Le coin de l'IA

barbabapat

12/01/2026 à 15h02

https://x.com/UnseenJapanSite/status/2008304365455524345

https://mustsharenews.com/grok-ai-swimsuit/

https://www.wired.com/story/grok-is-being-used-to-mock-and-strip-women-in-hijabs-and-sarees/

adhoc

12/01/2026 à 17h46

> Super même un sujet sur l'IA ça tourne en post de cus...<

Léz istoirs de cus de seespan :-))))

ℙ𝕖𝕣𝕧𝕖𝕣𝕤 ℙé𝕡è𝕣𝕖

12/01/2026 à 20h34

adhoc écrivait:
-----
> > Super même un sujet sur l'IA ça tourne en post de cus...<
>
> Léz istoirs de cus de seespan :-))))

Moi gème byen lez istoirs de cus

Prunelles

13/01/2026 à 15h24

Elon promet des robots chirurgiens dans 3ans. Max 4.
Comprenez qu'il va ouvrir les bons de réservations, que dans 10ans la situation n'aura pas évolué (sauf pour l'action Tesla...).
https://www.egora.fr/actus-pro/sante-numerique/avec-larrivee-des-robots-les-etudes-de-medecine-sont-desormais-inutiles

barbabapat

13/01/2026 à 16h19

Prunelles écrivait:
---------
> Elon promet des robots chirurgiens dans 3ans. Max 4.
> Comprenez qu'il va ouvrir les bons de réservations, que dans 10ans la situation
> n'aura pas évolué (sauf pour l'action Tesla...).
> https://www.egora.fr/actus-pro/sante-numerique/avec-larrivee-des-robots-les-etudes-de-medecine-sont-desormais-inutiles

Ca va être quelque chose le robot chirurgien qui discute avec le robot infirmière à la machine à café.

ELGY

13/01/2026 à 16h29

Bah si Elon l'a dit alors.....

J’imagine bien le Robot conspi barbaq.

vulpi

13/01/2026 à 16h49

Et il a prévu quoi pour les surnuméraires?
(cad ceux qui n'auront plus d'utilité et qui oseront réclamer de l'argent qu'ils n'auront pas gagné pour (sur)vivre... Vous n'y pensez pas à l'avenir des surnuméraires, on les laisse vivre ? On leur permet de se reproduire ? On fait un quota? Un zoo? Ce sont des questions qui viendront un jour)

barbabapat

13/01/2026 à 16h59

ELGY écrivait:
----
> Bah si Elon l'a dit alors.....
>
> J’imagine bien le Robot conspi barbaq.

https://brownstone.org/articles/when-physicians-are-replaced-with-a-protocol/


edit
https://automaton-media.com/en/news/ai-generated-manga-becomes-top-ranked-in-japans-biggest-e-book-store/

https://www.cmoa.jp/title/346887/

lemondestvaste

14/01/2026 à 12h03

j'ai bien aimé cette vidéo.

vulpi

14/01/2026 à 12h22

.

vulpi

14/01/2026 à 12h46

Erreur de manip.

barbabapat

14/01/2026 à 12h50

lemondestvaste écrivait:
--------------

> Et petite question : vous voyez comment l'avenir pour vos enfants ? on va vivre
> dans un monde profondément différent de celui que l'on connait ...

seespan

14/01/2026 à 14h02

Tiens il semble que la perception de l'efficacité de l'IA dans le codage soit un peut exagérée.

barbabapat

14/01/2026 à 16h24

seespan écrivait:
-------
>
> Tiens il semble que la perception de l'efficacité de l'IA dans le codage soit un
> peut exagérée.
>
>
C'est un outil qui apprend et qui apprend quand on le corrige. Le problème n'est pas l'état actuel, mais la façon dont il évolue par rapport à l'état actuel et sa capacité à intégrer les corrections qui lui sont proposées.

Au niveau de la musique, il était question de musiciens qui cherchaient au contraire à empoisonner les sources en produisant du bruit pour l'intégrer à la base de donnée des iA et chercher ainsi à les rendre moins performantes.

lemondestvaste

14/01/2026 à 18h31

Oui c'est une donnée qui m'a surpris, au bout d'un certains temps à force de se nourrir de contenu lui même créé par des ia, les créations s'appauvrissent. Je me demande dans combien de temps ça perdra en pertinence.

