Show Notes 13 September 2024
Story 1: This new AI modeler can turn pictures into 3D sculptures in seconds
Source: TechRadar.com Story by Eric Hal Schwartz
See also: https://arxiv.org/abs/2408.00653
See also a Tom’s Hardware feature about VFusion3D: https://www.tomshardware.com/tech-industry/artificial-intelligence/ai-driven-technique-can-generate-quality-3d-assets-from-2d-images-in-seconds
- My comments – The key trend reflected in this news feature has to do with multiple worldwide efforts to use AI to quickly transform 2D images into 3D “assets” / models. Note, not 3D sculptures, as the headline suggests, but 3D models/assets that can be used in a host of ways.
- First some background – what are “3D assets”?
- A 3D asset refers to any digital object created in three dimensions using specialized software. These assets can include models of characters, environments, objects, and more. They are widely used in various industries such as gaming, virtual reality (VR), augmented reality (AR), film, and e-commerce.
- 3D assets are often created by skilled artists and can be bought and sold on 3D stock marketplaces. They help save time and reduce costs in production by providing ready-made models that can be customized and integrated into different projects.
- Now here’s the news:
- The new Stable Fast 3D AI model from London-based stability.ai can produce a 3D asset from a single [2D] image in about half a second, which is a staggering speed compared to the usual rendering system.
- And even if it takes more than a second and a single image to perfect the 3D model, there’s a huge potential demand for this kind of tool in everything from architecture to video games.
- Stable Fast 3D produces a 3D model of an image that isn’t just a digital representation. The resulting model’s high-quality UV unwrapped mesh ensures that textures are mapped accurately onto 3D models….
- Optional side note – what is “UV unwrapped mesh”?
- A UV unwrapped mesh is a 3D model that has been flattened into a 2D representation to allow textures to be accurately applied. This process, known as UV unwrapping, involves projecting the 3D surface of the model onto a 2D plane. The term “UV” refers to the axes of the 2D texture space, as “X,” “Y,” and “Z” are already used for the 3D model’s coordinates.
- Here’s a simple breakdown of the process:
- Model Creation: A 3D model is created using polygons.
- Seam Marking: Seams are marked on the model where it will be “cut” and unfolded.
- Unwrapping: The model is unfolded along these seams to create a 2D layout.
- Texture Mapping: Textures are applied to this 2D layout, which then map back onto the 3D model.
- This technique is essential for ensuring that textures look correct and undistorted when applied to the 3D model.
- Sources:
- UV mapping – Wikipedia. https://en.wikipedia.org/wiki/UV_mapping
- Unwrapping The Secrets of UV Maps: Flattening Meshes For Texture …. https://blendernotes.com/unwrapping-the-secrets-of-uv-maps-flattening-meshes-for-texture-mapping/
- Master UV Unwrapping in Blender – Blender Base Camp. https://blenderbasecamp.com/master-uv-unwrapping-in-blender/
- UV Unwrapping Made Easy – The Ultimate Blender Guide. https://blendanimate.com/uv-unwrapping-made-easy-the-ultimate-blender-guide/
- This isn’t Stability AI’s first 3D mode-maker, but it is much faster than the earlier SV3D model, which took up to 10 minutes. Stable Fast 3D could theoretically make close to1,200 models and variants in 10 minutes.
- But the company can’t rest on its laurels as other generative AI developers experiment with similar ideas.
- OpenAI’s POINT-E model can make 3D models from text prompts
- and Adobe has its own experiments that could be deployed to its Firefly service.
- See also the link regarding VFusion3D developed by Meta and Oxford University scientists.
- Why there’s so much interest in 3D asset makers:
- It could help produce digital environments for virtual reality and games, allow for experimentation in designing buildings and decorating rooms, and even give boutique toymakers a chance to play with ideas that translate to the real world without the expense of producing a lot of failed prototypes.
- Even shopping online could change as you might be able to browse through interactive, 3D views of products.
Story 2: Researchers built an ‘AI Scientist’ — what can it do? — The large language model does everything from reading the literature to writing and reviewing its own papers, but it has a limited range of applicability so far.
Source: Nature.com Story by Davide Castelvecchi
Link: https://www.nature.com/articles/d41586-024-02842-3
- Could science be fully automated? A team of machine-learning researchers has now tried.
- ‘AI Scientist’, created by a team at Tokyo company Sakana AI and at academic labs in Canada and the United Kingdom, performs the full cycle of research from reading the existing literature on a problem and formulating hypothesis for new developments to trying out solutions and writing a paper.
- AI Scientist even does some of the job of peer reviewers and evaluates its own results.
- AI Scientist joins a slew of efforts to create AI agents that have automated at least parts of the scientific process.
- One of the co-creators noted, “To my knowledge, no one has yet done the total scientific community, all in one system.”
- Reality Check – The output is not earth-shattering so far, and the system can only do research in the field of machine learning itself.
