Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: robotics/AI – Page 163

What can theoretical physics teach us about knitting?

The practice of purposely looping thread to create intricate knit garments and blankets has existed for millennia. Though its precise origins have been lost to history, artifacts like a pair of wool socks from ancient Egypt suggest it dates back as early as the third to fifth century CE. Yet, for all its long-standing ubiquity, the physics behind knitting remains surprisingly elusive.

“Knitting is one of those weird, seemingly simple but deceptively complex things we take for granted,” says and visiting scholar at the University of Pennsylvania, Lauren Niu, who recently took up the craft as a means to study how “geometry influences the mechanical properties and behavior of materials.”

Despite centuries of accumulated knowledge, predicting how a particular knit pattern will behave remains difficult—even with modern digital tools and automated knitting machines. “It’s been around for so long, but we don’t really know how it works,” Niu notes. “We rely on intuition and trial and error, but translating that into precise, predictive science is a challenge.”

Taking AI Welfare Seriously

In this interview Jeff Sebo discusses the ethical implications of artificial intelligence and why we must take the possibility of AI sentience seriously now. He explores challenges in measuring moral significance, the risks of dismissing AI as mere tools, and strategies to mitigate suffering in artificial systems. Drawing on themes from the paper ‘Taking AI Welfare Seriously’ and his up and coming book ‘The Moral Circle’, Sebo examines how to detect markers of sentience in AI systems, and what to do about it. We explore ethical considerations through the lens of population ethics, AI governance (especially important in an AI arms race), and discuss indirect approaches detecting sentience, as well as AI aiding in human welfare. This rigorous conversation probes the foundations of consciousness, moral relevance, and the future of ethical AI design.

Paper ‘Taking AI Welfare Seriously’: https://eleosai.org/papers/20241030_T… — The Moral Circle by Jeff Sebo: https://www.amazon.com.au/Moral-Circl?tag=lifeboatfound-20?tag=lifeboatfound-20… Jeff’s Website: https://jeffsebo.net/ Eleos AI: https://eleosai.org/ Chapters: 00:00 Intro 01:40 Implications of failing to take AI welfare seriously 04:43 Engaging the disengaged 08:18 How Blake Lemoine’s ‘disclosure’ influenced public discourse 12:45 Will people take AI sentience seriously if it is seen tools or commodities? 16:19 Importance, neglectedness and tractability (INT) 20:40 Tractability: Difficulties in measuring moral significance — i.e. by aggregate brain mass 22:25 Population ethics and the repugnant conclusion 25:16 Pascal’s mugging: low probabilities of infinite or astronomically large costs and rewards 31:21 Distinguishing real high stakes causes from infinite utility scams 33:45 The nature of consciousness, and what to measure in looking for moral significance in AI 39:35 Varieties of views on what’s important. Computational functionalism 44:34 AI arms race dynamics and the need for governance 48:57 Indirect approaches to achieving ideal solutions — Indirect normativity 51:38 The marker method — looking for morally relevant behavioral & anatomical markers in AI 56:39 What to do about suffering in AI? 1:00:20 Building in fault tolerance to noxious experience into AI systems — reverse wireheading 1:05:15 Will AI be more friendly if it has sentience? 1:08:47 Book: The Moral Circle by Jeff Sebo 1:09:46 What kind of world could be achieved 1:12:44 Homeostasis, self-regulation and self-governance in sentient AI systems 1:16:30 AI to help humans improve mood and quality of experience 1:18:48 How to find out more about Jeff Sebo’s research 1:19:12 How to get involved Many thanks for tuning in! Please support SciFuture by subscribing and sharing! Have any ideas about people to interview? Want to be notified about future events? Any comments about the STF series? Please fill out this form: https://docs.google.com/forms/d/1mr9P… Kind regards, Adam Ford

Book — The Moral Circle by Jeff Sebo: https://www.amazon.com.au/Moral-Circl?tag=lifeboatfound-20?tag=lifeboatfound-20…

Jeff’s Website: https://jeffsebo.net/

Eleos AI: https://eleosai.org/

Chapters:

Continue reading “Taking AI Welfare Seriously” | >

AI designs an ultralight carbon nanomaterial that’s as strong as steel

Using machine learning, a team of researchers in Canada has created ultrahigh-strength carbon nanolattices, resulting in a material that’s as strong as carbon steel, but only as dense as Styrofoam.

The team noted last month that it was the first time this branch of AI had been used to optimize nano-architected materials. University of Toronto’s Peter Serles, one of the authors of the paper describing this work in Advanced Materials, praised the approach, saying, “It didn’t just replicate successful geometries from the training data; it learned from what changes to the shapes worked and what didn’t, enabling it to predict entirely new lattice geometries.”

To quickly recap, nanomaterials are engineered by arranging atoms or molecules in precise patterns, much like constructing structures with extremely tiny LEGO blocks. These materials often exhibit unique properties due to their nanoscale dimensions.

Online test-time adaptation for better generalization of interatomic potentials to out-of-distribution data

Molecular Dynamics (MD) simulation serves as a crucial technique across various disciplines including biology, chemistry, and material science1,2,3,4. MD simulations are typically based on interatomic potential functions that characterize the potential energy surface of the system, with atomic forces derived as the negative gradients of the potential energies. Subsequently, Newton’s laws of motion are applied to simulate the dynamic trajectories of the atoms. In ab initio MD simulations5, the energies and forces are accurately determined by solving the equations in quantum mechanics. However, the computational demands of ab initio MD limit its practicality in many scenarios. By learning from ab initio calculations, machine learning interatomic potentials (MLIPs) have been developed to achieve much more efficient MD simulations with ab initio-level accuracy6,7,8.