L'auteur de la video à un petit coté matrix, je ne sais pas si vraiment une super intelligence voudra se débarrasser du genre humain, ça promet ... quand il faudra affronter des robotos tueurs qui défendront les serveurs et sources d'énergies ...

vulpi

14/01/2026 à 20h02

On va avoir des usines à trolls ( ce n'est peut être pas le bon terme) qui auront pour but d'intoxiquer les IA avec des données erronées redondantes. Un peu comme certains contributeurs sur wikipédia.

vulpi écrivait:
-----
> Erreur de manip.


ça serait pas arrivé avec l'IA :-))))

vulpi

15/01/2026 à 13h30

colza2 écrivait:
------
> vulpi écrivait:
> -----
> > Erreur de manip.
>
>
> ça serait pas arrivé avec l'IA :-))))
😁 Peut être, mais j'arrive à faire des trucs improbables.
Perso j'utilise l'IA suisse, dont j'ai oublié le nom. Un gain de temps appréciable pour des recherches, mais ensuite il est important de bien contrôler ce qui est donné et le mettre à sa sauce.
C'est Public AI.

ℙ𝕖𝕣𝕧𝕖𝕣𝕤 ℙé𝕡è𝕣𝕖

15/01/2026 à 13h48

vulpi écrivait:
-----
...
> Perso j'utilise l'IA suisse, dont j'ai oublié le nom. Un gain de temps
> appréciable pour des recherches ...

De gonzesses ?

(pour fer plézir à Sipan ki adaure le cus)

vulpi

15/01/2026 à 14h57

Nan. C'est très basique. Temps de cuisson d'un poulet de 1,4 kg à la chaleur tournante. Synthèse sur l'alimentation en ultra -cyclisme avec liens. État de l'art pour l'obturation canalaire avec lien.
Gain de temps appréciable et ensuite je peux faire mes recherches perso.

seespan

15/01/2026 à 23h45

Pour les recherches la meilleur à l'heure actuelle ( à ma connaissance ) c'est perplexity.
Pas d'hallucination et ça te source toutes les réponses.
Après comme tout les grosses AI ( Gemine, open AI , grok, Claude ) je m'attend à un crash. Le prix de ventes est à perte et pas qu'un peut. Plus tu as tout les procès pour droit d'auteur qui sont en train de murir .

Rercherche perplexity sur la précision des scan intra et extra. 163 citation
I'll now compile a comprehensive research report comparing the precision of intraoral and extraoral scanners for dental use.

Precision Comparison of Intraoral and Extraoral Scanners for Dental Applications
The accuracy of digital impression systems has become a critical consideration in modern dentistry, with both intraoral scanners (IOS) and extraoral scanners (also known as laboratory or desktop scanners) serving distinct but complementary roles. Accuracy comprises two essential components: trueness (closeness to actual dimensions) and precision (repeatability across multiple scans). This comprehensive analysis examines current evidence comparing these scanning modalities across diverse clinical applications.

Fundamental Performance Characteristics
Intraoral Scanner Precision

Contemporary intraoral scanners demonstrate clinically acceptable precision across varied applications, though performance metrics differ substantially among manufacturers and clinical scenarios. Recent systematic reviews indicate that leading IOS systems achieve precision values ranging from 10–60 μm for full-arch scans. The 3Shape TRIOS 3 and TRIOS 5 scanners consistently demonstrate superior precision, with values of approximately 37.8 ± 4.53 μm for full-arch implant impressions. The Medit i700 exhibits comparable performance with precision of 40.6 ± 4.17 μm, while the Dentsply Sirona Primescan achieves precision values of 18.7 ± 5.6 μm to 49.1 ± 8.31 μm depending on scan complexity.
​

For single-tooth and short-span preparations, intraoral scanners demonstrate enhanced precision. Studies report values as low as 12.35 μm for the Medit i700 and 15–25 μm for TRIOS and True Definition systems in quadrant scanning. These measurements fall well within the clinically acceptable threshold of 30–75 μm established for prosthodontic applications.
​

Extraoral Scanner Precision

Extraoral scanners generally exhibit superior precision compared to intraoral systems, particularly for full-arch digitization. Laboratory scanners achieve precision values ranging from 4.33–37 μm across different models. The DOF scanner demonstrated exceptional precision with mean standard deviation of 4.33 μm, while the D810 achieved 6.56 μm. The 3Shape E3 extraoral scanner consistently produces precision values of 35.34 μm, outperforming most intraoral systems in laboratory conditions.
​

