Lifeboat Foundation

Google Maps is introducing a generative AI feature to help you discover new places, the company announced today.

Using large language models (LLMs), the new feature analyzes the over 250 million locations on Google Maps and contributions from over 300 million Local Guides to pull up suggestions based on what you’re looking for. For instance, if you want to find cool thrift shops in San Francisco, you can search “places with a vintage vibe in SF,” and Maps will generate shopping recommendations organized by categories, as well as “photo carousels and review summaries,” the company explains. The new feature is meant to feel more conversational than the ordinary search experience. If you ask a follow-up question like “How about lunch?” the AI will take your previous interest in vintage and find restaurants that meet the criteria, such as an old-school diner.

The company says the feature should be able to generate recommendations on even the most niche or specific query.

“Bodily maps of musical sensations across cultures”

Emotions, bodily sensations and movement are integral parts of musical experiences. Yet, it remains unknown i) whether emotional connotations and structural features of music elicit discrete bodily sensations and ii) whether these sensations are culturally consistent. We addressed these questions in a cross-cultural study with Western (European and North American, n = 903) and East Asian (Chinese, n = 1035). We precented participants with silhouettes of human bodies and asked them to indicate the bodily regions whose activity they felt changing while listening to Western and Asian musical pieces with varying emotional and acoustic qualities. The resulting bodily sensation maps (BSMs) varied as a function of the emotional qualities of the songs, particularly in the limb, chest, and head regions.

Although a significant number of neuromorphic devices applied to RC have been reported in recent years, the majority of these efforts have focused on shallow-RC with monotonic reservoir state spaces¹⁹. This can be attributed to the heavy reliance on monotonic carrier dynamics when using reported neuromorphic devices as reservoirs to map sequence signals, which gives rise to several noteworthy issues for RC when performing different spatiotemporal tasks. One major issue is that the narrow range ratio of spatial characteristics makes it difficult to extract the diversity spatial feature of sequence signal, which greatly limits the richness of the reservoir space state. As a result, during the process of mapping complex sequence signals, the reservoir state tends to overlap, making it difficult to effectively separate the spatial characteristics within complex information and subsequently reducing recognition accuracy. Another issue is the limited rang ratio of temporal characteristic, which hinders efficient extraction of temporal feature from sequential signals with diverse time-scales. For example, when performing dynamic trajectory prediction with abundant time-scales, the limited range ratio of temporal characteristic is difficult to adapt to the signal with different temporal feature, which severely limit the correlation of prediction. Despite researchers have achieved multi-scale temporal characteristics by increasing the number of signal modes in the input layer based on shallow-RC networks²⁰, as shown in the Supplement Information Fig. S1, the limitation of shallow-RC on spatial characteristics remain unresolved. Furthermore, increasing the input layer also means the requirement of more encoding design for sequence signals and the utilization of more physical devices to receive different modes of physical signals. This significantly increases the signal error rate and pre-processing cost of the input signals, which is detrimental to the robustness of RC. Therefore, developing new neuromorphic reservoir devices along with new RC networks to simultaneously meet large-scale spatial and temporal characteristics are highly required, which is crucial for achieving high-performance recognition and prediction in complex spatiotemporal tasks for RC networks.

Interestingly, primates in nature are able to quickly and accurately recognize complex object information, such as facial recognition, with the help of advanced synaptic dynamics mechanisms. Brain science research on primates has confirmed^20,21,22 that primates use a distributed memory characteristic for processing complex information. When the nervous system processes a task, each neuron and neural circuit processes only a part of the information and generates a part of the output. For example, as shown in Fig. 1a, when a primate observes an unfamiliar face, neurons in the temporal polar (TP) region (blue) respond to familiar eye features, forming TP feature memory. Neuron cells in the anterior-medial (AM) region respond to unfamiliar lip features, forming AM feature memory²³. In this way, all outputs are integrated by the cerebral cortex to form the final output result, significantly improving the computational efficiency and accuracy for complex information processing. The physiological significance of distributed memory characteristics in primates serves as inspiration for the design of physical node devices with distributed reservoir states in the reservoir layer of the RC system. These devices are intended to facilitate the distributed mapping of spatiotemporal signals. However, to date, no such devices have been demonstrated.

