Yea, this "Agential AI" BS is the latest scam. I think that if we'd just systematically start using "ML" instead of "AI", we could clear a lot of air and make all these discussions much more sensible. People are really thinking (because that's what's been sold to them) that there's "somebody there".
Many (speculative) things here. But this I firmly believe:
Machine learning (ML) can be a useful *aid/tool* in innovation/production of novelty.
But whether ML can automate (all of) scientific discovery - I doubt it. Why? Because ML does not have understanding & goals required for model selection.
Then we are in agreement. But I might go further and say that the current systems won't ever achieve this, because they are lacking understanding & goals - that is, agency. That's why we should speak in terms of "ML", not in terms of "AI". "AI" is a snake oil business con; we should drop using it.
Many (speculative) things here. But this I firmly believe:
Machine learning (ML) can be a useful *aid/tool* in innovation/production of novelty.
But whether ML can automate (all of) scientific discovery - I doubt it. Why? Because ML does not have understanding & goals required for model selection.
Tech & science need not be cognitively different, but they differ in context&aims. Model selection: past data + context&aims gives us ever better models of the solar system in terms of geocentric epicycles. The shift to heliocentricism was not a datamining issue; it required substantial reframing.
My p(doom) is pretty high. Not because I think there'll be an evil superintelligence that's going to decide it wants to wipe us out. It's because people are so stupid: we are going to make a lot of stupid things with this "AI" - like drown ourselves in slop - which will lead fucking everything up.
Tech innovations & science are different. You think Kuhn's The Structure of Scientific Revolutions is dross? But regardless of him, I do think think there is a potential theoretical gap: model selection is not a mechanical procedure. You can always conjure up more epicycles to make geocentrism work.
Well, somebody once said: "quantity has a quality of its own".. But the relationship of these two (seemingly contrary) notions is a profound conundrum of its own. In any case, there are different kinds on innovations. Science does include conceptual leaps too, something that LLMs might lack.
Mielestäni vakaumuksellisilla antikapitalisteilla pitäisi olla oikeus käyttää peliasuja, joissa ei ole mitään sponsorilogoja.
Enpäs tiennytkään, että tohtorit ovat automaattisesti tietokirjailijoita (eli väitelleet eivät kelpaa esikoistietokirjailijan apurahan saajiksi).
I reckon there are significant differences among LLMs. And differences in prompts might result in significant differences. However, I'd assume that some sort descent into drivel would eventually happen. Perhaps not meaningless. But pointless; some sort of mediocritization.
Niin, koko ajan on ilmeisempää, että Suuren ja Mahtavan Tekoälyn seurauksena on: #KnowledgeCollapse.
You mean the smell of dog shit - and we can only hope it's dogs - coming out of the frost? Delightful indeed.
Loistava vieraskynä, kiitos Kaarlo Hildén! Kirjoituksen pointti pätee korkeakoulutuksen lisäksi sellaisenaan myös myös tieteeseen:
Tieteen tavoitteen rajaaminen pelkkään kilpailukykyyn sivuuttaa inhimilliset haasteet ja sivistykselliset päämäärät – vaikka juuri niiden varaan kaikki muu rakentuu.
"Sivistys – kyky ajatella kriittisesti, toimia eettisesti ja ymmärtää ihmistä ja maailmaa – on sekä itseisarvo että taloudellisen menestyksen edellytys."
Simples.
Korkeakoulupoliittinen keskustelu on kyllä aika puupäistä. Lieköhän liian myöhäistä puhua järjen ääntä?
This is terrible. Surely there is something we can do?! Anyone? Bob Geldof?
To me, RATM is kind of in its own realm - truly progressive music (after the 1st album). Much of it is downright hip-hop. The other bands were very simple, easy, cheap in comparison, not true blend of genres. (I always liked the pejorative term "bounce metal".)
www.youtube.com/watch?v=DkkA...
Is RATM "nu metal"? Also, they did their stuff in the 90s. And: all of their stuff is suberb. (But you are right: the rest what people associate with this genre/era is shite.)
Oh, there are so many WWII things we could revisit: extermination camps & gas chambers, firebombing major cities into ashes, and, of course, the employment of nuclear weapons. The good 'ol days.
Oh no.. Powerslave is an album by Iron Maiden. It includes the song discussed here - Rime of the Ancient Mariner - and a song called "Powerslave" (the penultimate song before Rime). Also, while the poem is called "The Rime of the Ancient Mariner", the song is called "Rime of the Ancient Mariner".
But then they did this (which I find quite entertaining - and the audience clearly too):
www.youtube.com/watch?v=TLVb...
Had to have a quick look at what we've got. This is a great version:
www.youtube.com/watch?v=lma_...
Is that so? Don't know how many different versions I have heard. But the ones that I remember (like the Turn It On Again Tour version live) did sound pretty faitful to the original.
Scorchio!
Hah, I'm perfectly fine with trusting World of Genesis on this too. I find it just weird that it wasn't included on The Way We Walk.
Hah, I remember watching that on the telly. Can't say anything about the setlist though. But I was under the impression that it wasn't played on this tour (due to it being missing from The Way Walk Vol. 1, I suppose). But online setlists seem to include it as the last song, also for Knebworth:
Indeed it is.
Hmm.. I suppose Turn It On Again was performed on every tour, including the We Can't Dance Tour. But weirdly it's not included on The Way We Walk Vol I.
Here is a clue:
This feels like a helpful primer for understanding AI in general. It makes programmers more efficient on the front end because they can generate more code. But it makes them less efficient afterwards because it vastly increases the amount of time they spend tweaking and fixing bugs.
