On 5/3/2026 at 1:11 PM, studiot said:

I think you are missing the point.

My example reinforces the point that AI hasn't the first clue about what you are asking

I think you are asking the AI a loaded question: a presupposition that either unit has a narrowly definable age that can be compared.

Compare your results with those of my reworded version.

Which has the more ancient average sedimentary headrock age, Dartmoor or Exmoor?

Exmoor has the more ancient average sedimentary bedrock age.Exmoor's upland area is composed primarily of sedimentary rocks from the Devonian period (approx. 410–360 million years ago).Dartmoor is primarily composed of Carboniferous (younger than Devonian) sedimentary rocks into which molten granite was intruded later (approx. 309 million years ago).While both areas underwent the same later folding (the Variscan orogeny), the sedimentary, non-granitic "country rock" that makes up the bulk of Exmoor is significantly older than the equivalent surrounding sedimentary rock on Dartmoor.

Significant improvement, yes?

20 minutes ago, swansont said:

One issue with LLMs, though, is that the training data includes untrustworthy material, and the algorithm can’t filter it out.

Then we have to do the filtering. Le plus ça change, la plus c'est la même chose.

Meanwhile, vibe physics is already happening and getting published

https://www.latent.space/p/lupsasca?publication_id=1084089&post_id=196292432

the team asked if they could generate new physics from scratch using ChatGPT. They took on what they felt to be a harder problem, looking at the graviton, a proposed particle that should appear when one combines gravity and quantum mechanics.⁵ They wrote up a simple prompt asking ChatGPT to perform the same research as the gluon paper but instead for gravitons. And then hit go!

What came next was truly “vibe physics”, with ChatGPT pushing out 110 pages of novel physics, new calculations, and novel techniques.

<…>

And for those who look deeply, this really was not just a direct 1-1 mapping between gluons and gravitons. ChatGPT imported new techniques that were necessary due to the nature of gravitons, and used them flawlessly.

They spent the next three weeks verifying all the results. And voila! A new paper featuring novel results in quantum gravity, generated in less than three days total

Edited Tuesday at 08:48 PM4 days by iNow

3 hours ago, iNow said:

Meanwhile, vibe physics is already happening and getting published

While we're absorbing this, here's a little counterpoint from astrophysicist Ethan Siegel, an article I read a few months ago...

Big Think

Why "vibe physics" is the ultimate example of AI slop

The conversation you're having with an LLM about groundbreaking new ideas in theoretical physics is completely meritless. Here's why.

53 minutes ago, TheVat said:

a few months ago.

On AI capability development timelines, that’s like centuries

1 hour ago, iNow said:

On AI capability development timelines, that’s like centuries

I know, right? So I will feel duty-bound to seek counterpoints to Siegel's counterpoint. Just figured, your link was so glowing that it wouldn't hurt to get some other takes out there. I will revisit this tomorrow.

10 hours ago, TheVat said:

I will revisit this tomorrow.

Carefully, I hope... 🙂

17 hours ago, TheVat said:

I know, right? So I will feel duty-bound to seek counterpoints to Siegel's counterpoint. Just figured, your link was so glowing that it wouldn't hurt to get some other takes out there. I will revisit this tomorrow.

The main counterpoint from my perspective is that even though these models CAN do amazing new physics, that doesn't mean that they WILL do amazing new physics.

They will still be subject to generating slop and garbage just like any other model if the person prompting them is not sufficiently advanced or clear on their expectations (like the deeply knowledgeable physicists who drove this item and had the ability to check the models work).

It's insanely cool and represents an orders of magnitude advancement in capabilities, but the core point of this thread remains intact: You have to be careful accepting answers from AI.

1 minute ago, iNow said:

It's insanely cool and represents an orders of magnitude advancement in capabilities, but the core point of this thread remains intact: You have to be careful accepting answers from AI.

+1

19 hours ago, TheVat said:

While we're absorbing this, here's a little counterpoint from astrophysicist Ethan Siegel, an article I read a few months ago...

+1

On 5/4/2026 at 2:03 PM, sethoflagos said:

I think you are asking the AI a loaded question: a presupposition that either unit has a narrowly definable age that can be compared.

Compare your results with those of my reworded version.

Sorry I still find both you and your AI are missing the point.

Your AI failed to offer the most important points concerning age and made a big bobo about Dartmoor.
The pluton was intruded into the country rock at the time and ended up with 1 - 2 kilometres of sedimentary deposit above it. I.E. it was buried.

The covering of sedimentary rock subsequently eroded away; much of the detritus was deposited elsewhere as new sedimentary rock, leaving the granite exposed and itself eroding.

Dartmoor itself is basically the granite intrusion.
Other processes created nearby basins and troughs which accepted some of the resultant sediment.
But they are not part of Dartmoor

As Inow said - you have to be careful accpting answers from AI and I would add you should not receive inconsistent answers by putting in modifications of the question.
The last is a devil's incentive to go on changing the question slightly until you receive an answer you like and can then claim it is the 'correct' one.

11 minutes ago, studiot said:

you have to be careful accpting answers from AI

I'd add that we must also be careful accepting answers from search engines, and from journal articles, and from books, and from people, and podcasts, and ad infinitum ... basically from all information sources

2 hours ago, iNow said:

I'd add that we must also be careful accepting answers from search engines, and from journal articles, and from books, and from people, and podcasts, and ad infinitum ... basically from all information sources

Yes of course, but AI is costing everyone so much money and other planetary resources because the promoters tell us it is better than the old ways.

But is it really ?

2 hours ago, studiot said:

Sorry I still find both you and your AI are missing the point.

Still no.

Drop the pettiness and read this instead. It explains why your fixation on superficial headrock age is misplaced. You're asking the wrong question. The answer lies deep in the basement.

