AI Funding for Open-Source Artificial Intelligence

Posted 9 Nov 2008 at 16:01 UTC by mentifex Share This

Abstract: Many programmers of Open-Source software would like to dabble or specialize in artificial intelligence but are concerned about how to obtain AI Funding. The author of http://code.google.com/p/mind forth/ offers a few ideas.

Shoestring

The most primitive level of AI Funding is at the extremes of individual vis-a-vis teamwork; youth vis-a-vis maturity; and low net-worth vis-a-vis opulence. If you are a student or an IndependentScholar wanting to work on AI with minimal funding, there are ways to do it.

- Work on some DayJob to finance your AiProject.

- Use second-hand computer hardware to save money.

- Use OpenSource software to create OpenSource AI.


PreSolution

If you do not think that AiHasBeenSolved and you are eager to solve the problem of AI, then your opportunities for AI funding consist mainly of having the proverbial DayJob to support your efforts. Try to achieve some overlap between the DayJob and the AiProject. For instance, if your DayJob involves computers, you may learn things and gain access to things useful for your AiProject.


PostSolution

If you accept the assumption that AiHasBeenSolved with MindForth and its derivative implementations, then you already have a positive MindSet that may enable you to generate AI Funding for your own career at the dawn of AiEvolution. Consider the following options for generating CashFlow while you advance the StateOftheArt in AI.

  • You let it be known that, for a very small fee, you will install a customized AiMind in JavaScript on the WebSite or LapTop or DeskTop of whoever wants to have an AI with a unique KnowledgeBase -- that you will code for your grateful and appreciative client.

For example, if a friend of yours wants an AiMind to liven up a WebSite, you code the DoxoLogy of your friend right into the inerasable KnowledgeBase. Then you seed the MindModule of KbTraversal with particular concepts that will engender thoughts and ideas telling websurfers interesting tidbits about your illustrious friend. For instance, you pre-ordain that KbTraversal will activate either the TrueName or the NickName of your friend as a ConCept lodged in the vault of the innate KnowledgeBase. Then periodically, when things are otherwise quiet due to lack of InterAction between HumanUser and AiMind, the CustomAi will suddenly think outloud the name of your friend with either a statement of fact or a question of inquiry. People all over the world who have clicked on the AiMind link and now have your friend's CustomAi up on their screens, will be treated to a revelation of the wondrously special nature and all-around excellence of your friend (or corporate client). Subjects dear to the heart of your friend may also be concepts reactivated by your coding of KbTraversal. Suppose, for instance, that your friend or business client is especially interested in, nay, totally fascinated by the concept of extended warranties and service contracts. All you as an AiCoder have to do is prime the KbTraversal module to rotate through the conceptual activation and thought-generation of ideas about the concepts of "warranty" and "service". The AiMind will bring up each engineered topic, and users will have the opportunity to make statements or ask questions about each topic of thought. If you code "roses" into the KbTraversal as something to think about, people who love roses (all those people in Pasadena CA and Portland OR) may "hang out" at your friend's WebSite to talk about roses with the artificial AiMind. Along the way, money will flow your way to fund your noble efforts in AiEvolution.

  • For a teensy-weensy fee, you install an AiMind in somebody's RoBot.

Most RoBotics club members are more interested in HardWare than in SoftWare. They like hands-on experience with making things go. You, however, are a more contemplative type and you are more interested in the ultimate mysteries of mind and thought. You let it be known that you would love to install an AiMind in somebody's robot for a ridiculously small fee -- just enough to pay for your daily coffee and your subscription to ArtificialGeneralIntelligence Magazine. It is a match made in AiHeaven, as your friend or client plies you with cash to keep up the good work of adding just one more feature to the AI Synapture of the insipient RoBot.

In either of the above choices, building up an AiMind for a small fee can lead to a BigTime career in BigAi.


new branch of mathematics required, posted 10 Nov 2008 at 10:34 UTC by lkcl » (Master)

draw a 2x2 grid, and fill it out as follows:

  • top-left contains "classical newtownian mechanics"
  • bottom-left contains "chaos theory, singularities; percolation theory etc."
  • top-right contains "quantum mechanics"

  • ... what's in the bottom-right hand grid?

chaos theory and singularity mathematics depend on such rather awkward things as there being "a zeroth dimension"; division of zero by zero; division of infinity by infinity; locating the point at which a (deliberately positive-feedback-unstable) system flips between two different states.

quantum mechanics is (as best i can make out) based on snap-shotting mathematically-founded equations using a statistical sampling technique - obtaining a "rough and quick peek" at something that is inherently and otherwise impossible to guage [without that quick glance]. and, more to the point, this seems to be what nature actually _does_! if there's no observer, there's no interaction, and so there is an energy exchange which causes the "equation" to both change - and resolve.

