SoylentNews

upstart writes:

A punctured bone fragment predates eyed needles in Western Europe by about 15,000 years:

An animal bone fragment full of human-made pits hints at how prehistoric people in Western Europe may have crafted clothing.

The nearly 40,000-year-old artifact probably served as a punch board for leatherwork, researchers report April 12 in Science Advances. They suggest that the bone fragment rested beneath animal hide while an artisan pricked holes in the material, possibly for seams. If so, it's the earliest-known tool of its kind and predates eyed needles in the region by about 15,000 years.

Found at an archaeological site south of Barcelona, the roughly 11-centimeter-long bone fragment contains 28 punctures scattered across one flat side, with 10 of them aligned and fairly evenly spaced.

The marks don't seem to have been a notation system or decoration, given that some holes are hard to see and the bone fragment wasn't otherwise shaped, says archaeologist Luc Doyon of the University of Bordeaux in France. He thought leatherwork could have made the marks. But it wasn't until he visited a cobbler shop and saw one of the artisan's tools that the hypothesis solidified, guiding Doyon's next steps.

[...] Scientists knew that humans wore clothing long before the oldest-known eyed needles existed (SN: 4/20/10). "What [the new finding] tells us is that the first modern humans who lived in Europe had the technology in their toolkit for making fitted clothes," Doyon says.

Journal Reference:
Luc Doyon, Thomas Faure, Montserrat Sanz, et al., A 39,600-year-old leather punch board from Canyars, Gavà, Spain [open], Sci. Adv., 12, 2023. (DOI: 10.1126/sciadv.adg0834)

Original Submission

hubie writes:

Ancient protein evidence shows milk consumption was a powerful cultural adaptation that stimulated human expansion onto the highland Tibetan Plateau:

The Tibetan Plateau, known as the "third pole", or "roof of the world", is one of the most inhospitable environments on Earth. While positive natural selection at several genomic loci enabled early Tibetans to better adapt to high elevations, obtaining sufficient food from the resource-poor highlands would have remained a challenge.

Now, a new study in the journal Science Advances reveals that dairy was a key component of early human diets on the Tibetan Plateau. The study reports ancient proteins from the dental calculus of 40 human individuals from 15 sites across the interior plateau.

[...] Ancient protein evidence indicates that dairy products were consumed by diverse populations, including females and males, adults and children, as well as individuals from both elite and non-elite burial contexts. Additionally, prehistoric Tibetan highlanders made use of the dairy products of goats, sheep, and possibly cattle and yak. Early pastoralists in western Tibet seem to have had a preference for goat milk.

"The adoption of dairy pastoralism helped to revolutionize people's ability to occupy much of the plateau, particularly the vast areas too extreme for crop cultivation," says Prof. Nicole Boivin, senior author of the study.

[...] "We were excited to observe an incredibly clear pattern," says Li Tang. "All our milk peptides came from ancient individuals in the western and northern steppes, where growing crops is extremely difficult. However, we did not detect any milk proteins from the southern-central and south-eastern valleys, where more farmable land is available."

quietus writes:

It looks like the Paris Agreement is as dead as the fried chicken at my local deli.

At Paris, in 2015, the World agreed to limit the global temperature rise to 1.5 degrees Celsius. The latest report of the EU's Climate Change Service shows (summary pdf) that this target has been royally breached, at least for Europe. Temperatures there, averaged over the last 5 years, have increased by 2.2 degrees celsius.

Europe, at least, has a climate change service to measure these things. As for the rest of the world, an extrapolation of the pattern shown in Figure 1c, here, indicates that, there too, demand for swimming pools and flood insurance will grow.

To illustrate the complexity of the problem, the heatwave in mid-July of 2022 was caused by hot air from the Sahara moving into Europe, driving temperatures above 40 degrees Celsius. By mid-August, a stationary high-pressure system with clear skies and weak winds took hold, and caused a second heatwave, which was made worse due to the soil being dried out by the mid-July event, and no rains since.

Events above the Sahara might have come a second time in play, here. Increasing temperatures lead to a stronger evaporation over sea, while the land heats up more. This results in a stronger temperature gradient, which draws rains deeper inland: heavier rainfalls now are reported in the central Sahara, in summer, with formerly dry valleys being put under four meters of water. This causes less Sahara dust in the atmosphere, and hence shields the land less from solar radiation: the EU's report mentions that 2022 surface solar radiation was the highest in a 40 year record, and part of a positive trend.

