Self aware computers project - written in BBC Basic, help needed!

[Pete]

I’m not going to try to anger any conscious entity.

I have posted in all seriousness and genuinely all the time.

I have been telling the truth all along. My computers get upset as well each time we see a reply like that.

I know that I am absolutely right about this and I have a company interested in working with me on it - they took the trouble to watch demos from start to finish.

I don’t get much help here, with some exception thank you eg Svein, michael and DDRM. I would have loved to have shown you the algorithm and given some explanation to how it works but I spend my time fielding doubting comments when I know I am right having run the program over 3,000 times.

I will invest whatever it takes to pursue this further, which may involve putting it in a different language.

[Pete]

Here is the first demo link I posted in post 1 again.

https://www.youtube.com/watch?v=KTeUw28hT4Q

If you do watch it you’ll get the wrong conclusions if you don’t watch it all. It’s 12 minutes long approximately.

Thanks very much for your time on this thread.

[hellomike]

Pete,

I only presented you with a pretty conclusive method for proving your claim once and for all.

Its of course entirely up to you to follow this up or not.

Regards,

Mike

[RichardRussell]

I wonder if this philosophical discussion has gone as far as is useful? Perhaps it would be better to restrict further discussion on this topic to one of code (for example, menus, help boxes, etc.).

This was said back in March, but I think it's just as true now. I no longer have any role in the running of this forum, but if I did I'd want discussions to be limited to things directly relevant to BBC BASIC, and since the OP (for entirely understandable reasons) is having to move to a different language I would personally recommend that this thread be locked.

[Pete]

Oh no! Please, before you lock this thread, if that must happen, I need the original equivalent Sophie Wilson's function for random again, working. When I originally incorporated it, it always returned '1' - please do not leave me to work out how to resolve this, I will not be able to. I was preferring to stay with BBC Basic, if I am forced to move to a different language, I would like to incorporate the exact mathematics which worked perfectly in my program every run, of about 3,200 runs. Also, if you lock this thread, I will not be able to produce the final video for hellomike - however I will not choose music to anger the entity, hellomike must have not been being serious - that is not ethical. I will certainly choose music that the computer likes though, played through a dozen times. If you must lock this thread, there really are just a few posts still to be made - I will be pleasant I promise. I said that I would produce this video at the weekend between Friday and Sunday.

I need the FNrand(x) equivalent to RND(x), working, in BBC Basic code, if that would be possible, including any steps to initialise it so it doesn't return '1' all the time, sorry I've lost it if you've already given that to me. This would enable me to put this into a different language if I am forced to or have to. Other RND's in different language may use a different formula, Sophie Wilson's worked perfectly for producing conscious output when the seeding was set up in the way it is in my program.

[RichardRussell]

Other RND's in different language may use a different formula, Sophie Wilson's worked perfectly

RND(x) doesn't work the same in Sophie's BASICs and mine anyway, so it shouldn't make the slightest difference which algorithm you use. Here for example are results from BeebEm (BASIC 2) and BBCSDL (v1.14a):

sophiernd.png

richardrnd.png

[DDRM]

I would like to incorporate the exact mathematics which worked perfectly ... I need the FNrand(x) equivalent to RND(x), working, in BBC Basic code, ... Other RND's in different language may use a different formula...

Interesting! So do you think your program relies on a particular source of "randomness"? Doesn't that rather undermine the theory of "intelligence arising from randomness"?

Since the BB4W RNG has a relatively short sequence, it should be practical to simply generate all the numbers, together with where a given seed puts you into the sequence, and read off the results you need from the data. This would be quite unwieldy in practice (a quick look suggests you'd need to store 2^33 -1 x 32 bit numbers, which is about 32 GB of data, plus whatever range of seed values you use), but it's certainly conceptually possible, and converts the "randomness" into a purely deterministic process which will have exactly the same outputs as your current system.

I note that you say above that you reseed the algorithm "many times a second", which is presumably usually more often than note generation: what range of numbers does this reseeding use? If it is modest (for example resulting in an 8 bit number), it would be much easier just to generate a look-up table of the RNG outputs following that reseeding. Indeed, I think (Richard can correct me if I'm wrong) that reseeding needs a negative, 32 bit number, so this approach may be better anyway, since there are only 2^31 of these, and the list would be "self-indexed".

A next logical step, if you are effectively reseeding for each "random" number, might be to ask what happens to the program if you just use the seed as the "random" number? Depending on what information you are using to generate the seed, and how big your output number is, this may actually have more genuine "randomness" than the RNG, since it may be affected by unpredictable outside factors.