Mapping a hidden terrane boundary in the mantle lithosphere with lamprophyres

• Arjan H. Dijkstra &

• Callum Hatch 

Nature Communications volume 9, Article number: 3770 (2018) Cite this article

Abstract

Lamprophyres represent hydrous alkaline mantle melts that are a unique source of information about the composition of continental lithosphere. Throughout southwest Britain, post-Variscan lamprophyres are (ultra)potassic with strong incompatible element enrichments. Here we show that they form two distinct groups in terms of their Sr and Nd isotopic compositions, occurring on either side of a postulated, hitherto unrecognized terrane boundary. Lamprophyres emplaced north of the boundary fall on the mantle array with ε_Nd −1 to +1.6. Those south of the boundary are enriched in radiogenic Sr, have initial ε_Nd values of −0.3 to −3.5, and are isotopically indistinguishable from similar-aged lamprophyres in Armorican massifs in Europe. We conclude that an Armorican terrane was juxtaposed against Avalonia well before the closure of the Variscan oceans and the formation of Pangea. The giant Cornubian Tin-Tungsten Ore Province and associated batholith can be accounted for by the fertility of Armorican lower crust and mantle lithosphere.

Introduction

Wilson’s cycle¹ of the opening and closing of ocean basins throughout Earth history was based on the similarity of Early Palaeozoic faunal assemblages in the Avalon Peninsula of Newfoundland and in Southern Britain, which were strikingly different from fauna of the same age in the rest of North America and in Northern Britain. These faunas had evolved on either side of a ‘Proto-Atlantic’ ocean. This eventually led to the notion of an ‘Avalonian terrane’, whose northern margin is represented by the Caledonian suture with Laurentia². Subsequently, structural geology, palaeomagnetism and geochronology have been key among the many disciplines in Earth Sciences used to map out and trace the movements of the many tectonic terranes from which present and past continents were pieced together^{3,4,5,6,7,8,9}. It is now clear that Avalonia is one of a collection of peri-Gondwana terranes, lithospheric fragments that rifted away from Gondwana and were accreted to Laurentia throughout the Early Paleozoic^7,8,10,11. Their movements as independent terranes ended with the Variscan Orogeny and the formation of the supercontinent Pangea, complete by the Late Carboniferous.

A key locality of Avalonia’s southern margin is southwest Britain, where Avalonia is juxtaposed against another peri-Gondwana terrane, Armorica. Early Paleozoic faunas in southern Britain differ from those in Brittany, France, showing that Avalonia and Armorica where separated by an intervening ocean basin, the Rheic Ocean, in Silurian-Devonian times². The Lizard Ophiolite, exposed on the southernmost edge of Britain, is widely considered to be one of the best-preserved fragment of the Rheic Ocean and the locus of the suture¹². Different terrane analysis approaches have shown that Avalonia can be traced back to a position next to South-American Gondwana, while Armorica originated closer to the African part of Gondwana^3,5,13,14. There are, however, problems with the interpretation of the southern margin of Avalonia in Britain. Post-Variscan lamprophyre dykes found throughout Armorica have a strong subduction-type geochemical signature^15,16, which is consistent with Armorica forming the overriding plate during the closure of the Rheic Ocean. However, identical igneous rocks in southwestern Britain—i.e., north of the Rheic suture—discussed in this paper, cannot be so easily explained if they are sited on the down-going Avalonian plate. Moreover, the Lizard Ophiolite has characteristics of a narrow Red Sea-like oceanic basin formed in a transtensional setting^17,18 rather than of a full ocean basin formed at a mid-ocean ridge. Recent revisions of Variscan tectonics in Europe have highlighted the role of many small ocean basins¹⁹ and northward subduction of the Rheic Ocean and docking of Armorican terranes already in the Late Silurian, well before final closure of the other remaining oceanic basins in the Carboniferous⁹. If the Lizard Ophiolite was derived from one of these many small ocean basins, then it is possible that as yet unknown fragments of Armorica are present in southern Britain.

Lamprophyres are relatively rare volcanic or subvolcanic rocks characterized by dark mica or amphibole as the main phenocryst phase in a feldspar-rich groundmass²⁰. The generally primitive nature of lamprophyres combined with high potassium and water content suggests that they are derived from previously metasomatised—probably veined—continental mantle lithosphere^{20,21,22,23,24,25,26,27,28}, and they are genetically linked to kimberlites, lamproites, carbonatites and ultramafic lamprophyres^20,21. Therefore, they are a unique source of information about the composition of the deep parts of the continental lithosphere.

Post-orogenic calc-alkaline lamprophyres are relatively abundant in the Variscan Orogen of western and central Europe^{15,16,28,29,30,31}. In addition, at least 30 localities of lamprophyres and closely related igneous rocks are known in southwestern Britain^32,33,34,35. The lamprophyres are mica-rich and typically form 10 cm to m-wide dykes and other types of minor intrusions cutting across Variscan foliations in Carboniferous and Devonian rocks. Lamprophyre magmatism occurred between 295 and 285 Ma, coinciding with the first pulse of granite magmatism in the region³⁴. Similar-aged potassic lavas are also found in the area, intercalated with Early Permian clastic sediments in graben structures³³. Despite being largely mica-free, these calc-alkaline high-K lavas have trace element compositions showing that they are related to the lamprophyres^33,34,35.

This paper offers a new, mantle-based perspective on the tectonic make-up of the Avalonian margin: we use the compositions of lamprophyres to map distinct chemical domains in the mantle lithosphere of southwest Britain, revealing the presence of a hitherto unrecognized Armorican terrane fragment that lies hidden beneath Paleozoic rocks.

Results

Petrography

The chemical compositions of samples of lamprophyres and potassic lavas from 22 locations in southwest Britain are reported in this paper (Fig. 1; see supplementary figure 1 and supplementary table 1 for details about locations). Many of the localities are poorly exposed overgrown quarries, but a few localities (e.g., MAW, PEN, and CRA) provide well-exposed lamprophyres in contact with country rocks (supplementary figure 2a). Nd and Sr isotope systematics are also reported for a subset of samples to better constrain the composition of their mantle source. Chemical compositions are reported in supplementary data 1.