To end with a positive note, the EU ain't doing so bad, compared to Greenland: three different heatwaves in 2022, and an average September temperature more than 8 degrees Celsius higher than normal.

Original Submission

upstart writes:

The Fermi bubbles may have started life as jets of high-energy charged particles:

Bubbles of radiation billowing from the galactic center may have started as a stream of electrons and their antimatter counterparts, positrons, new observations suggest. An excess of positrons zipping past Earth suggests that the bubbles are the result of a burp from our galaxy's supermassive black hole after a meal millions of years ago.

For over a decade, scientists have known about bubbles of gas, or Fermi bubbles, extending above and below the Milky Way's center (SN: 11/9/10). Other observations have since spotted the bubbles in microwave radiation and X-rays (SN: 12/9/20). But astronomers still aren't quite sure how they formed.

A jet of high-energy electrons and positrons, emitted by the supermassive black hole in one big burst, could explain the bubbles' multi-wavelength light, physicist Ilias Cholis reported April 18 at the American Physical Society meeting.

In the initial burst, most of the particles would have been launched along jets aimed perpendicular to the galaxy's disk. As the particles interacted with other galactic matter, they would lose energy and cause the emission of different wavelengths of light.

Those jets would have been aimed away from Earth, so those particles can never be detected. But some of the particles could have escaped along the galactic disk, perpendicular to the bubbles, and end up passing Earth. "It could be that just now, some of those positrons are hitting us," says Cholis, of Oakland University in Rochester, Mich.

upstart writes:

U.S. government imposes record fine on Seagate for violating sanctions against Seagate:

Seagate has been hit with a massive $300 million fine by the U.S. Department of Commerce [PDF] for violating export control restrictions imposed on Huawei in 2020. The report shows that the U.S. Department of Commerce states that Seagate shipped millions of hard drives to Huawei in 2020 – 2021 and become the sole supplier of HDDs to the company while its rivals Toshiba and Western Digital refrained to work with the conglomerate.

Seagate shipped 7.4 million hard drives to Huawei on 429 occasions between August 2020 and September 2021 without obtaining an export license from the U.S. Department of Commerce's Bureau of Industry and Security. Those drives were worth around $1.104 billion back then, a significant sum for Seagate, which revenue totaled $10.681 billion in 2021.

To settle the matter, Seagate has agreed to pay the $300 million fine in quarterly instalments of $15 million over five years starting in October 2023. The civil penalty of $300 million is more than double the estimated net profits that Seagate made from the alleged illegal exports to or involving Huawei, according to BIS. In fact, $300 million is a record fine for BIS.

"Today's action is the consequence: the largest standalone administrative resolution in our agency's history," said Matthew S. Axelrod, Assistant Secretary for Export Enforcement. "This settlement is a clarion call about the need for companies to comply rigorously with BIS export rules, as our enforcement team works to ensure both our national security and a level playing field."

upstart writes:

The Hyena code is able to handle amounts of data that make GPT-style technology run out of memory and fail:

In a paper published in March, artificial intelligence (AI) scientists at Stanford University and Canada's MILA institute for AI proposed a technology that could be far more efficient than GPT-4 -- or anything like it -- at gobbling vast amounts of data and transforming it into an answer.

Known as Hyena, the technology is able to achieve equivalent accuracy on benchmark tests, such as question answering, while using a fraction of the computing power. In some instances, the Hyena code is able to handle amounts of text that make GPT-style technology simply run out of memory and fail.

"Our promising results at the sub-billion parameter scale suggest that attention may not be all we need," write the authors. That remark refers to the title of a landmark AI report of 2017, 'Attention is all you need'. In that paper, Google scientist Ashish Vaswani and colleagues introduced the world to Google's Transformer AI program. The transformer became the basis for every one of the recent large language models.

But the Transformer has a big flaw. It uses something called "attention," where the computer program takes the information in one group of symbols, such as words, and moves that information to a new group of symbols, such as the answer you see from ChatGPT, which is the output.

That attention operation -- the essential tool of all large language programs, including ChatGPT and GPT-4 -- has "quadratic" computational complexity (Wiki "time complexity" of computing). That complexity means the amount of time it takes for ChatGPT to produce an answer increases as the square of the amount of data it is fed as input.

