In Brief If you’re wondering what it takes to develop a self-driving car, know that Tesla is using a 1.8-exaFLOP AI supercomputer packed with 5,760 GPUs that train neural networks it hopes one day will power autonomous vehicles.
The machine was described by the automaker’s senior director of AI, Andrej Karpathy, during an online academic computer vision conference this week. It is used to develop Tesla’s super-cruise-control system Autopilot, and also what could be a fully self-driving system when finished. Tesla has been chasing the autonomous vehicle dream for years; the tech has so far proved elusive.
“This is a really incredible supercomputer,” Karpathy said. “I actually believe that in terms of FLOPS, this is roughly the number five supercomputer in the world.”
It should be noted the prowess of a supercomputers are typically measured in FP64 precision. It’s not clear what precision Tesla’s 1.8 exaFLOPs of compute is running at.
It has 780 compute nodes, each containing up to eight Nvidia A100 80GB GPUs. The super also has 10PB of NVMe storage. Tesla’s AI models churn through millions of ten-second clips of driving footage recorded at 36 frames per second during training. “Computer vision is the bread and butter of what we do and enables Autopilot. For that to work, you need to train a massive neural network and experiment a lot,” Karpathy added. “That’s why we’ve invested a lot into the compute.”
Can AlphaFold help discover new drugs?
DeepMind has teamed up with a non-profit research org to use its AI protein-folding model AlphaFold to develop drugs that tackle parasitic diseases.
Specifically, the Drugs for Neglected Diseases Initiative will use DeepMind’s machine-learning software to figure out if new drugs can treat Chagas disease and Leishmaniasis. These are spread in tropical Latin America climates, and develop after people are bitten by triatomine bugs and sandflies. They can fatal if left untreated.
“AI can be a game-changer: by predicting protein structures for previously unsolvable protein structures, AlphaFold opens new research horizons,” said Ben Perry, DNDi discovery lead, according to the Beeb. “It is heartening to see powerful cutting-edge drug discovery technologies enabling work on some of the world’s most neglected diseases.”
Google Cloud AI tool to detect defects in goods
Google launched a machine-learning software tool on its cloud platform that customers in the manufacturing industry can use to automatically spot damaged products or packages during manufacture.
The Visual Inspection AI is designed to process images of items passing through assembly lines, and identify defective or problematic gear. Customers will have to fine-tune the model on the goods they sell and the kind of defects they want to identify. By using the tool, clients can better discard broken items or fix issues on products before they’re shipped, or at least that’s the hope.
“The benefit of a dedicated solution [like Visual Inspection AI] is that it basically gives you ease of deployment and the peace of mind of being able to run it on the shop floor,” Google Cloud’s Dominik Wee, managing director of manufacturing and industrial, told VentureBeat.
“It doesn’t have to run the cloud,” he added. “At the same time, it gives you the power of Google’s AI and analytics. What we’re basically trying to do is get the capability of AI at scale into the hands of manufacturers.” ®
In a countersuit released last week, Musk put his side of the argument. According to him: Twitter misled investors; it breached the agreement by failing to provide enough information on spam accounts; another breach occurred when Twitter failed to consult with him on business moves such as firing senior employees; and its misstatement of user numbers constitutes a material adverse effect, which substantially alters Twitter’s value and therefore invalidates the deal agreement.
“Instead, they contain numerous, material misrepresentations or omissions that distort Twitter’s value and caused the Musk parties to agree to acquire the company at an inflated price. Twitter’s complaint, filled with personal attacks against Musk and gaudy rhetoric more directed at a media audience than this court, is nothing more than an attempt to distract from these misrepresentations,” said the lawsuit.
Strong words, but Musk will need strong evidence as well to convince the judge.
Musk’s core argument is about user numbers
From the moment the deal started to go sour, the focus was on the veracity of Twitter’s numbers. It is at the centre of Musk’s countersuit as well. He argues that the number of monetisable daily average users (mDAUs) – authentic, active accounts that can see adverts (hence monetisable) – is falsely inflated by Twitter miscounting the number of false and spam accounts on the platform. As well as being a threat to the ad income on which Twitter depends, Musk said his plan to introduce a subscription service for Twitter would be affected because there would be fewer customers to target than first thought.
