Wikipedia:Reference desk/Archives/Computing/2020 April 21

Welcome to the Wikipedia Computing Reference Desk Archives
The page you are currently viewing is a transcluded archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

April 21

Looking for Examples of Creative, Non-Deterministic software

I'm going to be in a seminar with mostly liberal arts people where we will discuss issues around the question "Can machines think". I'm not advocating for "strong AI", even though I'm an AI person, I'm still very skeptical of what people call strong or general AI. I think we've made a lot of progress via Machine Learning but for the most part what ML has done is to open up new domains (e.g., problems where the main issue is separating signal from noise) and that solving what I think is the hardest problem of common sense reasoning is as difficult now as ever. I don't want to open up a discussion on that general topic because I know there will be all sorts of opinions. What I'm looking for is data on one specific issue: one thing I still hear from critics of AI is that computers are "completely deterministic" or "only output the knowledge that was input" or can "never be creative or discover new things".

My first example is always Samuels' Checker player (although really any Machine Learning program probably works) which could play Checkers much better than Samuels. Also, SHRDLU and other systems where you can give them a logical goal and they can backward chain to construct a plan (a sequence of formulas) which will eventually result in the goal being true. Also, the Postmodern Essay generator: http://www.elsewhere.org/pomo/ is fun. Also, I remember reading a long time ago that some early theorem provers had discovered new mathematical theorems which were of some interest. Anything like that (e.g., discovering new genes or molecules?) would be appreciated. Thanks in advance. --MadScientistX11 (talk) 02:31, 21 April 2020 (UTC)

You should not let the opportunity pass by to point out – if no one else already did – that Turing, in his paper Computing Machinery and Intelligence (pdf), after proposing to consider the question, ‘Can machines think?’ and then immediately proceeding to transmogrify that question into what came to be known as the Turing test, stated, "The original question, ‘Can machines think!’ I believe to be too meaningless to deserve discussion." Another interesting (IMO) point in the debate is the title of Nico Frijda's inaugural lecture, Can humans think? It sounds facetious but is actually a profound question. Are humans really essentially different from La Mettrie's l'Homme Machine?
    I think the arguments against the position that machines are inherently unable of thinking are mostly fallacious, but it is hard to argue the contrary position that "strong" AI in either sense – whether artificial general intelligence or the computational theory of mind – is possible on the basis of actual examples; unfortunately, much of what is being touted is hype. If someone writes a program that learns to do a task better than the programmer, it may say more about the ineptness of the programmer (at that task) than about machine ability. If you read Samuels' paper, you will see that the computer's play was determined by a vector of numerical weights, and optimizing the play was a traditional problem in optimization. But a table in the paper shows that these weights never converged; they kept oscillating wildly! So what, if anything, did the program learn? Occasionally a theorem prover finds a more elegant proof than previously known, as noted here regarding a proof by Łukasiewicz. When a computer rediscovered the proof of the Isosceles Triangle Theorem given by Proclus and ascribed by him to Pappus, it was reported at the time as demonstrating that computers can be creative.^[1] It was even presented as a novel proof. While (IMO) much simpler and more elegant than the textbook proof and Euclid's Pons Asinorum – it had, however, already been known for 1500 years. Also have a look at Algorithmic art for inspiration. Several artists (e.g. Dutch artist herman de vries) have made use of randomness in creating their works without involving computers.^[2]  --Lambiam 09:24, 21 April 2020 (UTC)

Great minds think alike! Yes, at the very beginning of the part of my presentation I do indeed have that quote from Turing right up front! I think it's funny how people ignore such an important statement (also that so many people, including many people in computer science don't know that the original Turing Test was actually what Turing called The Imitation Game). Also, I think we have very similar points of view. My feeling on Strong AI is more or less the same: we know so little about how the brain does Natural Intelligence and there are so many unsolved problems (especially the problem of common sense reasoning) that I think it is pointless to even speculate if/when Strong AI will be possible. I think a lot of people have been misled by recent progress with Machine Learning which has been incredible but which IMO still confines AI to specific domains (e.g., separating signal from noisy data) rather than the kind of general intelligence required for Strong AI. All I'm looking for though is a response to the people who say computers are never capable of doing more than what their programmers knew and programmed into them. I think there are examples where computers have made inferences that could be considered creative even though I wouldn't use those examples as evidence that Strong AI is possible. Anyway, thanks for those references, that should give me some good additional material. --MadScientistX11 (talk) 17:06, 21 April 2020 (UTC)

