Fine-grained knowledge about manipulable objects is well-predicted by contrastive language image pre-training

iScience, Journal Year: 2024, Volume and Issue: 27(7), P. 110297 - 110297

Published: June 18, 2024

Language: Английский

---

Acute Neuropixels Recordings in the Marmoset Monkey

eNeuro, Journal Year: 2024, Volume and Issue: 11(5), P. ENEURO.0544 - 23.2024

Published: April 24, 2024

High-density linear probes, such as Neuropixels, provide an unprecedented opportunity to understand how neural populations within specific laminar compartments contribute behavior. Marmoset monkeys, unlike macaque have a lissencephalic (smooth) cortex that enables recording perpendicular the cortical surface, thus making them ideal animal model for studying computations. Here we present method acute Neuropixels recordings in common marmoset (Callithrix jacchus). The approach replaces native dura with artificial silicon-based grants visual access which is helpful avoiding blood vessels, ensures penetrations, and could be used conjunction optical imaging or optogenetic techniques. chamber housing simple maintain minimal risk of infection combined semichronic microdrives wireless hardware. This technique repeated penetrations over period several months. With occasional removal tissue growth on pial can performed year more. fully compatible enabling hundreds single neurons distributed throughout column.

Language: Английский

---

How Adversarial Rem-Dreams May Facilitate Creativity, and Why We Become Aware of Them

Published: May 10, 2024

The importance of sleep for healthy brain function is widely acknowledged. However, it remains unclear how the internal generation dreams might facilitate cognitive processes. In this perspective we review a computational approach inspired by artificial intelligence that proposes framework occurring during rapid-eye-movement (REM) can contribute to learning and creativity. framework, REM are characterized an adversarial process that, against dream reality, tell discriminator network classify internally created sensory activity as real. Such dreaming shown emergence real-world semantic representations in higher cortical areas. We further discuss potential contributions beyond learning, such balancing fantastic realistic elements, facilitating occurrence creative insights. characterize non-REM (NREM) dreams, where single hippocampal memory replayed at time, serving complementary role improving robustness environmental perturbations. finally explain subjects become aware but less NREM awareness phenomenon wake, lucid may appear.

Language: Английский

---

ModelVsBaby: a Developmentally Motivated Benchmark of Out-of-Distribution Object Recognition

Published: May 21, 2024

Deep neural networks have recently become remarkable computational tools for thinking about human visual learning. Recent studies explored the effects of altering naturalistic images and compared responses both humans models, providing valuable insights into their functioning how deep can shape our understanding Critically, much learning happens throughout early development. Yet, well-controlled benchmarks comparing AI models with young are scarce. Here, we present a developmentally motivated benchmark out-of-distribution (OOD) object recognition. Our benchmark, ModelVsBaby, includes set OOD conditions that long been studied in vision science literature, expected to be sensitive development object-recognition humans: silhouette, geon, occluded, blurred, crowded background, baseline realistic condition. Along stimuli, release unique dataset 2-year-old children stimuli. preliminary analyses show several interesting patterns: 2-year-olds achieve 80% accuracy silhouette condition, nearly as well condition (chance=12%). They also perform above chance, near 60% accuracy, on other challenging conditions. We evaluate image-text association (CLIP) trained varying amounts internet-scale datasets. The model performances enough data, all learnable by artificial learners. However, Realistic Silhouette learned fewer training data similar humans. stimuli infant responses, provide an essential steppingstone building aligned terms outcomes trajectory. This endeavor furnish creating better improving efficiency systems practical applications. Future work may use test more age groups, detailed comparison various flavors “developmental alignment".