Twitter has consistently stated that it estimates the number of false or spam accounts on the platform to be less than 5% of its mDAUs base, which stands at just under 238 million currently.
The suit says that Musk became alarmed about how Twitter accounts for its mDAUs when, three days after signing the deal agreement, it admitted it had overstated its mDAU total for three years, by between 1.4 million and 1.9 million users per quarter. Twitter denies that the user change was a “restatement” (it describes the alteration as “updated values”) but admits it did not give the information to Musk prior to the deal being signed on 25 April.
Musk is not happy with Twitter’s verification processes
After agreeing to buy the business with minimal due diligence, the suit says Musk was “astonished” to learn about how “meagre” Twitter’s processes for identifying spam accounts were. It said 100 accounts a day were sampled by human reviewers in order to come up with the less-than-5% figure. Twitter’s CEO and chief financial officer were unable to explain how these accounts were selected to be a representative sample.
“Musk realised that, at best, Twitter’s reliance on and touting of its process was reckless; at worst, it was intentionally misleading,” says the suit.
Twitter argues that it uses a much more layered process for weeding out dodgy accounts, including using automated systems. It also pointed to the detailed explanations of how it polices spam accounts, which had been given to Musk, the press, the Securities and Exchange Commission and the public via a Twitter thread by CEO Parag Agrawal. In the most notorious episode of this takeover saga, Musk replied to the latter with a poo emoji.
But according to the countersuit at least Agrawal and Musk agreed on one thing. The document states that on 8 April Musk sent the CEO an example of a spam tweet saying: “I am so sick of stuff like this.” Agrawal replied, acknowledging “[w]e should be catching this.”
Citing “preliminary expert estimates”, the countersuit claims that in early July one-third of visible accounts may have been false or spam. This means that the true proportion of spam accounts among Twitter’s user base is at least 10%.
It says users that see zero or almost no ads account for almost all the growth in monetisable daily users. The majority of ads are served to less than 16 million users, the suit claims.
Twitter says that although not every user sees ads on a given day, in the first quarter “significantly more than” 229 million accounts contributed to Twitter’s average quarterly user number.
Regarding the 10% number, Twitter says it was based on a publicly available web tool, botometer, that has designated Musk’s own account as a likely bot.
Twitter made decisions without consulting Musk
One of the clauses in the merger agreement states Musk must be told when Twitter is deviating from its obligation to conduct its business in the “ordinary course”. In the countersuit, Musk claims that Twitter has made several “significant” changes – including firing two executives, starting a hiring freeze and initiating a legal clash with the Indian government – that occurred without his consent.
Twitter’s response is that axing employees or acting to protect users’ rights in foreign jurisdictions are part of the day-to-day business of running a company.
Information was not forthcoming
Musk is also claiming that Twitter failed to provide him with all the data and information that he requested “for any reasonable business purpose related to the consummation of the transaction”. The suit says Musk was sent reams of “stale data” that didn’t answer his questions.
It says, pointedly, that Twitter was happy to send data such as “a copy of its agreement with the Golden State Warriors for courtside basketball tickets and VIP parking”.
After more back-and-forth arguments over increasingly detailed information requests, the suit claims “the only conclusion the Musk parties could draw from Twitter’s obfuscation and delay was that Twitter knew that it had something to hide”.
The hyperscalers and public cloud providers are barreling ahead, unfazed by a rapidly deteriorating economic outlook, according to a recent Dell’Oro Group report.
In fact, these internet behemoths stand to benefit from the current market conditions in more ways than one, analyst Baron Fung told The Register.
As chipmakers like Intel, Nvidia, Micron, and others face increased pricing pressure across their lineups due to declining demand, hyperscalers are well-positioned to take advantage of this and add more capacity on the cheap, he explained.
“Looking at the recent Q2 earnings, it was really pretty impressive from a growth standpoint,” Fung said of the cloud providers.
Amazon and Azure in particular saw robust revenue gains in their most recent quarters. AWS saw revenues climb 36 percent from the prior year, while Microsoft reported its cloud biz saw year-over-year growth of 40 percent. However, things weren’t as peachy for Google, which saw a otherwise strong quarter for cloud revenue tempered by a $858 million loss in income.