Comparative studies reveal that extraoral scanners maintain higher precision across extended scanning ranges. When scanning full dental arches, extraoral systems demonstrate precision measurements of 12,026 (SC scanner) and 7,687 (TS scanner) compared to intraoral values of 749–4,910. This advantage becomes particularly pronounced in cross-arch measurements where extraoral scanners show significantly superior precision (p = 0.018).
​

Trueness Performance Across Clinical Applications
Complete Arch Digitization

For full-arch impressions, both scanner types achieve clinically acceptable trueness, though with notable differences. Intraoral scanners produce trueness values of 24.5–95 μm for complete dentate arches. The TRIOS 5 demonstrates trueness of 54.9 ± 11 μm, Medit i700 achieves 60.5 ± 10.9 μm, and Primescan ranges from 30.5–72.3 μm. Studies consistently show that trueness deteriorates as scan distance increases, with diagonal cross-arch measurements exhibiting reduced accuracy compared to same-quadrant measurements.
​

Extraoral scanners demonstrate trueness ranging from 7.7–67.3 μm for complete arch scanning. The 3Shape E3 achieves exceptional trueness of 35.37 μm, significantly outperforming intraoral alternatives. When comparing identical specimens, extraoral scanning of impressions yields trueness values of 19.5–37 μm versus 20.7–33.35 μm for direct intraoral scanning. However, conventional polyvinyl siloxane impressions digitized by extraoral scanners show trueness of 24.5 ± 3.2 μm, superior to all tested IOS systems.
​

Implant Impression Accuracy

Full-arch implant scanning represents a particularly demanding application where scanner precision becomes critical. For six-implant edentulous cases, TRIOS 5 achieves the highest precision (37.8 ± 4.53 μm) and trueness (54.9 ± 11 μm) among intraoral scanners. Medit i700 follows closely with precision of 40.6 ± 4.17 μm and trueness of 60.5 ± 10.9 μm. Comparative analysis shows that intraoral scanners perform adequately for ≤4 implants, but conventional methods or photogrammetry demonstrate superior accuracy when >4 implants are present.
​

Extraoral scanners excel in implant applications, particularly when digitizing conventional impressions. Studies report trueness of 57.75 ± 65.17 μm for extraoral scanning versus 121.53 ± 89.55 μm for direct intraoral scanning of implant models (p = 0.001). For multi-implant cases, extraoral scanning of splinted impressions produces mean deviations of 66.05–78.58 μm compared to 25.23–28.15 μm for photogrammetric systems. The Ceramill Map400+ extraoral scanner demonstrates particularly strong performance with deviations of 32.4–93.7 μm across all implant positions.
​

Partial Arch and Single-Unit Preparations

For limited-span applications, intraoral scanners demonstrate competitive or superior performance compared to extraoral methods. When scanning three-unit fixed partial denture preparations, intraoral scanners achieve trueness of 28.09–74.50 μm and precision of 35.50–66.30 μm. Studies comparing IOS with extraoral scanning of conventional impressions show no significant differences for adjacent implants (p = 0.146), though extraoral methods perform better for distances spanning four or more implants.
​

Single-tooth preparations reveal intraoral scanner strengths, with Primescan demonstrating significantly superior margin accuracy compared to competitors in localized scanning. The mean deviations for intraoral scanners measure 19.6 μm mesiodistally and 16.4 μm buccolingually, while extraoral scanners show 24.0 μm and 22.5 μm respectively for similar measurements. This suggests that direct intraoral capture eliminates intermediate steps that may introduce distortion.
​

Comparative Analysis by Measurement Distance
Same-Quadrant Measurements

Statistical analysis reveals no significant differences in precision between intraoral and extraoral scanners for same-quadrant measurements (p = 0.47). Both modalities achieve clinically acceptable values, with intraoral systems ranging from 749–4,910 and extraoral from 2,107–12,026 in precision measurement indices. For anterior-to-posterior distances within one quadrant, precision values remain within 20–50 μm for both scanner types.
​

Cross-Arch Measurements

Extraoral scanners demonstrate statistically significant advantages for cross-arch measurements (p = 0.018). The SC extraoral scanner achieves precision measurements of 29,196 ± 8,467 for cross-arch distances, compared to 453–3,919 for intraoral systems. This disparity reflects fundamental differences in scanning mechanisms: extraoral scanners employ multi-axis motion and multiple cameras that compensate for cumulative positioning errors inherent in image-stitching algorithms used by most intraoral systems.
​

