I deeply hope that my brain doesn't give out before I can study all those //'s more closely.
n Nick Thompson [email protected] https://wordpress.clarku.edu/nthompson/ -----Original Message----- From: Friam <[email protected]> On Behalf Of u?l? ?>$ Sent: Tuesday, August 10, 2021 1:46 PM To: [email protected] Subject: Re: [FRIAM] Moral collapse and state failure Yes, if you squint. The trick is that both opaque and transparent ML algorithms are *engineered*. So, rather than intension/extension, a better frame would be the gen-phen map, forward-inverse map. The opaque boxes are the, probably irreversible, result of a complicated process. I think information is lost in that process. So even if you can somewhat reverse engineer how an opaque algorithm was built, it wouldn't be very accurate. A better approach would be to interview a bunch of ML people and catalog how *they* would have created such an opaque model. There's probably an analog of that in ethology. On 8/10/21 10:41 AM, [email protected] wrote: > I wonder if the interpretable/explainable distinction maps on to the > goal/function distinction which maps on to the phenomenon/epi-phenomenon > distinction which maps on the function spandrel distinction which maps on to > the intension/extension distinction which ..... > > Nick Thompson > [email protected] > https://wordpress.clarku.edu/nthompson/ > > -----Original Message----- > From: Friam <[email protected]> On Behalf Of u?l? ?>$ > Sent: Tuesday, August 10, 2021 1:23 PM > To: [email protected] > Subject: Re: [FRIAM] Moral collapse and state failure > > Yeah, it was long. I only got through half of it during my workout this > morning. > > I suppose it's right to say that the normative definition of moral > would exclude Trump (or people like him). But if we stuck to your idea > that a particular morality be *expressible*. (FWIW, I think the extra > qualifier "independently of oneself" is redundant, at least a little. > Any expression has to be at least somewhat objective ... spoken word > causes air vibrations, video recordings of someone talking, written > documents, etc.) > > So, there's a hot debate at the moment in machine learning about the > different usage patterns for interpretable ML vs explainable ML, whereas > "explainable" is weaker in that it doesn't give any direct access to the > mechanism, only describes it somewhat ... "simulates" it. Interpretable ML is > supposedly a kind of transparency so that you can see inside, have access to > the actual mechanism that executes when the algorithm makes a prediction. > > Targeting your idea that a moral code must be expressible, do you mean a > perfect, transparent expression of the mechanism a moral actor uses? Or do > you mean simulable ... such that we can build relatively high fidelity > *models* of the mechanism inside the actor? > > On 8/10/21 10:11 AM, Russ Abbott wrote: >> The Envy video looked like a lot of fun, but it was too long for me to sit >> through it. >> >> Regarding morality, my guess is that it's not predictability that leads >> people to consider someone moral, it's acting according to a framework that >> can be expressed independently of oneself. Society-wide utilitarianism would >> be fine; "someone much like Trump [who] says they're an exploitative, >> gaming, solipsist" and then behaves in a way consistent with that >> description, would not be considered moral no matter how consistently their >> behavior simply optimized short-term personal benefits. After all, to take >> your own Trump example, I doubt that many people would characterize Trump as >> moral. > > -- > ☤>$ uǝlƃ > > - .... . -..-. . -. -.. -..-. .. ... -..-. .... . .-. . > FRIAM Applied Complexity Group listserv Zoom Fridays 9:30a-12p Mtn > GMT-6 bit.ly/virtualfriam un/subscribe > http://redfish.com/mailman/listinfo/friam_redfish.com > FRIAM-COMIC http://friam-comic.blogspot.com/ > archives: http://friam.471366.n2.nabble.com/ > > > - .... . -..-. . -. -.. -..-. .. ... -..-. .... . .-. . > FRIAM Applied Complexity Group listserv Zoom Fridays 9:30a-12p Mtn > GMT-6 bit.ly/virtualfriam un/subscribe > http://redfish.com/mailman/listinfo/friam_redfish.com > FRIAM-COMIC http://friam-comic.blogspot.com/ > archives: http://friam.471366.n2.nabble.com/ > -- ☤>$ uǝlƃ - .... . -..-. . -. -.. -..-. .. ... -..-. .... . .-. . FRIAM Applied Complexity Group listserv Zoom Fridays 9:30a-12p Mtn GMT-6 bit.ly/virtualfriam un/subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: http://friam.471366.n2.nabble.com/ - .... . -..-. . -. -.. -..-. .. ... -..-. .... . .-. . FRIAM Applied Complexity Group listserv Zoom Fridays 9:30a-12p Mtn GMT-6 bit.ly/virtualfriam un/subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: http://friam.471366.n2.nabble.com/