Worsening macroeconomic factors may end up helping cloud providers as enterprises look for alternatives to capex-heavy server refreshes. We saw this phenomenon once before – in the early days of the pandemic.
These factors, combined with a wave of enabling technology – next-generation CPUs, GPUs, smartNICs, and CXL-enabled components to name a handful – will further accelerate hyperscaler spending, which is expected to grow 13 percent over the next five years, Fung said.
So it’s no surprise many chipmakers are optimistic about their cloud and datacenter-related revenues over the next few quarters, despite a slump in PC and gaming demand.
The analyst firm expects next-generation CPU platforms from the likes of Intel, AMD, and Ampere will be among the strongest drivers of hyperscale spending in the near term.
Intel and AMD are expected to launch their next-generation server processors later this year. Both of these chips pack a bevy of new features, including DDR5, and PCIe 5.0, in addition to having substantially higher core counts compared to the previous generation.
These chips are also among the first to support the CXL interconnect standard, “which will enable a new kind of paradigm in the datacenter,” according to Fung.
In its first iteration, the technology will allow systems builders to pack larger quantities of memory into servers than there are DIMM slots, using CXL memory-expansion modules. And in the years to come, the technology has provisions for tiered memory, memory pooling, and disaggregated compute architectures.
The operational and resource efficiencies enabled by the tech may eventually trickle down to customers in the form of lower prices, Fung added.
But it won’t just be the x86 stalwarts leading the charge in the datacenter. Fung also expects Arm chipmakers, like Ampere, to continue gaining traction in the hyperscale arena. Here, the chipmaker’s Altra and Altra Max processors have already attracted several high-profile customers including Microsoft Azure, Google, Cloudflare, and Oracle – to name just a few.
Finally, Dell’Oro predicts hyperscalers will drive edge infrastructure deployments – a market that Intel currently dominates – to 8 percent of the total datacenter infrastructure market by 2026. ®
The Irish-based study lead said food scientists, medical scientists and pharma companies must work together to produce functional foods to treat chronic conditions.
A team of researchers based at the Bernal Institute in University of Limerick (UL) have developed a new guide to designing functional foods to treat various chronic conditions.
Functional foods are foods that provide nutrition and act in a way that positively affects the body, similar to medicine.
According to the research, food has the potential to help in the treatment of heart diseases such as atherosclerosis.
“The capacity for our food to do more than provide us with nutrition is huge and relatively unexplored,” said study lead Daniel Granato, professor in food science and health at UL.
“Cardiovascular diseases are a main cause of death but they can be prevented. By bringing food scientists, medical scientists and pharma companies together we can employ the same methods used in producing medicinal drugs and produce foods that might mitigate health conditions,” Granato added.
The study has been published in Trends in Food Science & Technology, an academic journal. The UL researchers were joined on the project by academics from the Federal University of Alfenas and Universidade Federal de Minas Gerais in Brazil.
Granato and his team proposed an accurate computational approach to designing functional foods by predicting their bioactivity. This allowed the researchers to map how different food components benefit the body.
The study also drew attention to the potential of functional foods to treat illnesses and lessen the burden on the world’s health services. Functional foods are not too available on the market, despite their potential to help prevent conditions such as type-2 diabetes and glucose intolerance. These are both major contributors to heart disease.
Food science, cardiovascular disease therapy and computer modelling should be linked to produce functional foods that can mitigate atherosclerosis, according to Granato. He urged food and pharma companies to take note.
“This is critical to achieve United Nations Sustainable Development Goals in good health and wellbeing, as well as ensuring healthy lives and promoting wellbeing for all at all ages, by optimising discovery of bioactive compound sources, and reducing time to market for new functional foods,” he said.
Granato’s co-author and senior lecturer in the UL Department of Biological Sciences, Dr Andreas Grabrucker, said this approach could go far beyond heart disease.
“It will be the basis of a new research project at UL that aims to identify functional foods that lower the risk for neurodegenerative disorders such as Alzheimer’s disease,” he claimed.
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.