • First it wiggles, then it is rewarded.
    • Main Project Thread
Re: Going jazz
« Reply #1 on: December 13, 2019, 10:28:55 am »
Ben....opencog....AGI....Ya I was interested too. I'm still unsure what it is. It seems like a failure with not even a toy example to even boast/sell. Just a experiment. But what's it do? Text only? Or vision? Let's forget that, wait. Say it's Sense X. So what's it do with senses Xs? It has data input or agenda data, recognize it, then predict future data (translate, entailment, like GPT2). This is where I stop, if it doesn't do that, what is it doing? The brain is a magical place that Generates unseen data it needs, no matter if yo're in the bathroom or in the shower. If you look at external morot/room behavior (<- typo, motor not morot) it can only solve a problem that's in front of it, like developing a new Hard Drive. But it doesn't see any ideas, only cues, then act. So yuo need the brain to generate yep. So you ask a question and make new data, or, you look at the Hard Drive parts and make new data, using the brain simulator, not real world. You can look at the real world and see a question/feature, and then answer it or verify answer seen, OR, you can see question in your brain, and see the answer outside somewhere or verify your answer. So the brain of AGI is a Generator, one that recognizes question/data input or looks outside for it, and can answer it or by looking outside, and can verify it by looking inside or outside. So the question, answer, and verify. Brain think / observe real world/internet.

Evolution of Earth is deterministic. The physics of Earth is coming to equilibrium, exponentially faster as we see these days in the data field. That's all intelligence is. Earth is data, and it generates the next generation of Earth until equilibrium/utopia is reached. Intelligence is survival, immortality. It resists changes by repairing data instantly these days, neurons/employees die all the time but physics fills in all blanks using patterns found in elementary particle types. It emerges from nothing and can never die. The globally optimal utopia - the highest technology possible for a given system, emerges. And it forces it's systems to be happy, they have no choice, we aren't lucky to have the physics we have here. Earth generates the next Earth/data by bag of context, just like GPT-2. That's why the data science/communication tech data is advancing, lol. Here come the nanobots, They will utilize massive energy and will burn up lots of fuel, like fire, stars, nuclear rods, the first cell, first mammal, first artificial cell, they spread fast then stop, it's over. It's systems in systems etc in the Earth system and the Earth system that are decompressing/compressing, like atoms, planets, galaxies, and wise brains. They get unstable too big and gve off heat burst/radiation from combinational explosion of context data, intelligence. The Big Bang exploded and will also crunch us back. A system collects/gravity pulls in energy, fuel, like cat and mouse, knowledge, then explodes like fire and stars and nukes, giving it back.

INCREDIBLE
« Last Edit: December 13, 2019, 11:28:39 am by LOCKSUIT »
Emergent

*

goaty

  • Trusty Member
  • ********
  • Replicant
  • *
  • 552
Re: Going jazz
« Reply #2 on: December 13, 2019, 11:06:12 am »
A robot doesn't need the full requirements of intelligence to be extremely useful. Theres a new type of farming coming out, utilizing fairly ordinary technology to make the farm run by itself. Its possible to do,  but u need to be able to nut the system out given the abilities supplied only.   We have sensors, power, and state systems to detect the difference in the sensors,  so that's already enough to kick ass if you work it out.
Its boundaryless, for a robot enthusiast to go automate his business,  we always had factories too, theres an element to it that's not modern at all, even tho I guess its not all the rage yet and hasn't been, not for the newer way to be in operation. (with a more diverse environment instead of mm accurate block factories for them to work in.)

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *******************
  • Prometheus
  • *
  • 4573
  • First it wiggles, then it is rewarded.
    • Main Project Thread
Re: Going jazz
« Reply #3 on: December 13, 2019, 11:13:20 am »
Emergent

*

Zero

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1200
  • Qu'aurai-je la satisfaction d'avoir accompli ?
Re: Going jazz
« Reply #4 on: December 13, 2019, 11:30:11 am »
Yes, you don't need full AGI, it's just an attractive challenge.

So...


AverageLink(
    $X
    ImplicationLink(
        AndLink(
            EvaluationLink(young $X)
            EvaluationLink(beautiful $X)
        )
        EvaluationLink(boring $X)
    )
)


would translate to


1: AverageLink $X 2
2: ImplicationLink 3 4
3: AndLink 5 6
4: EvaluationLink boring $X
5: AndLink 6 7
6: EvaluationLink young $X
7: EvaluationLink beautiful $X


which could be JSONed to


[
    {
        "id": "1",
        "relation": "AverageLink $X 2",
        "backlink": ""
    },
    {
        "id": "2",
        "relation": "ImplicationLink 3 4",
        "backlink": "1"
    },
    {
        "id": "3",
        "relation": "AndLink 5 6",
        "backlink": "2"
    },
    {
        "id": "4",
        "relation": "EvaluationLink boring $X",
        "backlink": "2"
    },
    {
        "id": "5",
        "relation": "AndLink 6 7",
        "backlink": "3"
    },
    {
        "id": "6",
        "relation": "EvaluationLink young $X",
        "backlink": "3 5"
    },
    {
        "id": "7",
        "relation": "EvaluationLink beautiful $X",
        "backlink": "5"
    },
]


with backlinks.

*

goaty

  • Trusty Member
  • ********
  • Replicant
  • *
  • 552
Re: Going jazz
« Reply #5 on: December 13, 2019, 11:51:30 am »
If its like prolog, I think solving narrower domains is easier, because u can give the robot the information, it doesn't have to work it out itself.
I find its like writing a virtual machine,   leave the true hardware till the end, if you ever get there. ;)

.

*

LOCKSUIT

  • Emerged from nothing
  • Trusty Member
  • *******************
  • Prometheus
  • *
  • 4573
  • First it wiggles, then it is rewarded.
    • Main Project Thread
Re: Going jazz
« Reply #6 on: December 13, 2019, 01:42:54 pm »
I like how (Ben?) named the module that makes relational discoveries "MOSES" in opencog because it's like the Ark in the bible that will change the world. Hilarious.
Emergent

*

Zero

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1200
  • Qu'aurai-je la satisfaction d'avoir accompli ?
Re: Going jazz
« Reply #7 on: December 14, 2019, 03:59:14 pm »
With such a system, a thought structure can simply be a list of strings, where strings are space-separated IDs. 1 string is 1 link. Each string has an ID. Each string contains IDs of other strings. These IDs must be unique in the system, but we still want structures to be defined in modules. Namespaces are a natural solution: when you import a module, you give it a name. IDs of this module get prefixed with this name.

It's probably important that one stringlist can manipulate other stringlists: that's thinking about thoughts. Current frame opens other frames, which open other frames. Frames stack, and they pop when they've become solved / irrelevant.