Jim, It is not an attack ad hominem but your ruminations are becoming increasingly circular and viciously so. Just when I thought we had the simple project on track you found that either/and the issues are too subtle and the tools too gross. I think you need a few modernity updates too when it comes to applied computing, so here we go. There are tons of "good people" out there that developed hosted or virtual services so that you dont have to waste your time or money. Hadoop is of course here to stay and Hortonworks were working on a Windows version, but they may be giving up, instead virtual machines of their platform are good for most purposes http://hortonworks.com/thankyou-hdp20-cp/ . As we discussed before, a real-time approach may be more fruitful than a kind of AGI datawarehousing http://www.information-management.com/dmradio/forget-the-warehouse-why-the-focus-should-be-on-data-flows-10024732-1.html . The good people at http://www.mongohq.com/home will let you have 0.5GB on MongoDB for free (and lets not forget such projects are moving fast, it may already be better than what you remember), and I guarantee you that if your program makes any progress towards general intelligence it will discover on its own ways to register for further free accounts as needed with fake IDs. Redhats OpenShift makes most of the open source servers and services avalaible at a free tier, God Bless America!
In all truth, if one was to write a simple AGI program one would be advised to write is a module/add-on to some existing open source server, even as an Apache module, in order to have all the capabilities of (network) server programs without the overhead. You could even build simple programs on top of, oh the sacrilege, OpenCog! AT On Thu, Aug 22, 2013 at 9:22 PM, Mike Archbold <[email protected]> wrote: > This is a good summary of big data / hadoop. I noticed that Darpa is > on the Big-Data/Jeff-Hawkins-Prediction-compression bandwagon with > emphasis on time thrown in. > > On 8/21/13, Jim Bromer <[email protected]> wrote: >> http://strata.oreilly.com/2011/01/what-is-hadoop.html >> Hadoop commentary. >> >> On Wed, Aug 21, 2013 at 1:35 PM, Jim Bromer <[email protected]> wrote: >>> Well I was just telling someone that I should not have referred to my >>> "database management system," and instead should have used a phrase >>> like, "data management system," because most people would think of >>> something like a relational database system (like SQL) when I said >>> "database management system," whereas the phrase "data management," >>> could refer to modern data management systems like big data management >>> or something else. In order to start with something simple I am not >>> thinking of going big data myself, but my data management system is >>> not relational either. I was reading that Hadoop is a big data >>> management system (although I notice that they do refer to it as a >>> "database management system," in one link that I looked at even though >>> it was designed for situations where data might not fit into tables) >>> and it is open source but it looks like it is designed for an Apache >>> server. I did download an Apache server for my Windows computer and >>> it was not an end-of-days installation nightmare, but it did take some >>> hiking on the learning curve. I don't think it is running right now >>> but I really can't remember how it works. I might have even >>> uninstalled it. Anyway there should be some other open source data >>> management systems as well. >>> >>> I also looked at Mongo but there was something I did not like about >>> it. I did not stay with that very long since I had trouble getting it >>> to run from Windows but there was also an issue where I did not know >>> how to enforce a type so that if you started out with an array of >>> numbers, for example, you could start adding strings to the same >>> group. I guess you could, at the very least, do a read type check for >>> every write but I didn't get that far. >>> >>> I think that if you don't have a good basic data management system >>> then you are not going to be able to write a good AGI program. It is >>> just about a pre-requisite. There are trade-offs for writing your own >>> vs finding something on open source that you could use. I think >>> carefully looking for something from open source might pay off, but >>> then you will still have to create a higher tier of management for the >>> system anyway, so it is a very subtle issue. And what I am finally >>> starting to understand is that the preparatory work is vital as long >>> as you have some good AGI conjectures to test when you get there. Jim >>> Bromer >>> >>> >>> On Wed, Aug 21, 2013 at 6:11 AM, John Rose <[email protected]> >>> wrote: >>>> Speaking of such, and related: is there a SQL that replaces SQL, does >>>> anyone >>>> know? A language like SQL that has its utility and ease of use, >>>> powerfulness >>>> and flexibility that is more cognitively oriented? If not, what an >>>> opportunity exists now contemporarily to promote one. Maybe a language >>>> that >>>> emerges from NoSQL databases, but then it looks like SQL is doing that >>>> :) >>>> >>>> I don't know though... >>>> >>>> John >>>> >>>>> -----Original Message----- >>>>> From: Jim Bromer [mailto:[email protected]] >>>>> >>>>> The problems I had stepping through the templates were related to the >>>>> complications of the relations