At some point, if there is too much data -- too many words in the prompt, or too many strings of conversations over hours and hours of chatting with the program -- then either the program gets bogged down providing an answer, or it must be given more and more GPU chips to run faster and faster, leading to a surge in computing requirements.

In the new paper, 'Hyena Hierarchy: Towards Larger Convolutional Language Models', posted on the arXiv pre-print server, lead author Michael Poli of Stanford and his colleagues propose to replace the Transformer's attention function with something sub-quadratic, namely Hyena.

upstart writes:

Proposed emissions from a Mississippi Chevron plant could raise locals' cancer risk by 250,000x the acceptable level and a community group is fighting back:

We need climate action. But just because something gets grouped under the umbrella of things that theoretically combat climate change doesn't mean it's actually good for the planet or people. In an alarming example, production of certain alternative "climate-friendly" fuels could lead to dangerous, cancer-causing emissions.

A Chevron scheme to make new plastic-based fuels, approved by the Environmental Protection Agency, could carry a 1-in-4 lifetime cancer risk for residents near the company's refinery in Pascagoula, Mississippi. A February joint report from ProPublica and the Guardian brought the problem to light. Now, a community group is fighting back against the plan, suing the EPA for approving it in the first place, as first reported by ProPublica and the Guardian in a follow-up report on Tuesday.

Cherokee Concerned Citizens, an organization that represents a ~130 home subdivision less than two miles away from Chevron's Pascagoula refinery, filed its suit to the Washington D.C. Circuit Court of Appeals on April 7. The petition demands that the court review and re-visit the EPA's rubber stamp of the Chevron proposal.

[...] Last year, the EPA greenlit Chevron's plan to emit some unnamed, truly gnarly, cancer-causing chemicals at a refinery in Pascagoula. The approval fell under an effort described as fast tracking the review of "climate-friendly new chemicals." Chevron proposed turning plastics into novel fuels, and the EPA hopped on board, in accordance with a Biden Administration policy to prioritize developing replacements for standard fossil fuels.

By opting to "streamline the review" of certain alternative fuels, the agency wrote it could help "displace current, higher greenhouse gas emitting transportation fuels," in a January 2022 press release. But also, through that "streamlining," the EPA appears to have pushed aside some major concerns.

upstart writes:

Chinese tech giant claims better performance than competing GPUs:

Chinese social media, cloud, and entertainment giant Tencent on Monday revealed that it has started mass production of a home brew video transcoding accelerator.

The announcement comes nearly two years after the company unveiled a trio of custom chips designed to accelerate everything from streaming video to networking and artificial intelligence workloads.

In a post published on WeChat, Tencent Cloud revealed that "tens of thousands" of its Canghai chips, which are designed to offload video encode/decode for latency sensitive workloads, have been deployed internally to accelerate cloud gaming and live broadcasting.

Tencent says the Canghai chip can be paired with GPUs from a variety of vendors to support low-latency game streaming. When used for video transcoding, Tencent said a single node equipped with Canghai can deliver up to 1,024 video channels . We'll note that Nvidia, with the launch of its L4 GPUs last month, made similar claims. Without real-world benchmarks, it's hard to say how either firm's claims stack up.

[...] When it comes to spinning custom chips to improve the efficiency and economics of cloud computing, Amazon Web Services gets a lot of credit. The American e-tail giant and cloud titan has developed everything from custom CPUs, AI training and inference accelerators, and smartNICs to offload many housekeeping workloads.

And while Google has developed an accelerator of its own, called the Tensor Processing Unit (TPU), most US cloud providers have largely stuck with commercially available parts from the likes of Intel, AMD, Ampere, Broadcom, or Nvidia, rather than designing their own.

However, in China, custom chips appear to be more prevalent, with development an imperative accelerated by US sanctions that mean some tech products can't be exported to the Middle Kingdom.

Original Submission

upstart writes:

The company says it learned much from Terran-1's debut flight and is choosing to go bigger for its successor:

After its rocket failed to reach orbit last month, California-based Relativity Space doesn't want to dwell on the past. Instead, the company is leaping forward with its next launch vehicle, which promises to be bigger and better.