all quite fascinating - and quite useless for all practical intents and purposes when it comes to writing a computer-based mechanism that will allow the expression of intelligence to be made _through_ it (i refuse to use the word "artificial" in association with the word "intelligence").

that fourth corner - where quantum mechanically-based systems are *deliberately* set up in "inner positive-feedback loops" that have "outer negative-feedback loops" on top of those - thus encapsulating both stability and instability - that fourth corner - where quantum mechanics meets singularities and critical instability junctions - is where all the modern mathematics breaks down.

i'll say that again, because it's a stunner.

there simply is *no* modern branch of mathematics which covers both singularity mathematics (critical-point fluctuations) and quantum mechanics.

this, in the 21st century, is a definite "wtf" moment.

what's the relevance to computer-based intelligence?

well, the issue is that a friend of mine has been studying intelligence - consciousness - and his understanding is that the human brain causes critical-point fluctuations, and that it is capable of detecting and using "phase changes" as a means to store information. so, it's not so much the "stable state" that's important, but the *change* from one state to another that encodes the all-important information.

(it should come as no surprise to the computer scientist that the number of _changes_ between states is far, far greater than the number of possible states).

this _has_ been recognised in some of the pseudo-intelligence (aka "AI") research fields.

not only that, but to make a self-aware - "conscious" system requires a feedback mechanism that can include "recognition of self". in other words, it gets horribly, horribly recursive. no wonder that guy lost his marbles in "zen and the art of motor cycle maintenance".

so - the million dollar question is: how on earth, in a digitally-based discrete numbering system (computers) can you _possibly_ produce self-awareness, when, with the best will in the world, it's impossible to implement quantum mechanics efficiently, let alone a howling-positive-feedback quantum mechanics system?

so my friend, dr alex hankey, is endeavouring to create the branch of mathematics that could allow us to understand - at least recognise, finally - the existence of consciousness as a valid branch of science. i hope he succeeds, because we really could use a theory of mathematics that helps us make sense of how to do computer-based intelligence and consciousness, properly and methodically from a sound first-principles mathematical and provable theory, rather than through the trial-and-error approach that's been deployed by computer scientists for the past forty years.

no offense intended, mentifex :) i like trial-and-error creation of machine-based intelligence as much as you do!

Freelancing science: doing science without tenure, posted 11 Nov 2008 at 12:32 UTC by chalst » (Master)

I think mentifex will find the following weblog posts by Pawel Szczesny of interest:

  1. Growing in open source business model, where he talks about getting grants while outside academia;
  2. Freelancing science - today and tomorrow, where he descibes his situation as a freelancer with an unpaid university affiliation; and maybe also
  3. Bug tracking systems in science.

Meds, posted 15 Nov 2008 at 05:18 UTC by ncm » (Master)

Furthermore, one must be sure to take one's meds on schedule.

tomorrow, posted 15 Nov 2008 at 22:48 UTC by lkcl » (Master)

ncm, i'm going to reply, tomorrow, not today, because it's beyond my 9:30pm bed-time and i like to sleep well, these days. if i see an apology on here, from you for your comment, it'll be a nice reply. i like nice replies. if you're very lucky, i'll sleep in. the sun might even be shining. i might go for a walk, watch a film. that gives you... 18 hours in which to contemplate how nice a place advogato is with the lack of disparaging remarks that have been made for quite some time.

Re: Heh, posted 18 Nov 2008 at 22:11 UTC by zanee » (Journeyer)

When are you guys ever gonna bury the hatchet? This has easily been going on for a year or two now..

rrrr :), posted 20 Nov 2008 at 18:28 UTC by lkcl » (Master)

*sigh* - zanee, it's a little more than that :)

ok.

to put this into perspective, i'm being contacted occasionally by people who say that they read advogato and have a lot of respect for the things that they can learn from it.

the issue i have with disparaging and critical remarks is therefore that it shows people who are learning about free software, typically just coming out of university, that free software developers are intolerant and prefer to make personal remarks rather than learn from their peers.

we are our own ambassadors and i'm pissed off that some of us cannot be bothered to apply filters to find the value in what another of us says.

apology, posted 20 Nov 2008 at 18:36 UTC by lkcl » (Master)

given that there isn't an apology from ncm, i'm making one on his behalf.

to anyone reading these words, in the future, i apologise for you having to read a personal disparaging remark made by ncm at mentifex's expense. free software isn't ... how to say ... isn't a group of well-organised people who are "controlled" or "restricted" in what they can and cannot say, by virtue, for example, an "employee".

if ncm was employed by the "free software corporation", there would be rules and regulations: he would have been disciplined under the corporation's guidelines if mentifex was also an employee, and if mentifex was a _customer_ then ncm would have likely been fired, for affecting the company's profitability.

as there is no such "corporation", no such "control" can be exerted over ncm.

so it's down to individuals to enact self-censorship.

we speak for ourselves.