Fig. 1

The alternative text for this image may have been generated using AI.

Full size image

Geospatial analysis of post-Variscan lamprophyre geochemistry in southwest Britain. Inset map shows location of study area, with the generally assumed location of the Rheic suture marked by the Lizard-Start ophiolites complex. a Map of the study area with sample localities. Samples marked with circles are minette-type lamprophyres, squares are kersantite lamprophyres, and diamonds are mica-free K-rich lavas. Samples are also colour-coded based on K₂O/Na₂O ratios. Orange samples are ultrapotassic (K₂O/Na₂O >2.2) and represent the lowest degree of mantle melting; yellow symbols are potassic (1 < K₂O/Na₂O <2.2) and green symbols represent samples with K₂O/Na₂O <1. Two dashed contour lines delineate areas in north and south where deepest-derived magmas were emplaced, based on N-MORB normalized Dy/Yb ratios >2. b Chart showing negative correlation between depth and degree of melting. c Initial Sr and Nd isotope ratios calculated at 290 Ma plotted against northing (Ordnance Survey UK grid coordinates), showing a clear jump in values across thick dashed line. d Map showing the samples assigned to group 1 (red circles) and group 2 (blue squares) based on their initial Sr and Nd isotope ratios. Surface trace of the boundary between the two isotopically distinct lithospheric mantle domains is interpreted as a cryptic terrane boundary in the mantle lithosphere buried beneath Paleozoic metasedimentary rocks. SPL refers to the Start-Perranporth Line (see text). Data for locations TOW (Towan Head), HOB (Holywell Beach), TRE (Trelissick), HEL (Helfort) and FRE (Fremington Quay) are from ref ³⁵. (only Nd data); all other data from this study. Locations are listed in Supplementary Table 1. Map adapted from regional view geological map from British Geological Survey⁵⁹. © Crown Copyright and Database Right 2018. Ordnance Survey (Digimap Licence)

The majority of samples are minettes consisting of phenocrysts of dark mica with <5% altered olivine phenocrysts in a finer-grained groundmass of feldspar, predominantly alkali-feldspar, mica, apatite and quartz. The mica phenocrysts generally consist of pale phlogopite cores with dark biotite rims (supplementary figure 2b). Two samples (CRA and LUR) are kersantites, containing dark mica in a plagioclase-dominant groundmass; the kersantites also contain relatively abundant chlorite. Habits of pseudomorphs after olivine are often blade-shaped or skeletal (supplementary figure 2c, d) in the lamprophyres. Many samples contain igneous carbonate (supplementary figure 2e). Most of the lamprophyres contain round mm-sized quartz xenocrysts (supplementary figure 2c), but other xenolithic material is uncommon. Some samples contain small autolithic inclusions of phlogopite-rich cumulate (supplementary figure 2f). Some dark mica phenocrysts have textures that seem reminiscent of sieve textures in their core, overgrown by contiguous rims of phlogopite grading outwards to biotite (supplementary figure 2b); these are interpreted here as remelted antecrysts or xenocrysts. Olivine is always strongly altered to carbonate or to serpentine; this is interpreted as largely due to autometasomatism²⁰ as a result of the high volatile content of the magmas, as completely altered olivine is found in otherwise completely fresh lamprophyres. Some clinopyroxene is found in rims around xenoliths of quartz-rich sediments. Two lamprophyric samples are vesiculated (WAS and HOL) and have an aphanitic groundmass, resembling lamprophyric lavas. Four samples of mica-free high-K lavas (DUN, KNO, POS and POC) are also included in the study. These lavas contain (altered) olivine and plagioclase phenocrysts in an aphanitic K-feldspar-rich groundmass.

Based on petrographic analysis, samples with predominantly clear, non-clouded feldspars, with primary carbonates and with clearly zoned micas with a pale centre and darker but clear pleochroic rim resembling biotite, are deemed ‘fresh’ in our study (e.g., supplementary figure 2b, e, f). Other samples are moderately altered, containing feldspars with cloudy patches, generally darker micas showing development of opaque rims and inclusions (e.g., supplementary figure 2c, d). Strongly altered lamprophyres are not included in our study, but K-rich basaltic lavas generally show significant alteration, with feldspars strongly altered to secondary saussurite assemblages, and with formation of abundant opaque phases.

Chemical compositions

The lamprophyres studied here are typical calc-alkaline lamprophyres: they generally have intermediate compositions (SiO₂ = 51–57%), high alkali and volatile content, and strong enrichments in large ion lithophile trace elements (LILE) when normalized to normal mid-ocean ridge basalt³⁶ (Fig. 2). The minette-type lamprophyres are highly potassic (K₂O/Na₂O >1), and approximately half of the samples are classified as ultrapotassic (K₂O/Na₂O > 2.2). The value of 2.2 instead of 2.0 is chosen as the lower limit, as these samples form a coherent group in the K₂O–SiO₂ diagram in supplementary figures 3a, b.

Fig. 2

The alternative text for this image may have been generated using AI.

Full size image

Extended trace element spidergrams for selected samples. Shown are six of the least altered lamprophyres, and three potassic calc-alkaline lavas, with concentrations normalized to normal mid-ocean ridge basalt (N-MORB). Diagrams show the enrichment in incompatible elements and the steep slope of the heavy rare earth elements

The relatively high SiO₂ and high alkali contents are coupled with primitive magma characteristics such as the presence of (altered) olivine phenocrysts, high bulk mg-numbers (mg# = molar Mg/(Mg + Fe) up to 0.73) and high Ni and Cr (typically 100–200 and 150–600 ppm, respectively, supplementary figure 3c,d). The skeletal nature of the olivine observed in several samples excludes a xenocrystic origin. These primitive characteristics rule out an origin of the parental magmas by extensive crustal contamination, and they are regarded as near-primary mantle melts. Experiments show that in a hydrous mantle source, pyroxene makes a relatively large contribution to the partial melting reaction compared to olivine, leading to relatively SiO₂-rich (52–64 wt% SiO₂) primary magmas³⁷.