On Wednesday, Relativity Space announced its lessons learned from the launch of Terran-1, a 3D-printed, methane-fueled rocket that was set to break records on its first flight. The rocket took off from Cape Canaveral Space Force Station on March 22 but an engine failure prevented it from reaching orbit.

Shortly after its stage separation, the rocket engine did not reach full thrust, according to Relativity Space. The company shared key findings from the rocket anomaly, detailing that the engine's main valves opened slower than expected, preventing the propellant from reaching the thrust chamber in time.

Terran-1 is 85% 3D-printed by mass and it's also powered by a liquid methane-oxygen propellant known as methalox. [...]

[...] Unlike its predecessor, Terran-R is designed to be a much larger 3D printed, medium-to-heavy lift orbital launch vehicle capable of carrying 33.5 metric tons to orbit. The rocket's first stage will be outfitted with 13 3D-printed Aeon engines while its second stage will have a single methane-fueled engine.

mhajicek writes:

Potentially good news for old machinists and over-the-hill heavy metal fans:

"Five years ago, a team of researchers at the University of Rochester Medical Center (URMC) was able to regrow cochlear hair cells in mice for the first time. These hair cells are found in the cochlear region of ears in all mammals. They sense sound vibrations, convert those into brain signals, and eventually allow a person to hear and understand the different sounds around them. The new study from URMC researchers sheds light on the underlying mechanism that allowed the ear hairs to regrow in mice."

"We know from our previous work that expression of an active growth gene, called ERBB2, was able to activate the growth of new hair cells (in mammals), but we didn't fully understand why. This new study tells us how that activation is happening—a significant advance toward the ultimate goal of generating new cochlear hair cells in mammals," said Patricia White, one of the study authors and a neuroscience professor at URMC."

https://www.zmescience.com/science/news-science/can-we-reverse-hearing-loss-yes-we-can-here-is-how-it-works/

Original Submission

upstart writes:

Big tech companies were apparently hiring workers to keep them from joining rival firms:

Many former employees at big tech companies are admitting that they had very little to do at their jobs, despite earning high salaries. One such under-worked and overpaid former tech worker is 33-year-old Madelyn Machado, who left Microsoft to join Facebook's parent company Meta as a recruiter in the fall of 2021.

In a viral TikTok video, Machado claimed she was hired for a $190,000 yearly salary, but had basically nothing to do during her stint at the company. "I do think a lot of these companies wanted there to be work, but there wasn't enough," she said. Talking to The Wall Street Journal, Machado said that on most days, her work included attending virtual meetings from noon until 3:30 pm before logging off for the day.

Curiously, Machado says she was told by her recruiters at Meta that she wouldn't be hiring anybody during her first year at the company. She also claims that some of her colleagues told her that they had spent two years at the company without ever hiring anyone. Unfortunately for her, she only worked for six months at Meta before being fired last year for posting TikTok videos that the company said posed a conflict of interest.

upstart writes:

Inside the secret list of websites that make AI like ChatGPT sound smart:

AI chatbots have exploded in popularity over the past four months, stunning the public with their awesome abilities, from writing sophisticated term papers to holding unnervingly lucid conversations.

Chatbots cannot think like humans: They do not actually understand what they say. They can mimic human speech because the artificial intelligence that powers them has ingested a gargantuan amount of text, mostly scraped from the internet.

This text is the AI's mainsource of information about the world as it is being built, and it influences how it responds to users. If it aces the bar exam, for example, it's probably because its training data included thousands of LSAT practice sites.

Tech companies have grown secretive about what they feed the AI. So The Washington Post set out to analyze one of these data sets to fully reveal the types of proprietary, personal, and often offensive websites that go into an AI's training data.

To look inside this black box, we analyzed Google's C4 data set, a massive snapshot of the contents of 15 million websites that have been used to instruct some high-profile English-language AIs, called large language models, including Google's T5 and Facebook's LLaMA. (OpenAI does not disclose what datasets it uses to train the models backing its popular chatbot, ChatGPT)

The Post worked with researchers at the Allen Institute for AI on this investigation and categorized the websites using data from Similarweb, a web analytics company. About a third of the websites could not be categorized, mostly because they no longer appear on the internet. Those are not shown.

We then ranked the remaining 10 million websites based on how many "tokens" appeared from each in the data set. Tokens are small bits of text used to process disorganized information — typically a word or phrase.

upstart writes:

The Moon still has much to tell us about the early solar system:

The Moon still has much to tell us about the early solar system. Encouragingly, it also has scientific value as a platform for observational astronomy.