In this work, inspired by the distributed memory characteristic of primates, an ultra-short channel organic neuromorphic vertical field effect transistor with distributed reservoir states is proposed and used to implement grouped-RC networks. By coupling multivariate physical mechanisms into a single device, the dynamic states of carriers are greatly enriched. As reservoir nodes, sequential signals can be mapped to a distributed reservoir state space by various carrier dynamics, rather than by monotonic carrier dynamics. Additionally, a vertical architecture with ultra-short nanometers transport distance is adopted to eliminate the driving force of the dissociation exciton, thereby improving the feedback strength of the device and the reducing the overlap between different reservoir state space, which only cause negligible additional power. Consequently, the device serves as a reservoir capable of mapping sequential signals into distributed reservoir state space with 1,152 reservoir states, and the range ratio of temporal (key parameters for prediction) and spatial characteristics (key parameters for recognition) can simultaneously reach 2,640 and 650, respectively, which are superior to the reported neuromorphic devices. Therefore, the grouped-RC network implemented based on the device can simultaneously meet the requirements of two different spatiotemporal types task (broad-spectrum image recognition and dynamic trajectory prediction) and exhibits over 94% recognition accuracy and over 95% prediction correlation, respectively. This work proposes a strategy for developing neural hardware for complex reservoir computing networks and has great potential in the development of a new generation of artificial neuromorphic hardware and brain-like computing.

Mapping connectome hubs in the brain.

Shared with Dropbox.

Great science stands the test of time: After 10 years of its publication, this paper with is still listed as trending in.

@PNASNews webpage! +3.5M views & 1k citations.

Emotions are often felt in the body, and somatosensory feedback has been proposed to trigger conscious emotional experiences. Here we reveal maps of bodily sensations associated with different emotions using a unique topographical self-report method. In five experiments, participants (n = 701) were shown two silhouettes of bodies alongside emotional words, stories, movies, or facial expressions. They were asked to color the bodily regions whose activity they felt increasing or decreasing while viewing each stimulus. Different emotions were consistently associated with statistically separable bodily sensation maps across experiments. These maps were concordant across West European and East Asian samples. Statistical classifiers distinguished emotion-specific activation maps accurately, confirming independence of topographies across emotions.

The use of NISQ devices for useful quantum simulations of materials and chemistry is still mainly limited by the necessary circuit depth. Here, the authors propose to combine classically-generated effective Hamiltonians, hybrid fermion-to-qubit mapping and circuit optimisations to bring this requirement closer to experimental feasibility.

Over ten years ago, the Dark Energy Survey (DES) began mapping the universe to find evidence that could help us understand the nature of the mysterious phenomenon known as dark energy. I’m one of more than 100 contributing scientists that have helped produce the final DES measurement, which has just been released at the 243rd American Astronomical Society meeting in New Orleans.

Dark energy is estimated to make up nearly 70% of the observable universe, yet we still don’t understand what it is. While its nature remains mysterious, the impact of dark energy is felt on grand scales. Its primary effect is to drive the accelerating expansion of the universe.

The announcement in New Orleans may take us closer to a better understanding of this form of energy. Among other things, it gives us the opportunity to test our observations against an idea called the cosmological constant that was introduced by Albert Einstein in 1917 as a way of counteracting the effects of gravity in his equations to achieve a universe that was neither expanding nor contracting. Einstein later removed it from his calculations.

The latest star maps are rewriting the story of our Milky Way, revealing a much more tumultuous history than astronomers suspected.

By Ann Finkbeiner

Townships get little by way of public resources compared with rich suburbs. By mapping the problem, researchers like Raesetje Sefala hope to reverse it.