Food for thought ..., posted 20 Nov 2008 at 22:07 UTC by DeepNorth » (Journeyer)

lkcl: I am not sure that ncm is quite as far out of line here as you suggest. It *is* unmannerly, but then arguably so was the thing to which it was responding.

http://www.advogato.org/article/928.html

http://www.nothingisreal.com/mentifex_faq.html#x1-20001

http://www.advogato.org/article/769.html

http://www.advogato.org/article/809.html

http://www.advogato.org/article/832.html

http://www.advogato.org/article/910.html

I support diverse opinions and I am not overly judgmental (I think) due to the source of things. However, I took a look at the various stuff in question and whatever it is, it sure is not baked yet.

I am no AI expert, but I do know a little about it. I also touch upon things that (by co-incidence) help to answer why the human brain is so GD complicated in my own work. I did not see any immediate treatment of such ideas here.

I am a fan of John Anderson's 'spreading activation' idea. Empirically, whether his mechanism is correct or not it has terrific explanatory power and seems to happen all the time. I am hazy, because it has been years since I read Anderson's stuff on this, but I recall he at least sketched out how such a thing could be implemented and may even have implemented something. Mentifex' reference to spreading activation made me look a little longer than I would usually, but at the end of the day, I very much do not see a 'solution' here.

I have mixed feelings about labeling people and then judging their work by the label you give them. A lot of eccentric (there goes a label) people have have come up with great work. Arguably, *only* people at least a little eccentric *ever* come up with great work. However, necessary does not imply sufficient, if you know what I mean.

On balance, I think the ideas here have likely been given a fair(ish?) hearing and this is like the cautionary tale of the boy who cried 'Wolf'. A wolf did indeed show up eventually in that story, but you could hardly blame people for not coming running that last time.

Extraordinary claims call for extraordinary proof and in this case the proof is hardly extraordinary. In fact, it does not appear to be demonstrable at all.

It takes a lot of time to verify even relatively simple stuff. Given the work required to completely review such a thing in its current state, I would be looking for something a *lot* more sophisticated (you know, GUI, speech, etc) that does plenty of tricks before I roll up my sleeves to take a really good look.

I would support anyone looking at this with an open mind, but caution them not to take on too much pain doing the analysis.

Re: the above, posted 21 Nov 2008 at 16:24 UTC by zanee » (Journeyer)

The idea isn't so much that the certs are retarded or what he is saying is simply made up stuff. This has been argued forever and to my knowledge there hasn't be an appropriate remedy. The issue at hand is the diatribe that normally follows such post.

Ignoring it when it's been fruitless is probably the best thing.

Certs, posted 22 Nov 2008 at 19:03 UTC by DeepNorth » (Journeyer)

I like the certs, but they sure go astray. I can't imagine how some of the 'masters' here would rise above some of the 'journeyers' in any other system.

In fairness, though, I expect that you should be certifying people if you expect to get certified back. I have been remiss in that regard, but in my defense, I have a dilemma. Many of the people that I would call 'Journeyman' are superior to most of the people who are 'Masters' on this system. It seems condescending to certify someone as 'Journeyer' and (I am sure I will get clobbered by this) few of the 'masters' are visibly such. They might be that, but there is little evidence of that. It just does not seem constructive to go searching for someone already certified as a master and certify them again. Certifying someone who is certified as a 'master' as a 'journeyman' or less seems unmannerly somehow.

I read with interest the complaint about certifications of some individuals. The point was certainly valid. I confess I find it occasionally irksome myself. I try not to dwell upon it, but it is a shame that a relatively good idea suffers these flaws.

I am chagrined to admit that I have never been to a 'key signing party'. I have my reasons. However, it might be a good idea to have the occasional 'Advogato Certification Party' or drive or whatever. It might help to rationalize the certification a bit. Unfortunately, I have a feeling that it might be too much of a popularity contest. Sigh.

Apology, posted 26 Jan 2009 at 15:11 UTC by ncm » (Master)

This is the first time I've been back here since November. I might have apologized to Luke for leading him to lose sleep, but I see that he is one of those who certified mentifex as a "Master", in direct violation of the certification guidelines.

RE: new branch of mathematics required, posted 19 Feb 2009 at 14:48 UTC by ta0kira » (Apprentice)

This was an interesting reply. Maybe it should be a new article. While I think a new form of math is appropriate to continue in the direction AI is attempting to go in, I have to question that direction. AI is such a broad category. If you're just talking about algorithms to intelligently perform data analysis, maybe you're right. If you're talking about AI to control robotic agents, I definitely think AI is headed in the wrong direction of NP-hell. What AI seems to miss is time sensitivity; not only that, but time/accuracy trade-offs. Who out there is working on algorithms that provide increasing decision accuracy over time, which can essentially be stopped in an instant for a "good enough" assessment? I've seen some work out there for determination of an "appropriate" decision, but they still fall into the solved/unsolved temporal dichotomy.

Kevin

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!

X
Share this page