Language: Английский

---

How Adversarial REM Dreams May Facilitate Creativity, and Why We Become Aware of Them

Clinical and Translational Neuroscience, Journal Year: 2024, Volume and Issue: 8(2), P. 21 - 21

Published: May 31, 2024

The importance of sleep for healthy brain function is widely acknowledged. However, it remains unclear how the internal generation dreams might facilitate cognitive processes. In this perspective, we review a computational approach inspired by artificial intelligence that proposes framework occurring during rapid-eye-movement (REM) can contribute to learning and creativity. framework, REM are characterized an adversarial process that, against dream reality, tells discriminator network classify internally created sensory activity as real. Such dreaming shown emergence real-world semantic representations in higher cortical areas. We further discuss potential contributions beyond learning, such balancing fantastic realistic elements facilitating occurrence creative insights. characterize non-REM (NREM) dreams, where single hippocampal memory replayed at time, serving complementary role improving robustness environmental perturbations. finally explain subjects become aware but less NREM content- state-awareness wake, dream, lucid may appear.

Language: Английский

---

How to Turn a Leaky Learner into a Sealed One

Communications in computer and information science, Journal Year: 2024, Volume and Issue: unknown, P. 29 - 40

Published: Jan. 1, 2024

Language: Английский

---

High refresh rate display for natural monocular viewing in AOSLO psychophysics experiments

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: May 31, 2024

Abstract By combining an external display operating at 360 frames per second with Adaptive Optics Scanning Laser Ophthalmoscope (AOSLO) for human foveal imaging, we demonstrate color stimulus delivery high spatial and temporal resolution in AOSLO psychophysics experiments. A custom pupil relay enables viewing of the through a 3-mm effective diameter provides refractive error correction from -8 to +4 diopters. Performance assembled aligned was validated by measuring wavefront across field view range, as-built Strehl ratio 0.64 or better. High-acuity stimuli were rendered on imaged that frequencies up 54 cycles degree, corresponding 20/11 visual acuity, are resolved. The completed then used render fixation markers monitor, continuous retinal montage spanning 9.4 5.4 degrees angle acquired AOSLO. We conducted eye-tracking experiments during free-viewing high-acuity tasks polychromatic images presented display. Sub-arcminute eye position uncertainty achieved, enabling precise localization line sight monitor while simultaneously imaging fine structure central fovea. This refresh rate overcomes temporal, spectral, limitations AOSLO-based presentation, natural monocular

Language: Английский

---

Fast and robust visual object recognition in young children

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 16, 2024

Abstract By adulthood, humans rapidly identify objects from sparse visual displays and across significant disruptions to their appearance. What are the minimal conditions needed achieve robust recognition abilities when might these develop? To answer questions, we investigated upper-limits of children’s object abilities. We found that children as young 3 years successfully identified at speeds 100 ms (both forward backward masked) under disrupted viewing conditions. contrast, a range computational models implemented with biologically informed properties or optimized for did not reach child-level performance. Models only matched if they received more examples than capable experiencing. These findings highlight robustness human system in absence extensive experience important developmental constraints building plausible machines. Teaser The preschool rival those state-of-the-art artificial intelligence models.

Language: Английский

---

Asymmetries in foveal vision

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 21, 2024

Abstract Visual perception is characterized by known asymmetries in the visual field; human’s sensitivity higher along horizontal than vertical meridian, and lower upper meridian. These decrease with decreasing eccentricity from periphery to center of gaze, suggesting that they may be absent 1-deg foveola, retinal region used explore scenes at high-resolution. Using high-precision eyetracking gaze-contingent display, allowing for accurate control over stimulated foveolar location despite continuous eye motion fixation, we investigated fine discrimination different isoeccentric locations across foveola parafovea. Although tested were only 0.3 deg away show that, similar more eccentric locations, humans are sensitive stimuli presented Whereas magnitude this asymmetry reduced meridian comparable but, interestingly, reversed: objects slightly above gaze easily discerned when same below gaze. Therefore, far being uniform, as often assumed, vision perceptual asymmetries. Further, these differ not but also direction compared those present just ∼4deg resulting overall foveal extrafoveal fields.

Language: Английский

---

Acute Neuropixels recordings in the marmoset monkey

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Dec. 15, 2023

High-density linear probes, like Neuropixels, provide an unprecedented opportunity to understand how neural populations within specific laminar compartments contribute behavior. Marmoset monkeys, unlike macaque have a lissencephalic (smooth) cortex that enables recording perpendicular the cortical surface, thus making them ideal animal model for studying computations. Here we present method acute Neuropixels recordings in common marmoset (

Language: Английский

---