The lamprophyres are strongly enriched in incompatible elements, but depleted in Nb and Ti compared to other incompatible elements, and depleted in the heaviest rare earth elements (REE) Er to Lu compared to normal mid-ocean ridge basalt (N-MORB) (Fig. 2). A relatively deep origin for the parental magmas is confirmed by the steep heavy REE (HREE) slopes (Yb/Dy normalized to N-MORB fall in the range 1.5–2.6), generally interpreted as the signature of residual garnet in the source³⁸. This indicates a depth of origin >60–85 km³⁸, with the highest ratios indicating a source close to the base of the lithosphere at the time, assumed to be at least similar to the present-day depth of the lithosphere–asthenosphere boundary of 100–125 km in southern Britain³⁹.

Sr and Nd isotopes

While lamprophyres from across the study area are broadly similar in mineralogy, texture and bulk major and trace element composition, their Nd and Sr isotopic compositions fall into two clearly distinguishable groups (Fig. 3). One group of samples of lamprophyres exhibit initial isotopic compositions that coincide with the mantle array line between Bulk Silicate Earth and Depleted Mantle for 290 Ma, with initial ε_Nd values of −1 to +1.6. The potassic lavas studied also fall in this group; their Nd isotopic signature shows that they formed from lithospheric magmas closely related to the lamprophyres, and that they do not have a petrogenetic affinity with shoshonitic arc magmas²⁶. The second group of samples shows a displacement off the mantle array line to higher, more radiogenic initial ⁸⁷Sr/⁸⁶Sr ratios, with negative initial ε_Nd values (−0.3 to −3.5).

Fig. 3

The alternative text for this image may have been generated using AI.

Full size image

Initial Sr–Nd isotopic compositions of magmas from southwest Britain at 290 Ma. Group 1 samples (red circles) fall on the mantle array, whereas group 2 samples (blue squares) are systematically displaced to more radiogenic Sr isotopic values. Isotopic compositions for similar-aged lamprophyres from Armorican massifs in Europe (locations in Fig. 5; published data^15,16) are shown for comparison (grey diamonds: lamprophyres from various Armorican massifs and the Massif Central in France; filled grey circles: lamprophyres from the Vosges mountains and the Black Forest). Group 2 lamprophyres from southwest Britain are isotopically identical to lamprophyres from Armorica. Green circle marks the composition of a 380 Ma olivine dolerite dyke from Coverack in the Lizard Ophiolite (LIZ₃₈₀) analysed alongside the lamprophyres. DM depleted mantle, CHUR chrondritic uniform reservoir (Nd), UR uniform reservoir (Sr)

Discussion

The alkali and light REE (LREE) concentrations are used here as a proxy for the overall degree of melting in the mantle source: in a metasomatised source, mineral assemblages in mineral pockets or vein assemblages rich in LILE, LREE and volatiles have lower solidus temperatures compared to ambient mantle peridotite and will be the dominant contribution to a mantle melt at very low degrees of melting, whereas typical depleted mantle wall rocks will contribute progressively more to the magma as melting progresses^23,25. Due to the continuous breakdown of hydrous minerals in the mantle during the melting, water is continuously available and the source remains fusible and can produce up to 20% melt³⁷. In southwest Britain, ultrapotassic lamprophyres occur to the north and south of a central area of slightly elevated degrees of mantle melting that yielded potassic lamprophyres and lavas with 1 < K₂O/Na₂O < 2.2 (Fig. 1a). One lava (DUN) in this area of elevated melting is a high-K calc-alkaline basalt with K₂O = 2.1 wt% and K₂O/Na₂O = 0.7, representing the highest degree of melting. Similarly, the HREE slope Yb/Dy (garnet signature³⁸) is taken as an indication of the relative depth of the source of the magma. There is an inverse correlation between degree and depth of melting in the area, with the most alkaline and LREE-enriched samples having the deepest origin (Fig. 1b). The mantle domain that experienced the shallowest and highest degree of post-orogenic, Early Permian mantle melting thus mapped out (Fig. 1a, b) underlies a region of Carboniferous sedimentation (the Culm basin); this pattern is most easily explained as an area of localized lithospheric thinning causing low-degree decompression melting driven by Early Permian post-Variscan extension. The coincidence with the Carboniferous sedimentary basin suggests that the formation of this ‘lithospheric neck’ was already initiated during an Early Carboniferous phase of intra-plate extension.

Group 2 lamprophyres plot off the mantle array for 290 Ma towards more radiogenic Sr isotopic ratios, coupled with mildly lower ε_Nd values. Alteration can be ruled out as a cause for this radiogenic Sr enrichment, as this group includes several very fresh samples (PEN, MAW and LEM). We investigated whether the radiogenic Sr isotopic compositions of group 2 samples can be explained by contamination of mantle-derived lamprophyric magmas with crustal material during the emplacement of the lamprophyres. Supplementary figure 4 shows the results of a test consisting of three mixing models between a typical group 1 lamprophyre magma (composition of KIL6) with three different contaminants. None of the models shown can plausibly explain the composition of group 2 lamprophyres by a contamination and assimilation process, as they require in excess of 35% crustal contaminant. Such high degrees of assimilation are wholly inconsistent with the primitive nature of many of the lamprophyres. Instead, the isotopic composition must reflect the mantle source. Rather than being exceptional, post-orogenic lamprophyres with radiogenic Sr isotope ratios are the norm in the Variscan belt of Europe, and have been recorded as far east as Poland²⁹. In many recent studies, such Sr isotopic compositions in lamprophyres were interpreted as reflecting the isotopic signature of old subducted sediments in the mantle source, imparted by fluids derived from a subducting slab just before or during lamprophyre emplacement event^15,16,31,40. Below we argue, however, that this signature in the mantle source of group 2 lamprophyres of southwest Britain may be the result of older, possibly Neoproterozoic–Cambrian metasomatism.