Despite their successes, the crucial challenge of implementing MLIPs is the distribution shift between training and test data. When using MLIPs for MD simulations, the data for inference are atomic structures that are continuously generated during simulations based on the predicted forces, and the training set should encompass a wide range of atomic structures to guarantee the accuracy of predictions. However, in fields such as phaseion9,10, catalysis11,12, and crystal growth13,14, the configurational space that needs to be explored is highly complex. This complexity makes it challenging to sample sufficient data for training and easy to make a potential that is not smooth enough to extrapolate to every relevant point. Consequently, a distribution shift between training and test datasets often occurs, which causes the degradation of test performance and leads to the emergence of unrealistic atomic structures, and finally the MD simulations collapse15.

H-shaped bionic robot mimics cheetah’s sprint using electric charge

In recent years, roboticists and computer scientists have developed a wide range of systems inspired by nature, particularly by humans and animals. By reproducing animal movements and behaviors, these robots could navigate real-world environments more effectively.

Researchers at Northeastern University in China recently developed a new H-shaped bionic robot that could replicate the movements that cheetahs make while running. This robot, introduced in a paper published in the Journal of Bionic Engineering, is based on piezoelectric materials, a class of materials that generate an electric charge when subjected to mechanical stress.

“The piezoelectric robot realizes linear motion, turning motion, and turning motion with different radii by the voltage differential driving method,” wrote Ying Li, Chaofeng Li and their colleagues in their paper. “A prototype with a weight of 38 g and dimensions of 150 × 80 × 31 mm3 was fabricated.”

Psilocybin for the treatment of Alzheimer’s disease

Mushrooms are probably the most miraculous entities because each mushroom can aid in a different way to cure each illness in the human biology. Much like cannabis is actually a cure all for so many ailments in humans so in turn are mushrooms able to do the same.

Alzheimer’s disease (AD) stands as a formidable neurodegenerative ailment and a prominent contributor to dementia. The scarcity of available therapies for AD accentuates the exigency for innovative treatment modalities. Psilocybin, a psychoactive alkaloid intrinsic to hallucinogenic mushrooms, has garnered attention within the neuropsychiatric realm due to its established safety and efficacy in treating depression. Nonetheless, its potential as a therapeutic avenue for AD remains largely uncharted. This comprehensive review endeavors to encapsulate the pharmacological effects of psilocybin while elucidating the existing evidence concerning its potential mechanisms contributing to a positive impact on AD. Specifically, the active metabolite of psilocybin, psilocin, elicits its effects through the modulation of the 5-hydroxytryptamine 2A receptor (5-HT2A receptor). This modulation causes heightened neural plasticity, diminished inflammation, and improvements in cognitive functions such as creativity, cognitive flexibility, and emotional facial recognition. Noteworthy is psilocybin’s promising role in mitigating anxiety and depression symptoms in AD patients. Acknowledging the attendant adverse reactions, we proffer strategies aimed at tempering or mitigating its hallucinogenic effects. Moreover, we broach the ethical and legal dimensions inherent in psilocybin’s exploration for AD treatment. By traversing these avenues, We propose therapeutic potential of psilocybin in the nuanced management of Alzheimer’s disease.

Alzheimer’s disease (AD) is a progressive neurodegenerative disease that is the leading cause of dementia in the elderly population (Anonymous, 2021). It is characterized by the deposition of amyloid-beta (Aβ) plaques, tau neurofibrillary tangles, and neuroinflammation (Scheltens et al., 2021). The prevalence of dementia is expected to rise as the global population grows and ages, with projections estimating a significant increase in the number of cases (Anonymous, 2022b). In 2019, the total cost of healthcare, long-term care, and hospice services for individuals aged 65 years and older with dementia in the United States was estimated at $2.2billion, so AD imposes a substantial burden on individuals, families, society, and the economy (Anonymous, 2022a). The U.S. Food and Drug Administration (FDA) has approved seven drugs for the treatment of AD, including rivastigmine, donepezil, galantamine, memantine, memantine combined with donepezil, aducanumab and lecanemab.

Scientists spent 10 years on a superbug mystery — Google’s AI solved it in 48 hours

Professor José R Penadés told the BBC that Google’s tool reached the same hypothesis that his team had – that superbugs can create a tail that allows them to move between species. In simpler terms, one can think of it as a master key that enables the bug to move from home to home.

Penadés asserts that his team’s research was unique and that the results hadn’t been published anywhere online for the AI to find. What’s more, he even reached out to Google to ask if they had access to his computer. Google assured him they did not.

Arguably even more remarkable is the fact that the AI provided four additional hypotheses. According to Penadés, all of them made sense. The team had not even considered one of the solutions, and is now investigating it further.

Synaptic device array integrates sensing, memory, and processing for artificial vision

In a development for artificial intelligence, researchers have unveiled a synaptic device array that shows promise for enhancing artificial visual systems. This innovative array, measuring a compact 0.7 × 0.7 cm2, integrates the capabilities of sensing, memory, and processing to mimic the intricate functions of the human visual system.

Utilizing wafer-scale monolayer molybdenum disulfide (MoS2) and for enhanced electron capture, the array exhibits remarkable coordination between optical and electrical components. It is capable of both writing and erasing images and has achieved a 96.5% accuracy in digit recognition, marking a significant leap forward in the development of large-scale neuromorphic systems.

The human visual system processes complex visual data efficiently through an interconnected network that allows for parallel processing. However, current artificial vision systems face numerous challenges, including circuit complexity, high power consumption, and difficulties in miniaturization.