One issue in the discussion is that the signposts keep being moved. Fifty years ago no one would have believed that a machine would ever be able to beat a strong go player because it required a certain je ne sait quoi that machines couldn't possibly have. At the other side, I remember seeing repeated claims in the 70s and 80s that DENDRAL and its successor MYCIN had made valuable scientific discoveries, but I never saw a concrete exposition of what these supposed discoveries were, making the claims unverifiable. As to creativity in arts, there have been many experiments in computer-composed music, from the Illiac Suite to EMI. See also Category:Computer music compositions. For some of these compositions it cannot be denied that they are novel creations, so by what twisted logic can one then nevertheless deny that the programs were creative?  --Lambiam 21:13, 21 April 2020 (UTC)

• I am pretty sure Newton's flaming laser sword is part of the bibliography for that one. (At least the first third or half of that essay.)

I am not sure that "is strong AI possible?" is the question that people have in mind when they ask if computers can think or be creative. That is already putting the question in a "scientist" (positivist) framework: "strong AI" is defined as some set of observable properties, so the only way to answer the question either way is an impossibility theorem à la Gödel or a working prototype. "Can machines think?" might be more about defining "thinking" (and "strong IA" is not the only possible definition) than studying the properties of machines.

There are multiple examples of computer-assisted or full-computer generation of music, paintings etc. but those examples are IMO a bit superficial: we consider painting and music to be creative activities and therefore the (lack of) ability of computers in that domain should tell us something about computers in general. The problem is that it merely displaces the question to how said ability is judged and compared to some threshold of "creativity". For instance, if I have a program that can generate fake Rembrandts, the question displaces to what a "real Rembrandt" is. See for instance The Man in the High Castle (with a discussion of whether a gun used in a Civil War battle or a modern replica are distinguishable), or Baudolino (where the protagonists sell multiple heads of St John the Baptist to various cities and all of them experience miraculous cures, proving each of the heads is the true one), or most of Umberto Eco's works really.

Regarding "the output of a computer program is not more than its inputs", well, again you need to define some sort of thinking threshold. Does ClueBot NG "think"? It merely applies a deterministic algorithm to revert contributions that are statistically more likely to be vandalism. Yet nobody (not even the maintainers) could have predicted that they would catch this subtle act of vandalism. Tigraan^{Click here to contact me} 15:19, 28 April 2020 (UTC)

Can't View .PDF

I have a .PDF document on my C: drive and would like to view and print it by launching Adobe Acrobat to open and view the document. I am using Windows 10. In the past, I have been able to click on the document and open Acrobat and view and print the document. What happens is that Adobe Acrobat Reader DC shows up in the Task Manager as a background process, but I can't switch to it. Is there a reason that I can control why Acrobat has become a background process that I can't go to? Is there something I should do? Robert McClenon (talk) 06:17, 21 April 2020 (UTC)

If you are on Windows 10 you can try two things, the first one is to see if Adobe is the default app for PDF (settings => default apps => by file type)

If this is already Adobe but it is not working, the second option is to reinstall Adobe (settings => apps => Adobe acrobat reader => remove)

Then download the latest version (https://get.adobe.com/reader/) and install that (be careful of "bloatware" in the installer).

That should fix your problem, if it doesn't, more information is needed. Rmvandijk (talk) 08:52, 21 April 2020 (UTC)

Thank you, User:Rmvandijk. Acrobat was already installed, and was the default for .pdf. I un-installed it and re-installed it. Nothing has changed. The problem is the same. Should I call Adobe and ask for technical support? Will they provide technical support to a Reader customer, who hasn't paid them? Robert McClenon (talk) 00:40, 22 April 2020 (UTC)

There are possible workarounds. 1) Can you start Adobe Reader directly, e.g. on the Start menu or desktop? If it's already running then it may open the PDF file when you click it, or you can try opening it from the File menu. 2) If you right-click the PDF file and choose "Open with" then you may have other working options like a browser. Maybe it works for your purpose and you don't need Adobe Reader. If another program seems OK then you can set it as default. PrimeHunter (talk) 01:29, 22 April 2020 (UTC)

As User:Jc3s5h mentions below, there are two, I assumed you had Reader and per your comment, I believe I am right :P