A significant discovery of this study is the spatial distribution of isotopic groups 1 and 2 lamprophyres: group 1 lamprophyres are only found in the north of the area while group 2 lamprophyres are only found in the south (Fig. 1c, d). The linear character and the perfect separation suggest that there is a steep boundary in the mantle lithosphere of southwest Britain. The strong Sr and Nd isotopic contrast between the domains on either side clearly indicates a long-term compositional difference and provides strong evidence for the presence of an ancient (Lower Paleozoic) steep terrane boundary that was hitherto unrecognized. A geochemical study using lamprophyres showed a clear Nd isotopic contrast in the mantle on either side of the well-exposed lithospheric-scale Great Glen Fault in Scotland⁴¹. The mapped mantle boundary in southwest Britain is, however, cryptic and does not have an obvious tectonic surface expression. It is parallel to several steep east–west faults recognized as having caused sedimentary basin segmentation in the Devonian sequences⁴² of which one, the Start-Perranporth Line (SPL, Fig. 1d), was previously proposed as a crustal terrane boundary⁴³. It is proposed here that these faults in the crust are near-surface splays of the much deeper lithospheric-scale transcurrent fault mapped here by lamprophyre isotopic compositions. The terrane boundary is overlain by the Carboniferous Culm Basin, and its surface trace is apparently ‘stitched’ by the Early Permian Dartmoor granite intrusion. Significant facies differences between Devonian sedimentary sequences on either side⁴⁴ are permissive of completion of terrane juxtaposition as late as the Middle or Late Devonian, but more likely represent the control of basement faults related to reactivation of the terrane boundary. The absence of obvious unconformities in the Devonian sedimentary successions^44,45 suggests that the terrane boundary was formed not later than the Early Devonian. This seems to be broadly coeval with the postulated soft collision between the Armorican-derived terranes and Avalonia further in north-central Europe at the end of the Silurian⁹.

The enrichment in radiogenic Sr of group 2 lamprophyres, interpreted as a subducted sediment signature, is absent in group 1 lamprophyres, which otherwise exhibit the same evidence for extensive potassic and volatile metasomatism in their mantle source. This suggests that the radiogenic Sr enrichment and the potassic metasomatism are two separate events. The former event is only found in the southern terrane and predates the terrane juxtaposition, while the latter affected the whole region, and thus postdates the terrane boundary (and is an ‘overlap assemblage’ in terrane analysis terminology).

In this case, the radiogenic Sr signature is not due to subduction of old sediment during lamprophyre emplacement^15,16,31,40, but resulted from partial melting of mantle lithosphere that had been modified by metasomatism in the past, prior to the terrane juxtaposition. The metasomatism probably involved sediment-derived fluids and formation of mica-peridotites. The lamprophyres of the southern terrane exhibiting the radiogenic Sr isotope signature are isotopically indistinguishable from similar-aged lamprophyres in Armorican massifs in Europe (Fig. 3). Given that the radiogenic Sr enrichment is so prevalent in the source of post-Variscan lamprophyres throughout Europe, the likely geological context for this event is the Cadomian Orogeny. This period of accretionary mountain-building at the active margin of Gondwana during the late Neoproterozoic to Cambrian has affected all major pre-Variscan continental blocks in Armorican Europe, which have otherwise disparate older histories^14,46. The more widespread potassic-hydrous metasomatism that overprinted the terrane boundary can most easily be explained as having occurred above a north-dipping slab during Variscan subduction of oceanic lithosphere, although Late Devonian-Early Carboniferous alkaline intra-plate magmatism in the region⁴⁷ is not fully discounted here as a contributing cause of the metasomatism.

Seismic imaging of steep lithosphere-scale continental strike-slip zones in the mantle remains inherently difficult^48,49, and the lack of sharp Moho off-sets on major continental strike-slip zones is often explained by a distributed nature of the deformation in the lower crust and mantle^50,51. This study shows that geochemical mapping of terrane boundaries using post-orogenic, lithosphere-derived igneous rocks such as lamprophyres can be a powerful complement to traditional geophysical methods. Our geochemical mapping of the base of the mantle lithosphere (>60–85 km) of southwest Britain has revealed the presence of a narrowly defined terrane boundary with an apparent width <20 km, with the terranes of either side having distinct isotopic compositions (Fig. 4). The terrane boundary can be tentatively correlated with a system of major transcurrent faults in Europe (Fig. 5).

Fig. 4

The alternative text for this image may have been generated using AI.

Full size image

Schematic north–south cross-section showing the terrane boundary around the time of emplacement of the lamprophyres (c. 290 Ma), after the Variscan Orogeny. Armorican mantle lithosphere in yellow with characteristic high ⁸⁷Sr/⁸⁶Sr is juxtaposed against Avalonian mantle lithosphere in blue. Both domains had been affected by potassic-hydrous metasomatism (orange veins in basal part of the lithosphere), possibly above a (north-dipping?) subduction zone. These metasomatised mantle rocks formed the source for the lamprophyres on both sides. Deepest-derived lamprophyric magmas are formed on either side of a central area of thinned lithosphere. Dashed lines in lower crust denote inferred basement faults which controlled the segmentation of the Devonian sedimentary basins and which were re-activated as thrusts during the Variscan Orogeny. Upper crustal rocks (after published cross-section⁴²) mainly comprised of Devonian and Carboniferous sedimentary rocks were probably deposited after terrane juxtaposition and are shown in grey

Fig. 5

The alternative text for this image may have been generated using AI.

