Linas, you lost me at Category theory... Nevertheless, I also find the idea of integrating symbolic system into neural networks amusing. The proof I really do is a recent speculation from a paper I'm writing from time to time, just to gather my thoughts (maybe I got this from you at some point, I really don't remember now):
Nevertheless, symbolic approach may support structure forms on top of which > artificial neural networks could operate, thus forming a synergy between > the two seemingly opposite philosophies in designing AI. > But then I develop things in symbolic direction because with this, I'm currently interested only in improvement of OpenCog URE engine, as far as I plan to offer contribution around here if the language I'm building passes the stages a, b, c, d, and also e, just in case. Realizing ideas take time, and life is too short to do it all, while I'm not a fan of bossing around... I also like to see creativity in other people too... sri, 27. ožu 2019. u 20:04 Linas Vepstas <[email protected]> napisao je: > Hi Sergei, > > On Wed, Mar 27, 2019 at 12:14 PM Sergei Kaunov <[email protected]> wrote: > >> Amused by your work, Linas, and you describe it very interesting. Where >> should we watch further progress on the topic? >> > > Thanks! Progress is hard, because several things have to happen in > parallel. > > -- I (or you, or anyone else) has to think hard across multiple different > difficult mathematical abstractions. > > -- Once I (or you, or anyone else) has some "great idea", it has to be > transmitted to others, either by email or by writing papers. > > -- Others can understand those "great ideas" only if they have the > appropriate mathematical background: in Ivan's case: hilbert-systems and > natural deduction and proof theory and type theory and category theory and > neural networks and deep learning and statistical mechanics (for example, > the "objective function" that neural net guys love to minimize/maximize is > exactly the same thing as a boltzmann distribution from statistical > echanics) So my personal progress on the topic is blocked by the > inability to communicate it to others. If you don't understand what I'm > talking about, its deeply frustrating for me. (And this is symmetric: > sometimes, I am told about great ideas, which I don't understand, because I > lack the background knowledge) > > -- A "great idea" is great only if it *actually works*. And, here, that > means (a) writing software (b) running experiments (c) analyzing data. (d) > describing experimental results to others. So, not only are steps a,b,c, > extremely time consuming, but step (d) is often mis-understood/neglected, > because the intended audience didn't understand why the experiment is > important. > > -- After you've conquered steps a,b,c,d then and only then can you do step > e) build an insanely great demo that will wow everyone who sees it, even if > they are a complete moron. For example, "deep fakes". You don't need math > to know that something unusual is happening there. > > The pressure I'm under, that I feel, is that I've got a collection of > "great ideas", I'm trying to articulate them, having trouble finding an > audience, struggling with steps a,b,c,d and meanwhile everyone is shouting > out loud "you guys are a bunch of losers because you don't have step e) you > suck!" and dealing with the psychological and financial fallout from that. > > I'm not unique, here -- most researchers/scientists deal with these same > issue. The commonly accepted solution for this is to create collaborations > and teams -- "division of labor" -- and tehre's chicken-and-egg problems to > solving that, also. > > --linas > > >> >> -- >> You received this message because you are subscribed to the Google Groups >> "opencog" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To post to this group, send email to [email protected]. >> Visit this group at https://groups.google.com/group/opencog. >> To view this discussion on the web visit >> https://groups.google.com/d/msgid/opencog/0b258284-cb0c-45ae-8d34-9f74d7ee00f9%40googlegroups.com >> . >> For more options, visit https://groups.google.com/d/optout. >> > > > -- > cassette tapes - analog TV - film cameras - you > > -- > You received this message because you are subscribed to the Google Groups > "opencog" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at https://groups.google.com/group/opencog. > To view this discussion on the web visit > https://groups.google.com/d/msgid/opencog/CAHrUA355QGo1zA8oCUZ3guEwMFymqJ7-PA5ydTJdMx2Sf%2B-ypQ%40mail.gmail.com > <https://groups.google.com/d/msgid/opencog/CAHrUA355QGo1zA8oCUZ3guEwMFymqJ7-PA5ydTJdMx2Sf%2B-ypQ%40mail.gmail.com?utm_medium=email&utm_source=footer> > . > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "opencog" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/opencog. To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAB5%3Dj6Xt0ze729QzdMVFyU4Q%2BYPB7hib_7n%3D4iNguzmJ2Em2SA%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