- In specific, AI Scientist is lacking what most scientists would consider the crucial part of doing science — the ability to do laboratory work.
- AI Scientist is based on a large language model (LLM). Using a paper that describes a machine learning algorithm as a template, it starts by searching the literature for similar work.
- The team then employed the technique called evolutionary computation, which is inspired by the mutations and natural selection of Darwinian evolution.
- It proceeds in steps, applying small, random changes to an algorithm and selecting the ones that provide an improvement in efficiency.
- To do so, AI Scientist conducts its own ‘experiments’ by running the algorithms and measuring how they perform.
- At the end, it produces a paper, and evaluates it in a sort of automated peer review. After ‘augmenting the literature’ this way, the algorithm can then start the cycle again, building on its own results.
Story 3: New memory tech unveiled that reduces AI processing energy requirements by 1,000 times or more – New CRAM technology gives RAM chips the power to process data, not just store it.
Source: Tomshardware.com Story by Jeff Butts
See also: https://www.nature.com/articles/s44335-024-00003-3
See also: https://www.digitaltrends.com/computing/cram-in-memory-reduces-ai-energy-1000x/
- Artificial intelligence (AI) computing requires tremendous amounts of electricity, but targeted research might hold the key to greatly reducing that. A team of researchers at the University of Minnesota Twin Cities has developed technology that could reduce the energy consumption required by AI processing by a factor of at least a thousand.
- In essence, they’ve created a shortcut in the normal practice of AI computations that greatly reduces the energy requirement for the task.
- In current AI computing, data is transferred between the logic components processing it and where data is stored (memory/storage – a.k.a. RAM). This constant shuttling of information back and forth is responsible for consuming as much as 200 times the energy used in the computation, according to this research.
- Thus, the researchers have turned to Computational Random-Access Memory (CRAM) to address this.
- The Computational Random-Access Memory the research team has developed places a high-density, reconfigurable spintronic in-memory compute substrate within the memory cells themselves.
- Optional side note, what is spintronics?
- Spintronics, short for “spin transport electronics,” is a field of study that focuses on the intrinsic spin of electrons and their associated magnetic moment, in addition to their fundamental electronic charge, in solid-state devices. Unlike traditional electronics, which rely solely on the charge of electrons to store and process information, spintronics leverages the electron’s spin state (up or down) to add an extra layer of functionality.
- This technology has significant implications for data storage and transfer, potentially leading to more efficient and faster devices. Spintronic devices are already used in applications like hard drives and magnetic sensors.
- Sources:
- Spintronics – Wikipedia. https://en.wikipedia.org/wiki/Spintronics
- Spintronics Applications, Definition, Devices, Future Works – BYJU’S. https://byjus.com/physics/spintronics/
- Introduction to Spintronics – UMD. https://www.physics.umd.edu/rgroups/spin/intro.html
- What is Spintronics? – Definition from Techopedia. https://www.techopedia.com/definition/15025/spintronics
- Using Computational Random-Access Memory, the data never leaves memory, instead undergoing processing entirely within the computer’s memory array.
- According to the University of Minnesota Twin Cities research team, this allows the system running the AI computing application an energy consumption improvement “on the order of 1,000x over a [current] state-of-the-art solution.”
Story 4: Cleaning up the aging brain: Scientists restore brain’s trash disposal system
Source: ScienceBlog.com Story by University of Rochester
See also: https://neurosciencenews.com/csf-neurology-aging-27549/
- First some background:
- The human brain has a specialized system called the glymphatic system that helps clean out toxins. This system works primarily during deep sleep and involves a series of tubes that carry fresh cerebrospinal fluid into the brain. This fluid mixes with the waste-filled fluid surrounding brain cells and then flushes the mix out of the brain and into the blood.
- Additionally, the meningeal lymphatic vessels in the outer brain membrane also play a role in regulating the brain’s fluids and flushing out excess proteins that can build up. This process is crucial for maintaining brain health and preventing conditions like Alzheimer’s disease.
- Sources:
- Are toxins flushed out of the brain during sleep? – Harvard Health. https://www.health.harvard.edu/mind-and-mood/are-toxins-flushed-out-of-the-brain-during-sleep
- The brain’s drain: How our brains flush out their waste and toxins. https://www.newscientist.com/article/2210825-the-brains-drain-how-our-brains-flush-out-their-waste-and-toxins/
- Does the brain flush out toxins while you sleep? | Live Science. https://www.livescience.com/health/sleep/does-the-brain-flush-out-toxins-while-you-sleep
- How Sleep Clears the Brain | National Institutes of Health (NIH). https://www.nih.gov/news-events/nih-research-matters/how-sleep-clears-brain
- Alzheimer’s, Parkinson’s, and other neurological disorders can be seen as “dirty brain” diseases, where the brain struggles to clear out harmful waste. Aging is a key risk factor because, as we grow older, our brain’s ability to remove toxic buildup slows down.