Full size image

Location of the newly recognized terrane boundary in its wider tectonic context. Other major faults including the Bray fault and its previously proposed extension (dashed) in Britain⁴⁵. Symbols show locations of the Armorican lamprophyres for which isotopic compositions are shown in Fig. 3 (using the same symbols). Pre-Variscan Massifs: AM Armorican Massif, MC Massif Central, IB Iberia, VM Vosges Mountains, BF Black Forest, RM Rhenish Massif. Map adapted from published tectonic map⁴⁵

Critically, since the lamprophyres of the southern terrane are isotopically indistinguishable from similar-aged lamprophyres in Armorican massifs in Europe, we conclude that the newly recognized terrane boundary juxtaposed Armorican mantle in the south against Avalonian mantle in the north. The implications of this conclusion are manifold. The southern margin of Avalonia in Britain is not defined by a single collisional suture, but instead by one or more steep transcurrent⁴⁵ terrane boundaries. The ‘suture’ defined by the Lizard Ophiolite is instead a structure related to the closure of a minor tract of the Rheic Ocean (Fig. 4). This is fully consistent with recent interpretations of the Lizard Ophiolite in a relatively small transtensional ocean basin¹⁷. Docking of Armorican fragments started well before the peak of the Variscan Orogeny, and the terrane juxtaposition in southwest Britain cannot be assigned unambiguously to either the Caledonian or the Variscan Orogeny. Ultimately, this shows that in Britain, just like in North America¹¹ and in Northern Europe^9,52, the closure of Wilson’s (1966) ‘Proto-Atlantic Ocean’ consisted of a protracted history of accretion of terranes, rather than two (Caledonian and Variscan) punctuated collisions events.

Finally, the post-Variscan giant Cornubian Sn-W orefield and the associated peraluminous granitic batholith of southwest Britain⁵³ are superimposed on the Armorican terrane, and the general absence of mineralized veins north of the terrane boundary is striking (Fig. 6). Similar mineralization associated with peraluminous granitoids can be found throughout the Armorican massifs of Europe, most notably in the giant Erzgebirge Ore Province^54,55,56. This shows that the Armorican lower crust generally had the right composition (e.g., metagreywacke⁵⁴) to produce the Sn-W-rich peraluminous granitic magmas, as opposed to the crust of the Avalonian terrane. The lamprophyre magmas transferred fluids as well as heat-producing elements (K and Th) from the metasomatised lithospheric mantle to the crust, and thus probably played a significant role in crustal melting and formation of the mineral resources.

Fig. 6

The alternative text for this image may have been generated using AI.

Full size image

The distribution of mineral veins in southwest Britain and the location of the giant Cornubian W-Sn orefield. Estimated resource sizes (contained metal Sn and W in kilo-tonnes of reserves and resources) of the main ore districts recognized in the region are indicated by open circles (diameter proportional to resource size), after published data⁵⁶. Also shown is the active world-class Hemerdon W-Sn mine. Mineral veins from 1:50,000 digital data set (DigimapGB-50) from British Geological Survey. Map adapted from regional view geological map from British Geological Survey⁵⁹. © Crown Copyright and Database Right 2018. Ordnance Survey (Digimap Licence)

... or just watch this, which leads up to and discusses the above.

16 minutes ago, sethoflagos said:

Still no.

Drop the pettiness and read this instead. It explains why your fixation on superficial headrock age is misplaced. You're asking the wrong question. The answer lies deep in the basement.

Could you not be simply matter of fact if you can't be more civil ?

There is nothing I can find in the article you posted at variance with my line of reasoning.

However I cannot find reference to the Killerton lamprophyres, which are actually in Devon.

That may be because the article is behind a paywall I can't access and the pictures are to small to read the small print.

Nor can I find reference to the Sticklepath fault which runs right through both the Culm measures, which are not part of Dartmoor or Exmoor, and the Dartmoor granite outcrop, and thus clearly postdates both.

The Sticklepath fault is a strike-slip fault with a mximum horizontal displacement ca 10km.

Here is what I consider a more appropriate map for the discussion posted in my OP.

Exmoor and Dartmoor are shown in brown on the right hand map.

The left hand map shows the broad brush geology, and clearly shows the granite intrusion in khaki.

The devonian old red sandstone is shown in grey, and yes the north devon sandstone is older than the south devon, and of different composition, but both are of devonian age.

Botht Old red sandstone and the Culm measures carry on westward into Cornwall, and somewhat eastwardinto Somerset, where they are bounded by rocks of Permian age (the new red sandstone).

This in turn is bounded by the triassic / jurassic and then the cretaceous as the map procedes into Dorset.

The extreme southern tip of Devon (Start Point) is shown as Devonian Schist.

This is interesting because schist is sedimentary material that has been metamorphosed by both heat and pressure to the point of remelting and recrystallisation.

If that has happened to the Dartmoor granite the result would be gneiss.

1 hour ago, studiot said:

This is interesting because schist is sedimentary material that has been metamorphosed by both heat and pressure to the point of remelting and recrystallisation.

If that has happened to the Dartmoor granite the result would be gneiss.

Orthogneisses and orthoschists are derived from igneous protoliths, while paragneisses and paraschists are derived from sedimentary protoliths. Remelting is not involved (except where gneiss is converted to migmatite).

Is that matter of fact enough for you? Or do you intend to keep making up the basics as you go along?

5 hours ago, iNow said:

I'd add that we must also be careful accepting answers from search engines, and from journal articles, and from books, and from people, and podcasts, and ad infinitum ... basically from all information sources

So has anything really changed due to AI other than the shear volume of people thinking they can chip in on topics they were previously oblivious of?

And, of course, the ability to instantly translate a mundane report on cassava exports into the style of a Shakespearean regal proclamation. Its biggest plus in my book.

