September 27, 2004

Funny one [?|!]

I was "working" a path through our recent discussions about AIs, levels of information and co-evolution and I encountered and interesting landscape.

As we know them, supports of Artificial pre-intelligent hardware/software assemblies are made in such a way it is quite easy to dump them, clone them in fact.
Even for my laptop, a monthly full backup and daily partial ones, warranty that I could be able to resurrect it in the case of a major drawback. The most essential information is duplicated at a distant disk as soon as I connect to the Net.

AIs will benefit from this not only for immortality (as long as backups will be available) but also being able to mix individual experiences, the very same way smart synchronization is made between two directories/folders; copying elements from/to the two of them, in order to get two identical sets at the end of the process. Something like "living the same as the other one" in an accelerated rhythm. Accumulate experience, learn, by connecting and exchanging information. Ultra-fast and accurate studying.

If AIs already have an advantage over NIs, this might be the one that will make the difference. It tends to make the individual ubiquitous, at least for observation and data gathering, omniscient, as it can share methods of info treatment with specialized fellows, and as omnipotent as possible if it agglomerates the collaboration of others to achieve a goal.

Can you imagine the power of the process?
The first example that I though of is a scientific gathering where during the registration process, the evening before the opening of the meeting, everybody shares the entire data set (and methods of analysis, results, observations and conclusions) with everybody else, before starting any critical analyses.
Of course this is a ridiculous example, as continuous connection through a Net is the best solution, the same way we exchange over Internet. A permanent scientific gathering.

The schema is in use in scientific circles already, but very slow and somehow hindered by NIs egos and vanity, determining a degree of resistance when it comes to acknowledge the work of colleagues as superior of his own.

Would that be different for AIs? If yes, then we will be History as soon as AIs will appear.
On the other hand, we are already following this path, using tools to increase our capacity of info gathering and treatment. It might be that we slowly transform ourselves to AIs :-)
I am unable to calculate a Pearson coefficient by mind, but I do use more then a million of such calculations every week, seeking correlations. I always felt that this makes of me a kind of Borg. The same way you are one, reading on your screen what I wrote some time ago on one of Blogger's Hard-Drives.

7 Comments: (go down to newest )

  • Blogger kasimir-k ::

  • The dumping/cloning is something that NIs do too - kind of. Where a (pre)AI can clone all of it's contents, NIs "dump" more processed data: write books, posts on blogs, research reports etc. The ubiquitous AI-individual can be thought as parallel to humans as a specie: this too is immortal (unless it dies, of course :-) and can share experiences. So these features are shared by AIs and NIs - just accomplished quite differently :-)

    But before technology reaches autonomy it will still changes our experience sharing in interesting ways. Let's start with data storage: using personal storage devices is soon replaced by storing data over the network. Data will be accessible anywhere, and there will be no need for back-ups, as data is stored on three locations and verified on access. The storage of data will become fully transparent: we can just access it when and where ever we want. And of course not only our data, but data shared by others...

    That means of lot of stuff, and it's already tough to find some things on the web. Search engines are great help, but really interesting things get with the semantic web: "The Semantic Web is an extension of the current web in which information is given well-defined meaning, better enabling computers and people to work in cooperation."But it's not just sharing files, archived experiences. We already share in real time audible and visible experiences with mobile phones - with a head set display and earphones and a bit more bandwidth this will be pretty close to "being there".

    But back to AIs: it will indeed be interesting to see what kind of egos they will have, if any at all. What functions does an ego have for us, and how will AIs achieve the same?

    Somehow to me it feels accurate to say that we are transforming to AI (probably not many AIs, but just one). And when thinking coefficient calculations or Blogger's hard drives, it's amusing sometimes to go into details, how things really work, what systems are needed to accomplish this. Hard drives, the text there occupying far smaller area than is visible to bare eye. Well, then the processor, basically the most complex piece of engineering humans have come up with. The OS: millions of lines of code. TCP, IP, and the dozens of routers tossing the packets around. HTTP, HTML, CSS and scripting, and web browsers that can take in a bunch of bytes and put out a nicely laid out document.

    Just to read one web page requires incredibly complex system, and not only in one or two places: dozens of devices around the globe have to co-operate, each of them utilising many programs each of which has to work together with other parts of the chain.

    Basically my problem here is that I'm miserably unable to communicate how vast and complex system the web overall is, maybe because it is so much beyond what an individual can comprehend.

    Recently when I have been wanting some psychedelic reading I've gone to

    11:39 AM  


  • Blogger Rezlaj ::

  • Last night I was reading a short story by Isaac Asimov called Jokester. In Asimov's stories, computers are huge. The more power they have, the bigger the machine is, reaching the sizes of several football fields.