Studies examining maxillary full-arch scans confirm that trueness deteriorates with increased scan distance for IOS. The absolute mean trueness and precision values for all intraoral scanners measure 76.6 ± 79.3 μm and 56.6 ± 52.4 μm respectively, with greater imprecision occurring in molar-to-contralateral-molar measurements compared to same-quadrant distances.
​

Technological Factors Influencing Precision
Scanning Technology and Algorithms

Scanner precision correlates strongly with acquisition technology rather than surface topography characteristics. Intraoral scanners employing confocal microscopy (TRIOS series), structured light with AI-assisted library matching, and high-frequency contrast analysis (Primescan) demonstrate superior precision compared to earlier-generation systems. The Medit i700's AWS module with single-camera off-axis aperture design achieves exceptional precision through circular target point tracking that minimizes multi-camera alignment errors.
​

Extraoral scanners utilizing blue LED technology demonstrate better precision (21.7 μm trueness, 19.2 μm precision) compared to white light (32.9 μm, 24.4 μm) and laser-based systems (28.2 μm, 29.2 μm). Structured light extraoral scanners provide rapid scanning but may exhibit reduced repeatability in narrow or deep areas, while LED systems offer superior repeatability due to shorter wavelengths.
​

Environmental and Operator Factors

Ambient lighting significantly influences intraoral scanner precision. Studies demonstrate that TRIOS 3 achieves optimal accuracy at 1,000 lux (room light only) with median deviation of 26.33 μm, while 5,000 lux illumination increases deviation to 46.38 μm (p < 0.001). Temperature, humidity, and intraoral conditions (saliva, bleeding, patient movement) further impact IOS accuracy, variables largely absent in controlled extraoral scanning environments.
​

Scanning distance affects precision differently across systems. Research shows that 3Disc scanners perform optimally at 0 mm, TRIOS at 2.5 mm, and Helios at 5 mm scanning distances. Calibration status has minimal impact on modern IOS precision, with studies showing no significant differences before and after 324-day calibration intervals (p > 0.05).
​

Clinical Application Considerations
Edentulous Arch Scanning

Edentulous arch digitization presents unique challenges where scanner selection significantly impacts outcomes. For completely edentulous maxillary cases, indirect digitalization using extraoral scanners (D8I) demonstrates superior trueness for complete surface (7.95 μm), palate (9.11 μm), and border areas (20.22 μm). However, for alveolar ridge areas, intraoral scanners (Primescan: 16.45 μm, TRIOS 4: 8.96 μm) achieve competitive trueness.
​

Desktop scanners show the highest precision in digitizing completely edentulous cases, with RMS deviations significantly lower than TRIOS or Medit IOS systems across all anatomical regions. Peripheral borders, inner seal areas, and poorly traceable soft tissues (soft palate) exhibit maximum discrepancies with IOS, suggesting extraoral scanning of conventional impressions remains preferable for complete denture fabrication.
​

Material and Surface Considerations

Scanner performance varies with scanned substrate characteristics. Primescan ranks first in 11 of 15 substrate categories when tested across natural tooth surfaces, restorative materials, and dental substrates. Color and spectral properties of impression materials affect extraoral scan accuracy, with blue impression materials showing superior digitization under blue-light scanners compared to yellow or pink materials.
​

Surface texture and geometry influence precision metrics. Curved surfaces and groove areas demonstrate poorer trueness and precision for both scanner types. Sharp edges on prepared teeth improve extraoral scanner accuracy (trueness: 27.5–33.3 μm, precision: 22.7–31.7 μm), while smooth natural contours may reduce measurement reliability.
​

Systematic Review and Meta-Analysis Evidence
Quantitative Synthesis

Meta-analytical findings from systematic reviews indicate no significant overall difference in accuracy between digital (IOS and photogrammetry) and conventional impressions (SMD: -0.03, 95% CI: -0.20 to 0.14, p = 0.74). However, substantial heterogeneity exists (I² = 68%), reflecting variability in study design, scanner models, and measurement protocols.
​

Subgroup analysis reveals that for cases with ≤4 implants, no significant accuracy difference exists between digital and conventional methods. Conversely, conventional impressions show slight superiority when >4 implants are present (SMD: 0.22, 95% CI: 0.05 to 0.39, p = 0.01). Photogrammetry demonstrates marginally greater accuracy than conventional impressions (SMD = 0.15, 95% CI: 0.02 to 0.28), while standard intraoral scanners show comparable performance.
​