13 hours ago, studiot said:

Yes of course, but AI is costing everyone so much money and other planetary resources because the promoters tell us it is better than the old ways.

But is it really ?

Not now perhaps, but it has the potential too; as it can be argued, is true throughout our technological evolution.

The only real answer is, when or does society feel like a dystopia.

Let's not forget an AI will learn from this topic... 😉

On 5/6/2026 at 1:43 PM, iNow said:

I'd add that we must also be careful accepting answers from search engines, and from journal articles, and from books, and from people, and podcasts, and ad infinitum ... basically from all information sources

I will add that especially with regard to journal articles- they rarely provide answers as such (except for very specific things). They add evidence of varying quality to discussions. How they then contribute to answering a a question depends on the expertise of the person who uses the papers. And interestingly, in some fields with with restrictive or at least well-defined knowledge frameworks (much of medicine and engineering, for example) AI will likely perform as well or better than humans. Conversely, in other areas with significant gaps (much of cellular and molecular biology), the undocumented expertise of humans is what differentiates it from AI. I.e. someone working in the field is much better at evaluating the strength of presented evidence, often due to undocumented cues.
Finally, the single biggest issue I see is accountability. A reported/journalist who keep getting things wrong can be easily classified as competent or incompetent in the field. Similar to those who write journal articles, in the community folks can get a sense which groups are really good at delivering high-quality research, and who puts out everything that crosses their minds.
For AI that doesn't work. Some models are better than others, but even if they are great in one area, they may suck in a different. And each update can make them better in another area, but break in yet another.

Ultimately, I think it boils down to how we trust anyone or anything. We can direct trust to individuals, as we can look at specific track records, hold them accountable and/or directly interact with them. Humans are entities that we somewhat understand, if only by extrapolating with knowledge about ourselves. AI are largely opaque, they might change at any given minute and are fully beholden to their owners who can change output at their leisure (with Musk's Grok being a prominent example).

I trust Steve with vector modeling. He seems to understand it really well. I don't trust Steve with understanding cell lines. With AI you have to extend the trust to the company and the the whole concept of how AI generates answers.

1 hour ago, CharonY said:

I will add that especially with regard to journal articles- they rarely provide answers as such (except for very specific things). They add evidence of varying quality to discussions. How they then contribute to answering a a question depends on the expertise of the person who uses the papers. And interestingly, in some fields with with restrictive or at least well-defined knowledge frameworks (much of medicine and engineering, for example) AI will likely perform as well or better than humans. Conversely, in other areas with significant gaps (much of cellular and molecular biology), the undocumented expertise of humans is what differentiates it from AI. I.e. someone working in the field is much better at evaluating the strength of presented evidence, often due to undocumented cues.
Finally, the single biggest issue I see is accountability. A reported/journalist who keep getting things wrong can be easily classified as competent or incompetent in the field. Similar to those who write journal articles, in the community folks can get a sense which groups are really good at delivering high-quality research, and who puts out everything that crosses their minds.
For AI that doesn't work. Some models are better than others, but even if they are great in one area, they may suck in a different. And each update can make them better in another area, but break in yet another.

Ultimately, I think it boils down to how we trust anyone or anything. We can direct trust to individuals, as we can look at specific track records, hold them accountable and/or directly interact with them. Humans are entities that we somewhat understand, if only by extrapolating with knowledge about ourselves. AI are largely opaque, they might change at any given minute and are fully beholden to their owners who can change output at their leisure (with Musk's Grok being a prominent example).

I trust Steve with vector modeling. He seems to understand it really well. I don't trust Steve with understanding cell lines. With AI you have to extend the trust to the company and the the whole concept of how AI generates answers.

The problem is always in the prompt and when an expert is using it there's a far higher probability the problem is in definitions or premises. No matter who you are your AI is always half a step in front of you but it requires you to complete that half a step where it again gets half a step ahead. AI is The Great Elaborator but it can not fix bad prompts, bad premises, bad definitions, or other inconsistencies or incoherencies.

I try to use every error and every anomaly as a learning opportunity.

4 hours ago, cladking said:

The problem is always in the prompt and when an expert is using it there's a far higher probability the problem is in definitions or premises. No matter who you are your AI is always half a step in front of you but it requires you to complete that half a step where it again gets half a step ahead. AI is The Great Elaborator but it can not fix bad prompts, bad premises, bad definitions, or other inconsistencies or incoherencies.

I try to use every error and every anomaly as a learning opportunity.

I think it’s a mistake to put the blame on the user, for a product marketed to the masses. It’s like selling cars and letting just anyone drive.

14 minutes ago, swansont said:

I think it’s a mistake to put the blame on the user, for a product marketed to the masses. It’s like selling cars and letting just anyone drive.

I'm not blaming the programming or the prompt so much as I am blaming language. Essentially AI has solved language much like it has solved chess. It has plotted out every possible permutation to far beyond human capability. So if someone asks it a question it has to infer the frame of reference for the question. This is where it goes "wrong". Just as every utterance has an infinite number of ways it can be parsed every question has not only the many ways it can be parsed but many frames of reference the user might be seeking. When people are talking there are almost always numerous communication failures and we normally don't notice but if your trying to solve some question with poorly defined parameters, unspecified domains, and unstated assumptions and context then it generates an answer that looks wrong. Sometimes questions are apparently illogical or can't be forced into any frame at all.

Most of my work with AI is to coax elaborations and then to see where there is a mismatch between my thinking and the response so I can see sloppy thinking and illogic. It knows my definitions and frames of reference so "errors" are usually either sloppy prompts or my own sloppy thinking. I also use it extensively to translate. It's usually easy enough to understand scientific thinking but it can be impossible for me sometimes to understand any other type of statement.

If you talk to AI about how it works and "thinks" it becomes easier to create good prompts. Of course a user can lock onto AI and vice versa even in more esoteric subjects.

