My nephew works for DeepMind in AI. When I look to the articles he’s published, I can barely understand the titles, much less the content of the articles.
Graphene making headlines again, first as a supplement to concrete to strengthen it. Stronger concrete, less concrete, less steel, less CO2. Then, graphene found locked in with diamond forming structures in a meteorite never seen before. Value - structure characteristics would presumably allow for high speed battery charging and now tha tthey know what they look like and can estimate what conditions formed them they can work on replicating them. The super material reinforcing rinks, cars and buildings - BBC News "By adding as little as 0.1% graphene into cement and aggregate, you can potentially use less material to get the same performance," explains Mr Baker. Reducing the amount of concrete used in construction for instance by 30%, could lower global CO2 emissions by 2-3%, he estimates. Besides the roller disco, Mr Baker and his colleagues have also trialled the graphene-infused concrete, known as concretene, in a gym floor in Wiltshire and some road projects, including a section of the A1 several hundred metres long in Northumberland. The team will also pour concretene in an as-yet undisclosed project in the United Arab Emirates this year. These early trials have comprised fairly straightforward projects, Mr Baker explains - floor slabs, rather than walls or elevated platforms, which might be more risky. So far, the concrete has performed as expected, though. Strange, never-before-seen diamond crystal structure found inside 'Diablo canyon' meteorite (msn.com) While studying lonsdaleite in the meteorite, the researchers found something odd. Instead of the pure hexagonal structures they were expecting, the researchers found growths of another carbon-based material called graphene interlocking with the diamond. These growths are known as diaphites, and inside the meteorite, they form in a particularly intriguing layered pattern. In between these layers are "stacking faults," which mean the layers don't line up perfectly, the researchers said in a statement. Finding diaphites in the meteoritic lonsdaleite suggests that this material can be found in other carbonaceous material, the scientists wrote in the study, which means it could be readily available to use as a resource. The finding also gives the researchers a better sense of the pressures and temperatures needed to create the structure. Graphene is made of a one-atom-thick sheet of carbon, arranged in hexagons. Although research on this material is still ongoing, the material has many potential applications. Because it is both as light as a feather and as strong as a diamond; both transparent and highly conductive; and 1 million times thinner than a human hair, it could one day be used for more targeted medicines, tinier electronics with lighting-fast charging speeds, or faster and bendier technology, the researchers said. And now that researchers have discovered these graphene growths inside meteorites, it's possible to learn more about how they form — and thus how to make them in the lab. "Through the controlled layer growth of structures, it should be possible to design materials that are both ultra-hard and also ductile, as well as have adjustable electronic properties from a conductor to an insulator," Christoph Salzmann, a chemist at University College London and co-author of a paper describing the research, said in the statement.
MIT boffins make AI chips '1 million times faster than the synapses in the human brain' (msn.com) A team at MIT reports that it has built AI chips that mimic synapses, but are a million times faster, and are additionally massively more energy efficient than current designs. The inorganic material is also easy to fit into current chip-building kit. "Once you have an analog processor, you will no longer be training networks everyone else is working on. You will be training networks with unprecedented complexities that no one else can afford to, and therefore vastly outperform them all. In other words, this is not a faster car, this is a spacecraft," said lead author and MIT postdoc Murat Onen. "The speed certainly was surprising. Normally, we would not apply such extreme fields across devices, in order to not turn them into ash. But instead, protons ended up shuttling at immense speeds across the device stack, specifically a million times faster compared to what we had before. And this movement doesn't damage anything, thanks to the small size and low mass of protons. It is almost like teleporting."
From a first principles point of view, I would say that this is a staggering breakthrough, if they can demonstrate this out in public. Meaning, Cold Fusion was stunning until everyone figured out that the data was completely fudged. I wish there were more details on the actual chip design, including materials and manufacturing. To be clear, I am not brilliant. Me, and my team, spend our days and nights reviewing the work of some truly brilliant people and find flaws, failures, oversights and omissions. We are "detail people" who challenge every breakthrough not because we are jaded, cynical or even jealous. We do it to help the brilliant people learn to be even more brilliant in their designs and applications. I say all that so that I can admit that I do not understand where all of the ultimate applications for AI start and end. Further, without understanding the principles of the development, I cannot understand if what they have created can be reproduced, scaled and be able to provide widescale benefits to large numbers of people...or if it even needs to. I will say this, however. In 1988 I was part of a small group at UF, NC State and Purdue that produced some of the earliest operational FETs in GaN with electron mobilities above single digits. As of 2022, companies are just now starting to deploy GaN MIMIC in wide ranging applications from high powered radar, to radar jammers to base station transmitters. Meaning, time from first breakthrough to practical application is often measured in decades and not years, months or weeks.
Thanks. Does the fact that it is MIT reporting it and they noted that it could be easily incorporated into the existing chip making process mean anything? And it appears they already have plans to integrate it to use with AI.