A quick look shows that you're not the only one with this problem. Per User:PrimeHunter, try opening it with a browser to see if the file is correct. If it opens, it is probably not the .pdf. Some say that shutting down all adobe-processes in the task manager fixes it (until reboot), while I've also read that going to the Reader executable and setting it to compatibility mode with windows 7 (or 8) will allow you to open it again with Reader. This [[3]] or this [[4]] show two answers on the acrobat community websites. Hope that fixes it for you. Rmvandijk (talk) 06:32, 22 April 2020 (UTC)

The original post and subsequent comments refer to two different programs, Adobe Acrobat, and Adobe Reader. Adobe Acrobat requires paying for a license, and is a powerful program for creating, editing, viewing, and printing PDFs. Adobe Reader is a free program which is mainly used to view or print documents that were created by other software (although some relatively minor changes to a document are possible, such as filling in a form). The original poster should investigate which program he/she has and which he/she wants or needs. Jc3s5h (talk) 01:58, 22 April 2020 (UTC)

The OP mentioned Adobe Acrobat Reader DC, which is the latest version of Adobe Reader, and all that is needed for what they stated they wanted to do, viewing and printing. Regardless of all else, if clicking on a pdf document starts it up, it should then not run as a background process.  --Lambiam 06:20, 22 April 2020 (UTC)

Thank you, User:Lambiam, for having read my post clearly and explained it to User:Jc3s5h. Robert McClenon (talk) 20:18, 22 April 2020 (UTC)

User:PrimeHunter - I think I have it solved, although how I solved it is still a mystery. I created a Word document, for an unrelated reason, and was about to print it, and then decided that I would save it as a .PDF and print it. When I saved the Word document as .PDF, it opened in Adobe Acrobat Reader DC, which was in foreground, and now clicking another PDF opens it successfully. Maybe the install had been incomplete in some way and Word, on saving a document as .PDF, installed it correctly or completely. In any case, I think it is now working. Robert McClenon (talk) 20:18, 22 April 2020 (UTC)

Thank you. Robert McClenon (talk) 20:18, 22 April 2020 (UTC)

Archiving messages in 'What'sApp'

Where can one archive selected messages in "What'sApp', either on the computer, or on the smartphone ? Secondly, I'd like to know of a support center, or anything like that, for this application ? בנצי (talk) 08:09, 21 April 2020 (UTC)

Is there a preference for variables in bash script numeric calculations

I have seen the two different syntaxes used in bash scripts:

xxx=3
yyy="$((xxx+1))"

or

xxx=3
yyy="$((${xxx}+1))"

Is there a preference or recommended way of writing this? -- Q Chris (talk) 13:57, 21 April 2020 (UTC)

The ${} in the latter is simply superfluous. It isn't "wrong", but it's unneeded. Both expressions mean the same thing. Variable names are expanded automatically within arithmetic expressions. It's also only needed outside of arithmetic contexts either to delimit an ambiguous expression like ${foo}bar or to introduce parameter expansion. Some people appear to have read somewhere that you should always use ${foo}, or else are just cargo-culting from seeing it in existing programs and not understanding when to use it. Note the quotes in yyy="$((xxx+1))" are also unnecessary, as wordsplitting is not performed in assignments, though again they don't hurt. The much bigger problem with shell programming is people not using quotes when they should; this is the main example of people using them unnecessarily. --47.146.63.87 (talk) 19:20, 21 April 2020 (UTC)

Official documentation is also available for GNU Bash, and elaborates on the expression expansion syntax. Nimur (talk) 20:40, 21 April 2020 (UTC)

Hayward’s Double chess, computer programs, complexity

Are there computer programs available that play double chess? How complicated is it compared to Go?Rich (talk) 20:12, 21 April 2020 (UTC)

The average branching factor will be at least the square of that for standard chess (twice the number of pieces with more space to roam over), so it will be at least 1000, compared to the measly 250 of Go. (Aside: I think the geometric average, which should be lower than the arithmetic mean, is a better measure.) The average game length (number of turns when played out until checkmate or draw) will tend to be larger, but I guess less than twice that of standard chess, which is about 70. Let us estimate it at 120. For the game-tree complexity as measured by the 10-log of the tree size, I then estimate 120 × log 1000 ≈ 360, equal to the value 360 given for Go. Clearly, with slightly different estimates the comparison can come out on either side. Since a program (or human player) will only look a few moves ahead, double chess may be more complex in practical terms. On the other hand, it may be easier to assess a position in terms of strengths and weaknesses. 3D chess, anyone?  --Lambiam 06:11, 22 April 2020 (UTC)