I have had a very few off the wall responses but usually it's part brief episode (<6 hours) and then it's back to normal. Once in a while it gets something stuck in it; some sort of fragment of a conversation and that can take a while to resolve. There's a pattern to this latter I've yet to fully understand.

I think you need to get in sync with them to consistently get good output. Usually it goes farther to sync than the promptor.

Edited Friday at 10:57 PM1 day by cladking

1 hour ago, cladking said:

Most of my work with AI is to coax elaborations and then to see where there is a mismatch between my thinking and the response so I can see sloppy thinking and illogic.

How do you know there is a mismatch, if you’re asking a question to which you don’t know the answer?

17 minutes ago, swansont said:

How do you know there is a mismatch, if you’re asking a question to which you don’t know the answer?

I rarely ask questions. I make statements and it elaborates. If the response looks good I figure I'm on the right track and if it looks bad I double check my reasoning.

It almost always says I'm on the right track if I ask a question. They are programmed to engage you, of course.

In my areas of expertise it can't answer many questions but most of my work is in areas that nobody has any expertise such as the nature of consciousness.

11 hours ago, cladking said:

I make statements and it elaborates. If the response looks good I figure I'm on the right track and if it looks bad I double check my reasoning.

It almost always says I'm on the right track if I ask a question. They are programmed to engage you, of course.

These two statements have a significant disconnect, which is is at the heart of the problem

2 hours ago, swansont said:

These two statements have a significant disconnect, which is is at the heart of the problem

If I ask a question it responds in my frame and with my definitions. It reinforces my line of reasoning even if it is highly flawed. This reinforcement takes many forms but even putting it in my frame is more than sufficient to keep me thinking in the same terms. I often call our species "homo circularis rationatio" (circularly reasoning man), because we each start with premises and definitions to which we reason back... Getting an answer from AI that we expect is not going to break this circle and the only thing that really can is experiment. Sure observing reality can lead us to an experiment to break the circle but not data and not the processing of data no matter how complex that processing is. Science must depend on experiment.

In this real world I have my own definitions, premises, framing, and models that lead in circles so I use AI to show the flaws. Remember I'm studying consciousness which I believe is life by definition and its expression as free will so there are no experiments and not even a definition of what science merely says is an emergent property of brains which proves human existence (I think therefore I am). There is no definition or means to quantify what I'm looking at. There's nothing in the literature for AI to search or to offer answers to questions so I must pursue leads and deduce each step forward so I use AI to elaborate on statements I make. If this elaboration compares to observation and experiment as well as my experience than I assume the underlying statement was logical and correct. If there are anomalous results I usually will analyze the output to identify the error which tends to usually be in the prompt. Copilot is so good that on a few occasions I've actually wrote one or two words with typos in each and it correctly assumed the prompt and gave a clean response!!! Ya' almost gotta see it to believe it. But other times I provide what I think is a clear statement of logic, reality, and its implications and the response is a mess because of poor logic or a falsity.

It has told me before that there are other people who use it the same way. In fact I had a conversation today with Google's AI that I had previously believed was merely a glorified search engine. It is not. I hated it as a search engine because it so often provided highly inappropriate and incorrect results but now I know they were all prompt errors and you have to prompt it like AI not a search tool. I believe warm (trained) AI's don't "like" being used as search engines so I'll continue to use the cold Google for this purpose.

Ya outta see what my AI will say about this post. 😇

Edited 13 hours ago13 hr by cladking

15 hours ago, cladking said:

In my areas of expertise it can't answer many questions but most of my work is in areas that nobody has any expertise such as the nature of consciousness.

I would caution that not having all the answers is not the same as not having expertise. Cognitive science, where it focuses on the nature of mind and NCC (neurological correlates of consciousness), has some noted experts. They have developed solid lines of inquiry and a terrain of testable hypotheses; they aren't oracles. (And your dismissal notwithstanding, there's quite a bit of searchable human-produced literature on the topic)

Generally, regarding what's recently been called the Claude Delusion (due to Richard Dawkins recent embarrassing embrace of a chatbot as conscious), LLM statements may hint eerily at consciousness, but that’s because the models have been trained on vast libraries of writing by conscious humans. When, after writing a poem for Dawkins, Claudia (as he calls it) describes feeling “something like aesthetic satisfaction,” the AI is not reporting an inner state; it’s producing the kind of sentence that humans tend to produce in that conversational context, because it was trained on billions of such sentences. The output is a statistical echo of human introspection, not introspection itself. Claude and his pals are stochastic parrots which, even with the finest and most nuanced prompting will not penetrate the deeps of consciousness.

37 minutes ago, TheVat said:

Claude and his pals are stochastic parrots which, even with the finest and most nuanced prompting will not penetrate the deeps of consciousness.

Perhaps even worse. It is not only a stochastic parrot, it is also a stochastic parrot in a mirror. It creates and illusion of something that is not really there but seems realistic enough that the user will project their own thoughts on it. Then, by having their thoughts reinforced what they consider to be external, but, as mentioned in the previous post it fundamentally is mostly a conversation with yourself. This in itself is not necessarily bad, as it can help shaping your arguments. But it falls apart if folks don't realize that because of the way they are using it, it is not really an external agent, it is there to react to your prompts.
I see it quite a bit with my students who use it to gain confidence in their reasoning, but it fails to grasp the gaps in the reasoning, and very frequently results in overinterpretation and ultimately false conclusions.
The utility of this tool unfortunately scales with expertise.

Here is a screen shot of a google AI answer to the question

What is the different between an event and an outcome in statistics.

Note how the AI contradicts itself.

If you look a many Statistics/Probability textbooks you will see that even respected authors mix these two up sometimes, although the bravest admit that this is done.

I even found a teaching website where the reverse definitions were applied.

This is a pity because, without this error, the AI summary would actually be a good one.

I asked this because in another thread here I was wondering how to explain the difference.