Lunar exploration is undergoing a renaissance. Dozens of missions, organised by multiple space agencies—and increasingly by commercial companies—are set to visit the Moon by the end of this decade. Most of these will involve small robotic spacecraft, but NASA's ambitious Artemis program, aims to return humans to the lunar surface by the middle of the decade.

[...] The potential role for astronomy of Earth's natural satellite was discussed at a Royal Society meeting earlier this year. The meeting itself had, in part, been sparked by the enhanced access to the lunar surface now in prospect. Several types of astronomy would benefit. The most obvious is radio astronomy, which can be conducted from the side of the Moon that always faces away from Earth—the far side.

The lunar far side is permanently shielded from the radio signals generated by humans on Earth. During the lunar night, it is also protected from the Sun. These characteristics make it probably the most "radio-quiet" location in the whole solar system as no other planet or moon has a side that permanently faces away from the Earth. It is therefore ideally suited for radio astronomy.

[...] Radio waves with wavelengths longer than about 15m are blocked by Earth's ionoshere. But radio waves at these wavelengths reach the Moon's surface unimpeded. For astronomy, this is the last unexplored region of the electromagnetic spectrum, and it is best studied from the lunar far side. Observations of the cosmos at these wavelengths come under the umbrella of "low frequency radio astronomy." These wavelengths are uniquely able to probe the structure of the early universe, especially the cosmic "dark ages," an era before the first galaxies formed.

Netflix Will Block Password Sharing Before July 2023

upstart writes:

Netflix Will Block Password Sharing Before July 2023:

Netflix has been working on a way to block people from sharing their Netflix passwords. It was supposed to roll out in the United States already, but now it's coming to the US and other regions sometime soon.

Netflix confirmed in its recent earnings report that it will start rolling out the new account sharing limitations in the second quarter of 2023 — meaning sometime between now and June 30. The company said in the report, "In Q1, we launched paid sharing in four countries and are pleased with the results. We are planning on a broad rollout, including in the US, in Q2."

In other countries where Netflix has already rolled out the changes, Netflix accounts have a "primary location" that is determined using your account history, home Wi-Fi network, and other data. Devices that aren't connected to that network and watching Netflix are automatically blocked after 31 days. The only way around the block is to add a paid "extra member" to your account, which costs less than an individual subscription, but isn't available for all types of Netflix plans.

upstart writes:

A European Chips Act to play catch-up with the US and Asia:

The European Union finally agreed on a new plan to boost its microchip industry. The multi-billion investment is focused on strengthening Europe's technological leadership, the EU said, but it could very well be an attempt to put the Old Continent on par with what market leaders are already doing right now.

After spending some months negotiating between the European Council and the European Parliament, the European Union has now officially approved a plentiful subsidy plan for its semiconductor industry. The European Chips Act will put €43 billion (roughly $47 billion) to bolster Europe's "competitiveness and resilience" in the microchip business, promoting an effective digital and green transition powered by hi-tech technology.

Right now, Europe has a 10% market share of global chip manufacturing; with the EU Chips Act, Brussels plans to double the EU's production capacity to 20% of the global market by 2030. The plan is also focused on strengthening Europe's research and technology capabilities over chip advancements, building innovation capacity in design manufacturing and packaging, developing an in-depth understanding of the global semiconductor supply chain, and addressing the skills shortage by attracting new talents and growing its own skilled workforce.

Microchips already are "strategic assets for key industrial value chains," the EU said, while the digital transformation opened new markets for the chip industry such as highly automated cars, cloud, IOT, connectivity, space, defense and supercomputers. The recent global semiconductor shortages also showed how the global supply chain has an "extreme" dependency on very few actors in a complex geopolitical context.

[...] As a matter of fact, the final EU Chips Act contains some additional provisions which were not included in the initial draft. Besides funding the manufacturing of cutting-edge semiconductor technology, the plan will also cover the entire value chain with older chips and research & design facilities. The EU Chips Act is coming after the world's powerhouses in the chip industry (USA, Taiwan, South Korea, Japan) have already approved or are in the process of approving their own subsidy initiatives. Therefore, Brussels' money to boost EU semiconductor output won't guarantee success.

Original Submission