Not likely. All wafer fabs are highly, highly specialized for the material systems that they work with. Silicon, Gallium Arsenide, Gallium Nitride, Lithium Tantalate (filters), etc.... Without any information on how the made these chips it is very difficult to understand whether their manufacture could be easily scaled up in a new production facility or not. Generally, most things created in a university lab are a long way from commercial, volume production.
About the same time you posted this I was listening to the BBC world service and a story on the passing of James Lovelock, the creator of the Gaia hypothesis. They replayed an old interview with him. In discussing artificial intelligence, Lovelock believed that computers would easily surpass humans because copper transmits signals faster than brain cells.
thanks. how does that reconcile with this statement from the development team? Are they just being overly optimistic The inorganic material is also easy to fit into current chip-building kit. here are other articles on it MIT Discovers New Material for Processors — Semiconductors Better Than Silicon? | Tech Times MIT Discovers Semiconductor That Can Perform Far Better Than Silicon (scitechdaily.com) hopefully these new build chip plants are able to take advantage of this significant advance in materials
So, here is the thing. There are currently no compound semiconductor, not GaAs, GaN or now the boron arsenide that can be built to the same scale as a silicon wafers. My current employer is a GaAs/GaN manufacturer, though we design many products in multi-chip modules that include Si chips in addition to the compound semiconductors, or standalone Si chip based devices as well. To build a boron-Arsenide based device, you would need to retro-fit a current fab, which is usually so expensive you just build a new one (which is also extremely expensive), more optimized for the new technology. There have been several material systems that are "beter" than Si through the years. I never seen even one ever put into production due to cost or engineering control issues. Now, that is all going to change soon because Si is at the 5 nm node - which means that is a gate length of about 10-12 atoms wide if the gate were linear. In other words, we are hitting the limits of Si technology, so for leading edge applications, something will have to come along and fill the gap. However, even then, most of the world's applications do not require cutting edge processing speed, so cost will dictate the use of Si for a long, long time to come.
Thanks, the whole thing is baffling to me that things can be built at that scale. hopefully some of the new chip factories being constructed will be able to incorporate the latest materials in the process of designing and building their new plants.
None of the new production coming on line will accomodate anything other than Si. Intel's $20B investment in Ohio (really $7B, US Tax payer is paying the rest), the multifab center in Texas that Samsung is preparing and the TSMC Mega-Fab in AZ are all Silicon based and designed for 300mm wafers and possibly scalable to 450mm wafers. No compound semiconductor material is manufactured on anything larger than 200mm wafers.
Should such capital intensive projects be designed to be adaptable for material advantages if SI capacity is reaching 100%? Is that possible? WRT the funding, it is my understanding that most of that is taxes that would not have been collected, it isnt payment from taxpayer to company. In real estate is it Tax Incremental Financing. TIF is also one way to help fund affordable housing.
No, fabs are not backward compatible in general (and I saw backward because new technologies often come out on smaller wafer sizes - tools designed for 12" wafers (300mm) are not compatible with smaller 6" or 8" wafers). You and I and the rest of the tax payers just paid for 75% of a brand new $12B Texas Instruments wafer fab in Richardson, TX. Hell, we should at least get a free calculator out of this!!! The Latest CHIP ACT that both the House and Senate have now approved set aside $39B for direct investment, ie not tax breaks, but direct investment in manufacturing facilities. Companies that receive this direct investment of tax payer dollars are ALSO eligible for a 25% "Investment Credit" on top of that.
Understood. We could either slap tariffs on chips made elsewhere or we could provide offsetting income to get critical infrastructure here. National defense..Russia is running out of chips for smart weapons and has no capacity to build their own. That is making a big difference in their ability to accurately target long range targets. We need that capacity onshore
the main article is re: using lasers to make zbits act differently. Lab technique with potential to helps stabilize quantum computing by creating a new form of matter. Buried deeper in the article is something I had not seen before. Makes one wonder if parallel dimensions are a real possibility New phase of matter could protect quantum computers against errors (msn.com) The new research comes just months after a series of studies into time crystals, a phase of matter which repeats in time, similar to how a regular crystal's structure repeats in space. What that means is that the particles in the crystal perpetually switch between two states without requiring the input of more energy and without losing any energy. These crystals are the first objects to break what is known as "time-translation symmetry," a rule in physics that says that a stable object will remain unchanged throughout time. Time crystals avoid this rule, being both stable and ever-changing. So, for example, ice when stable will remain ice and will only change when temperature or another factor makes it unstable. A time crystal would change even when in its ground state, acting differently than all other phases of matter. Last year, scientists from Stanford and the Max Planck Institute for Physics of Complex Systems, as well as scientists at QuTech, a collaboration between the Delft University of Technology and the Netherlands Organisation for Applied Scientific Research (TNO), figured out for the first time how to create these theoretical crystals. In June, scientists from Lancaster University, Royal Holloway London, Landau Institute and Aalto University in Helsinki succeeded in linking two-time crystals in a two-level quantum system, in which two independent quantum states were able to occupy both states simultaneously. .