- However, new research in mice conducted by a team of scientists at the University of Rochester demonstrates that it’s possible to reverse age-related effects and restore the brain’s waste-clearing process.
- In healthy and young brains, the glymphatic system does a good job of flushing away these toxic proteins, however, as we age, this system slows, setting the stage for these diseases.
- Once laden with protein waste, cerebrospinal fluid in the skull needs to make its way to the lymphatic system and ultimately the kidneys, where it is processed along with the body’s other waste.
- The new research combines advanced imaging and particle tracking techniques to describe for the first time in detail the route via the cervical lymph vessels in the neck through which half of dirty cerebrospinal fluid exits the brain.
- One of the research team noted, “Unlike the cardiovascular system which has one big pump, the heart, fluid in the lymphatic system is instead transported by a network of tiny pumps”.
- The team then set out to see if they could revive the tiny pumps [lymphangions] and identified a hormone-like compound drug commonly used medically to induce labor and known to aid smooth muscle contraction.
- The tiny pumps are lined with smooth muscle cells, and when the researchers applied the drug to the cervical lymph vessels in older mice, the frequency of contractions and the flow of dirty cerebrospinal fluid from the brain both increased, returning to a level of efficiency found in younger mice.
Honorable Mentions:
Story: Man Receives World’s First Lung Cancer Vaccine
Source: ExtremeTech.com Story by Adrianna Nine
Link: https://www.extremetech.com/science/man-receives-worlds-first-lung-cancer-vaccine
- A man from London has become the first human to receive an experimental vaccine aimed at preventing lung cancer. At the United Kingdom’s National Institute for Health Research (NIHR) on Tuesday, doctors delivered the 67-year-old’s introductory dose of BNT116. The messenger RNA (mRNA) immunotherapy will hopefully “train” the man’s body to fight non-small cell lung cancer, which makes up roughly 85% of lung cancer diagnoses worldwide.
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
Story: Scientists may have found a radical solution for making your hamburger less bad for the planet
Source: The Washington Post via MSN Story by Shannon Osaka
- DAVIS, Calif. — Sushi, a four-week-old Holstein calf, was lying in a pen under the hum of a metal fan when a group of professors and graduate students arrived to sample his stomach. The male calf greeted the researchers with a friendly nibble of their clothing, then flopped back down lazily on a bed of rice hulls.
- But even as the calf slumbered, deep in his four-chambered stomach, minuscule organisms were hard at work. Fungi, bacteria, and other tiny creatures were breaking down feed into energy and chemicals, setting in motion an ancient process that today heats the Earth more than every flight across the world combined.
- Scientists here are hunting for a way to transform Sushi’s gut — so that he no longer releases planet-warming methane.
- The 125-pound calf belongs to the first wave of a multiyear, approximately $30 million experiment by scientists at the University of California at Davis and the Innovative Genomics Institute to change the inner workings of the cow stomach. Cattle, one of the most-consumed creatures on the planet, produce enormous amounts of methane, a powerful greenhouse gas that is responsible for 30 percent of global warming.
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
Story: New Genetically Engineered Wood Can Store Carbon and Reduce Emissions
Source: Science Daily Story from University of Maryland
Link: https://www.sciencedaily.com/releases/2024/08/240812123203.htm
- Researchers at the University of Maryland genetically modified poplar trees to produce high-performance, structural wood without the use of chemicals or energy intensive processing. Made from traditional wood, Engineered wood is often seen as a renewable replacement for traditional building materials like steel, cement, glass and plastic. It also has the potential to store carbon for a longer time than traditional wood because it can resist deterioration, making it useful in efforts to reduce carbon emissions.
- But the hurdle to true sustainability in engineered wood is that it requires processing with volatile chemicals and a significant amount of energy, and produces considerable waste. The researchers edited one gene in live poplar trees, which then grew wood ready for engineering without processing. The study was published online on August 12, 2024, in the Journal Matter.
- “We are very excited to demonstrate an innovative approach that combines genetic engineering and wood engineering, to sustainably sequester and store carbon in a resilient super wood form, said Yiping Qi, a professor in the Department of Plant Science and Landscape Architecture at UMD and a corresponding author of the study, “Carbon sequestration is critical in our fight against climate change, and such engineered wood may find many uses in the future bioeconomy.”
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
Story: Robots learn to feel human touch without artificial skin – AI-powered robot allows entirely new ways for humans to interact with machines
Source: UK’s Independent Story by Anthony Cuthbertson
Link: https://www.independent.co.uk/tech/robots-skin-touch-humans-ai-b2600365.html
- Scientists have discovered a way to give robots a sense of touch without the need for expensive artificial skin.
- A team from the German Aerospace Centre instead used artificial intelligence to interpret the signals from a robot’s internal sensors in order to accurately detect external force and pressure.
- The researchers said the new approach would negate the need for costly biometric skins and sensors, while also enabling a “shift from conventional modalities towards adaptability, flexibility, and intuitive handling”.