I was using for the AGI part. The >>>> so-called >>>>> Concepts used different classes that all used the same templates. When >>>> used >>>>> as 'Concepts' I found that there were times when three or more arrays >>>> (from >>>>> different classes) were involved in a single operation and as a result >>>>> I >>>> there >>>>> were times I could not figure out which class 'owned' the template that >>>> was >>>>> being called. So this was a preliminary AGI issue (if my theories are >>>> workable). >>>>> It was really confusing since the operation of one class could be >>>> dependent on >>>>> the operation of the others. I could have tracked what I was doing with >>>>> a >>>> paper >>>>> and pencil or something but it was so confusing that I could not even >>>> remember >>>>> what the different algorithms were doing. By defining the common >>>> operations >>>>> of the arrays as individual classes I should be able to debug >>>>> complicated >>>>> interactions of the classes better. And the differentiation of the >>>> classes is >>>>> easier to track as well. And finally the debugging representation of >>>>> the >>>> stack is >>>>> more straightforward. You are not supposed to define an operation as a >>>>> template until you have all the instantiations debugged but my >>>>> intention >>>> was to >>>>> simplify the original design process. It didn't work. >>>>> >>>>> On Sun, Aug 11, 2013 at 10:44 PM, Jim Bromer <[email protected]> >>>>> wrote: >>>>> > Well, I guess you have to use joins for larger SQL databases but I >>>>> > don't define relations. I can't remember what I did to cause the >>>>> > errors that I first got with Access, but I was still running into SQL >>>>> > errors the last time I used it and I still have trouble with SQL >>>>> > helpers. Since some SQL errors are so well known there should be a >>>>> > SQL compiler or interpreter that can point out the most obvious >>>>> > mistakes or the language should be changed so that syntax that causes >>>>> > common beginner's errors should not be allowed without using a >>>>> > keyword >>>>> > to designate it. But of course they can't change the basics of the >>>>> > language because of legacy code. >>>>> > >>>>> > I would never program with SQL if it wasn't still in vogue for >>>>> > business operations. - Jim Bromer >>>>> > >>>>> > On Sun, Aug 11, 2013 at 9:03 PM, Jim Bromer <[email protected]> >>>>> wrote: >>>>> >> When I first started using Access for a business related project 10 >>>>> >> to 15 years ago I could not understand why I got such bad results on >>>> most >>>>> joins. >>>>> >> It took me months to discover that it was an issue with Database >>>>> >> programs that were originally designed in the 1970s. I became >>>>> >> outraged, because, for some reason I had once been familiar with the >>>>> >> problem. (Perhaps I had tested some ideas in the 1980s and >>>>> >> discovered >>>>> >> through reading that the particular error that annoyed me was >>>>> >> typical >>>>> >> with that kind of db programming and the discovery that they were >>>>> >> still using the same thing 20 years later seemed absurd to me.) I >>>>> >> eventually learned to do some Access and SQL programming but I >>>>> >> usually use fairly simple commands (I don't use joins in my >>>>> >> commands) >>>>> >> and then mix the data in the programs where I have more advanced >>>>> >> control over the actions. (Of course more complicated commands may >>>>> >> be used to increase security. But you would only use those for well >>>>> >> defined and debugged actions.) >>>>> >> >>>>> >> It would be like suggesting that Ben Goertzel's programmers use >>>>> >> Fortran. (I mean - like - whaaatt?) >>>>> >> >>>>> >> I also figured that if I wrote my own database management system I >>>>> >> could make it much more efficient and that is still probably true. >>>>> >> On the other hand, since I haven't proven my ideas even in the >>>>> >> simplest preliminary tests it would have made much more sense to >>>>> >> start with very simple list management algorithms. Once you learn >>>>> >> how to save data and retrieve data to a file (it's not getting >>>>> >> easier) it is probably simplest just to create arrays and save them. >>>>> >> The only problem is that you then need an index into the lists (as >>>>> >> well >>>> as an >>>>> implicit system of organization of the lists). >>>>> >> >>>>> >> I thought that if I got a basic system working and then put it in >>>>> >> template form (I don't know if you know C++) then I could just >>>>> >> implement the individual components of the db management system as >>>>> >> if >>>>> >> it were a high level script. However, it took longer than I thought >>>>> >> and once I started using it on some tests the errors became very >>>>> >> complicated. I could not get the compiler to give me any idea what >>>>> >> the values of the data objects of the template algorithms were when >>>>> >> I >>>>> >> was in debug mode and I didn't know how to write parallel algorithms >>>>> >> to display what the values of data were when I stepping through it. >>>>> >> (I thought about adding a dump as an unused branch and then since >>>>> >> the >>>>> >> compiler did step through the template, I could just use set the >>>>> >> next >>>>> >> statement but I was burned out by then.) And because you want to >>>>> >> use >>>>> >> little variations in the different kinds of lists (to make it more >>>>> >> efficient) you are going to want to do a lot of differentiation >>>>> >> anyway. >>>>> >> (Templates are good when you use data of different types but you >>>>> >> intend to use the algorithms in the same kind of way with only minor >>>>> >> variations.) >>>>> >> >>>>> >> So comparing the use of arrays vs using some db management system, I >>>>> >> would say if you know how to save and retrieve arrays, then use >>>>> >> them. >>>>> >> However, since the arrays are probably going to be huge for an AGI >>>>> >> program you are going to have to distribute them and that means you >>>>> >> need to use indexes into the arrays that are saved to the file in >>>>> >> different segments. And you need some indexical information that >>>>> >> will help you locate some particular objects in large arrays. But I >>>>> >> would say start with something simple and test some of your ideas >>>>> >> out >>>>> >> before you start making the db management system more complicated. >>>>> >> But I am almost up and running so it would not make sense for me to >>>>> >> try >>>> to >>>>> start over again to try to make it even simpler. >>>>> >> >>>>> >> So I explained why I had absolutely no desire to use SQL and still >>>>> >> wouldn't but in answering you I reexamined a possible strategy to >>>>> >> simplify my programming that I might be able to use. Even if I can't >>>>> >> use this method of simplification in the database management part >>>>> >> of >>>>> >> my program since it is mostly written I could use similar strategies >>>>> >> in the 'conceptual' management systems of the (planned) AGI parts.- >>>>> >> Jim Bromer >>>>> >> >>>>> >> >>>>> >> On Sun, Aug 11, 2013 at 5:27 PM, Piaget Modeler >>>>> >> <[email protected]> >>>>> >> wrote: >>>>> >>> >>>>> >>> Venturing into the Void.... >>>>> >>> >>>>> >>> "Umm...why not use MySql or Sql Server or Mongo DB for your object >>>>> store?" >>>>> >>> (Some of these are free.) >>>>> >>> >>>>> >>> ~PM >>>>> >>> >>>>> >>> ----------------------- >>>>> >>> >>>>> >>> > Date: Sun, 11 Aug 2013 17:18:33 -0400 >>>>> >>> > Subject: [agi] Re: My Relatively Simple AGI Project >>>>> >>> > From: [email protected] >>>>> >>> > To: [email protected] >>>>> >>> >>>>> >>> > >>>>> >>> > I am trying to refurbish a database management program that I >>>>> >>> > wrote a few years ago and I keep finding additional lists >>>>> >>> > (indexes) that I have to implement (the class of each of the >>>>> >>> > lists >>>>> >>> > is almost complete) and I just cannot keep it simple. I haven't >>>>> >>> > even started testing any of my AGI theories and the program is >>>>> >>> > already becoming too complicated. I intend on using a lot of data >>>>> >>> > related to a concept (a concept like collection of related data) >>>>> >>> > because one basis of my theories is that it takes a lot of >>>>> >>> > knowledge >>>> about >>>>> a subject to 'know' >>>>> >>> > one simple fact about the subject. So this means that all my >>>>> >>> > lists >>>>> >>> > will add to the complications of the program once I start testing >>>> it. >>>>> >>> > >>>>> >>> > The non-programming resolute skeptics do not realize how >>>>> >>> > complicated a "simple" program can be. My database management >>>>> >>> > program is simple because I am using variations on the same basic >>>>> >>> > algorithms. In fact, the first step of my refurbishing the >>>>> >>> > program >>>>> >>> > was to take the class for the lists and the indexes out of a >>>>> >>> > template that I had written for them and use write individual >>>>> >>> > algorithms for the common algorithms that I had defined in a >>>>> >>> > template 6 or 7 years ago. I found that I did not have enough >>>>> >>> > details to debug the system using the templates even though they >>>>> >>> > were working well and I was not able to figure out how to keep >>>>> >>> > track of run time errors when there was a problem. (That is not >>>>> >>> > an >>>>> >>> > impossible task, but it was just one more thing multiplied over >>>>> >>> > and over again.) >>>>> >>> > >>>>> >>> > Programming is complicated. It may turn out that don't have good >>>>> >>> > AGI because the management system for a program that is supposed >>>>> >>> > to be creative is in itself going to be extremely complicated. So >>>>> >>> > people get stuck into traditional programming methods just >>>>> >>> > because >>>>> >>> > they were partly designed to make programming a little more >>>>> >>> > simple. - Jim Bromer >>>>> >>> > >>>>> >>>>> >>>> >>>> >>>> >>>> ------------------------------------------- >>>> AGI >>>> Archives: https://www.listbox.com/member/archive/303/=now >>>> RSS Feed: >>>> https://www.listbox.com/member/archive/rss/303/24379807-f5817f28 >>>> Modify Your Subscription: https://www.listbox.com/member/?&; >>>> Powered by Listbox: http://www.listbox.com >> >> >> ------------------------------------------- >> AGI >> Archives: https://www.listbox.com/member/archive/303/=now >> RSS Feed: https://www.listbox.com/member/archive/rss/303/11943661-d9279dae >> Modify Your Subscription: >> https://www.listbox.com/member/?&; >> Powered by Listbox: http://www.listbox.com >> > > > ------------------------------------------- > AGI > Archives: https://www.listbox.com/member/archive/303/=now > RSS Feed: https://www.listbox.com/member/archive/rss/303/14050631-7d925eb1 > Modify Your Subscription: https://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