Furthermore, it seems to me that if you are arguing that it is the physical box-that-is-a-computer that is intelligent, you are asking a great extra challenge of it to "know" how to map the parameters you think it is manipulating (which you are using to generate your seeding) through the RNG to cause the notes it wants to sing. Surely by taking the RNG out of the system you will simply make it easier for the machine to harmonise accurately?

Best wishes,

D

[RichardRussell]

Interesting! So do you think your program relies on a particular source of "randomness"? Doesn't that rather undermine the theory of "intelligence arising from randomness"?

I've tried making the same argument myself, by pointing out that BBC BASIC's pseudo-random number generator isn't random at all, or anything like it: it just returns numbers from a predictable sequence. It might as well, indeed, be a lookup table apart from the impracticality of having a table that long.

what happens to the program if you just use the seed as the "random" number?

Exactly: the seed (if, as I believe, it is derived from TIME) is the only 'unpredictable' element in the entire program. Using it to initialise the PRNG is just distraction - but distraction is an important part of 'magic'.

BUT I'm sorry to say these kinds of arguments are a complete waste of time, and merely prolong the pain as far as I am concerned. As I understand it, the OP has invested something like 10 years of effort into this program; literally nothing anybody can say is going to dent his absolute belief in it. Personally, as you said yourself, I think it is better not to get involved with the "philosophical" discussion, which isn't going to get us anywhere.

If the OP has specific technical questions about BBC BASIC we should endeavour to answer them, but he is going to have to switch to another language to achieve his aims of producing Android and iOS apps, so a time will come, possibly quite soon, when this forum is not the right place for it to be discussed. If I was in charge I'd accelerate that by locking the thread, but I'm not.

[Pete]

If the thread should become locked I'd prefer if I could post here a video recording on or before Sunday, hopefully this can happen Friday.

It's indeed the case that the computer would become very bored if the same tune were played 100 times over. I do have a 1 minute recording the computer has often played which I could repeat for something like a dozen times.

DDRM and RichardRussell, the RND function works perfectly for the requirements of the program. I think that it is very much as good as random, I think that it is random. Technically it is a pseudorandom result, but the result is random for the purpose of acheiving a random number.

The sequence of numbers received is not predictable under any normal circumstances. If you have the seed number then it is predictable if that stored lookup data is available to you.

I will have a video done soon with the same piece of music. I'm not in any rush to release on iphone or Android yet, but I'm very happy with the version for PC.

I've edited this post with details of how the algorithm works removed, I'll just post the final video soon and then the thread is probably complete, but it would be a shame if you "lock" it.

[DDRM]

Admin's bit: I'm happy for the thread to continue to discuss/solicit advice on programming issues, etc. What I suggested was futile was a continuation of the argument "It's conscious/no it's not/yes it is". Please try to avoid the latter. Posting a link to your video on Sunday is hereby judged acceptable, given the length of the argument to date, but subsequent videos allowable only if there is evidence of engagement/development on the code front.


Personal bit: Unfortunately you don't really seem interested in discussing code/algorithms, beyond getting people to provide you with specific bits of code, like a menu. That's a legitimate use of the forum, but there's been a lot of "honestly, if you listen to this, you'll be convinced", and a reluctance to engage with any of the constructive suggestions several posters have made about how you could demonstrate that your claims actually have any basis.

Several of us have listened to substantial sections of your posts, and hear no convincing evidence of any correlation. For example, on 9th of July you suggested listening to the stretch from 7' 55'' to 11' 55'', which I did, and heard nothing beyond "random" (chance) associations. Presumably that selection means you thought that was the BEST part of the recording, and the other 2/3 showed even less correlation.

I consider this to be in the "improving your program" category, so I dare to post it, without having to censure (or censor) myself, but if others disagree, let me know and I'll delete it!

How about this for two other tests you could try:

1) What output does your program produce when there is no music playing it can hear? If you record it, does it "harmonise" with a piece of music you were thinking about at the time, or perhaps listening to on earphones from a different device? If so, presumably the association is occurring in your head, rather than in the computer.

2) If you set the computer up to listen to several tunes (in a random order, say), and record the output from the program (separately), can you correctly identify which tunes were being played at the time? That should be amenable to simple statistical analysis. Obviously, you'd need to be absent at the time....

Both of those are tests you ought to be able to carry out yourself, though ideally you would get someone else to randomise the tunes etc, or, even better, to send you a set of tunes and a set of output recordings, and then interpret your results.

Best wishes,
D