    In this story you have the computer, the Multivac as he calls it. Humans have been dumping information in it like crazy in the hopes that the computer will be able to process more variables and give us better answers than what we have today. But, as the computer got more and more information they reached a point where it was almost impossible to retrieve it. So they (the government I suppose) researched and found that the only human minds capable of formulating the right questions were the Chess Grand Masters. They have capacity of processing permutations and massive information to a point of becoming almost intuition.

    I'm not going to tell the whole story. But it's funny. Asimov wrote the story decades ago, before the internet and we are already going that way.

    Recall search engines from '94 and look at them now. They've gotten so complex already that it's not that easy to look for information anymore. And I don't mean silly information like product prices. I mean more hidden information about research that is not well-known for example.

    The Internet is a kind of living creature. I think it's the best candidate for AI we have ever created. The more information we add we will have to find better ways of going through it. Search engines are already talking about trying to interpret the meaning of the words and searching what you want, not what you write. That's a very advanced concept as far as computers go. Our mind is hardwired to give and extract meaning.

    An AI will need a way to store all the information and retrieve it. The Internet is about information and information retrieval. Imagine the Internet becoming concious some time in the future. I hope it's not evil! The problem is that I'm sure a conciousness of this type will see as as inferior beings for sure. It's up to the AI to have pity on us and spare our lives to mantain its hardware.

    12:04 PM  


  • Blogger Steve Jurvetson ::

  • Well, it sounds like these AI’s would not just need to aggregate and share input, they could replicate themselves as well (like Agent Smith in the Matrix). And then they might like to argue about who is the master copy. =)

    Football field size computers??? Come on.

    Think Big.

    Of course, you have to start with a moderate engineering project, disassembling Jupiter.

    9:18 PM  


  • Blogger Rezlaj ::

  • In that case it would be a collective conciousness I suppose. I still can't decide whether a machine would go for a collective conciousness like the Borg or individual conciousness like us.

    I can argue for and against both scenarios. On the one hand, I'd think that we would make the AIs similar to us. Afterall, us humans, like to commit hubris every chance we get and play God.

    But, being a programmer, it is hard for me to think of programs as independent. I don't know how many of you are programmers, but the way we make programs is opposite to individuality.

    Take the data normalization rules. They clearly try to prevent unnecessary data replication and promote efficiency in search and modeling. Usually, your data model is designed separately from the main application. This makes the data independent from the application, which is very good. You could have different applications accessing the same set of data.

    So, according to everything that I've been tought and that I've learned over the years, a collective conciousness is the one that I can imagine best. But that doesn't mean that the other option is not better. Maybe I'm too close to the problem. I would like for full blown AIs to be independent of each other, but wishfull thinking doesn't count.

    2:28 AM  


  • Blogger Popa ::

  • There is a comic book called Tom Strong, written by Alan Moore (if you never read Alan Moore, read it now! Get "V for Vendetta" to start). In one of the comics, the enemy was an human that created a self-replicating machine with his personality built-in. He spreads the schematics for one module on internet and waits for someone to build it. Once there is one module, this module creates another module and so on untill he occupies all earth or is completely destroyed.

    "I am the modular man. Everything that is will be me."

    6:05 PM  


  • Blogger kasimir-k ::

  • Steve, thanks again for a mind bending link! A full blown Matrioshka Brain eats my planet wide computers for breakfast! (Then again, my planet computer does not care about the second law of thermodynamics :-)

    But yes, back to the individuality of AI(s). I'd envisage an AI based on the internet, not constituting of just single computer, but a very large number of them, each being able to access data collected/produced by others. Even though each computer there could be to a certain degree autonomous, the processes of the AI are not tied to any single one computer. In this light it is difficult for me to imagine more than one AI.

    It's like if the theories and thoughts of scientists were not tied to any single person, but were processed by who ever who is available at the moment. This is not the case, because our communication capacity is limited - we can't communicate a massive theory to a colleague in an instant, so our thoughts are very much tied to us. This is not the case in a distributed AI: the data and processing is separate of the individual computers.

    But how about motivation? We research things because of curiosity. Many are also driven by the recognition by peers and other very humane goals. But will AI have curiosity? What will keep AI improving itself? Or do we have to tell it to expand?

    12:38 PM  


  • Blogger OldCola ::

  • Motivation could be expressed as a simple desequilibrium as long as an impossible goal isn't reached.
    This is the kind of things relatively easy to express in terms of algorithms.
    individuality is also simple to express. We do dispose of the immunity paradigm, at various levels, and it is quite easy to define, to encode.
    I suppose most of the effort should be turned to imagination. That's the difficult part and I suppose the only one quality I really recognize as intelligence.
    This isn't only about imagining new thing but also for symbolic representation ;-)
    See what I mean ?

    5:40 PM  


    Post a Comment

    << Home