Qualitative Evidence Synthesis

Umbrella reviews encompassing 10 systematic reviews published 2020–2024 consistently identify TRIOS 3 (3Shape) and Primescan (Dentsply Sirona) as demonstrating highest complete-arch accuracy. Among 30+ IOS models evaluated, trueness and precision emerge as the most frequently assessed outcomes, measured via RMS deviation or mean absolute deviation.
​

Patient-reported outcomes favor intraoral scanning universally, with reduced procedural time averaging 8.59 minutes (95% CI: 6.78–10.40) compared to conventional impressions. However, accuracy limitations persist in partially edentulous and fully edentulous arches, where extraoral scanning of impressions or photogrammetry may provide superior results.
​

Practical Clinical Thresholds
The clinically acceptable accuracy threshold for prosthodontic applications ranges from 30–150 μm depending on restoration type. Most modern intraoral scanners achieve mean full-arch deviations <100 μm, with all tested systems in recent studies meeting basic clinical requirements. However, precision demands vary by application:
​

Single crowns: 10–30 μm tolerance
​

Multi-unit fixed prostheses: 30–75 μm tolerance
​

Full-arch restorations: 75–150 μm tolerance
​

Implant-supported prostheses: 30–150 μm passive fit tolerance
​

Extraoral scanners more consistently achieve the lower end of these ranges (4–50 μm), while intraoral scanners typically perform in the mid-range (20–100 μm) with greater variability based on scan complexity and distance.
​

Comparative Scanner Rankings
Top-Performing Intraoral Scanners (by precision)

Medit i700: 12.35–40.6 μm precision
​

TRIOS 5: 37.8 ± 4.53 μm precision
​

Primescan: 18.7–49.1 μm precision
​

TRIOS 3/4: 20–70 μm precision
​

CS3600: Variable performance, 45–80 μm precision
​

Top-Performing Extraoral Scanners (by precision)

DOF: 4.33 μm mean standard deviation
​

D810: 6.56 μm mean standard deviation
​

3Shape E3: 35.34 μm precision
​

NeWay: 15.7 μm precision
​

InEos X5: 21.62 μm precision
​

Summary of Key Findings
Precision Comparison

Extraoral scanners demonstrate superior precision compared to intraoral systems, particularly for full-arch and cross-arch measurements. Laboratory scanners achieve precision values of 4–37 μm versus 10–60 μm for intraoral scanners across complete arch applications. This advantage stems from multi-axis motion, controlled scanning environments, and elimination of intraoral variables (saliva, movement, limited access).
​

For same-quadrant and single-tooth applications, precision differences become statistically insignificant, with both modalities achieving clinically acceptable values. Intraoral scanners may offer slight advantages for direct capture of preparations, eliminating intermediate impression-making steps that introduce potential distortion.
​

Trueness Comparison

Trueness measurements show more nuanced patterns. Extraoral scanners produce trueness values of 7.7–67.3 μm compared to 24.5–95 μm for intraoral systems in full-arch applications. However, this advantage primarily manifests when scanning impressions or casts under optimal conditions. Direct intraoral scanning eliminates dimensional changes associated with impression material polymerization and stone setting.
​

Modern high-end intraoral scanners (TRIOS 5, Medit i700, Primescan) approach extraoral scanner trueness for limited-span applications, with values of 19.21–60.5 μm that fall within clinical acceptability thresholds.
​

Clinical Recommendations

Evidence supports the following scanner selection guidelines:

Quadrant impressions and single-unit preparations: Intraoral scanners provide equivalent or superior accuracy to extraoral methods
​

Full-arch dentate impressions: High-end IOS (TRIOS, Primescan, Medit i700) achieve acceptable accuracy; extraoral scanning offers marginal precision advantages
​

Multiple implant cases (>4 implants): Extraoral scanning of conventional impressions or photogrammetry demonstrates superior accuracy
​

Complete denture applications: Extraoral scanning of impressions provides significantly better precision for peripheral borders and mobile tissues
​

Laboratory workflow: Extraoral scanners offer consistently higher precision and trueness for model digitization
​

Both intraoral and extraoral scanning technologies have achieved clinical maturity, with accuracy sufficient for most prosthodontic applications. Scanner selection should consider the specific clinical scenario, spanning distance, tissue characteristics, and workflow integration rather than relying solely on precision metrics. The trend toward AI-assisted scanning, improved optical systems, and enhanced algorithms continues to narrow the performance gap between these complementary technologies.
--
หมอจัดฟัน

vulpi

16/01/2026 à 07h59

Je te rejoins, il y a un bulle financière et le ménage va se faire.