Meta, the tech giant previously known as Facebook, revealed Monday that it’s built one of the world’s fastest supercomputers, a behemoth called the Research SuperCluster, or RSC. With 6,080 graphics processing units packaged into 760 Nvidia A100 modules, it’s the fastest machine built for AI tasks, Chief Executive Mark Zuckerberg says.
That processing power is in the same league as the Perlmutter supercomputer, which uses more than 6,000 of the same Nvidia GPUs and currently ranks as the world’s fifth fastest supercomputer. And in a second phase, Meta plans to boost performance by a factor of 2.5 with an expansion to 16,000 GPUs this year. Read More
Daily Archives: January 24, 2022
Annual reflections of 2021 (Lillian Li)
2021 was an exceptional and rough year for Chinese tech. Things were fixed, but nothing was solved.
Rather than the protracted malaise that pervaded the faceoffs between tech giants and their proxies for much of the late 2010s, 2021 scattered the entire game board. With that, uncertainty defines the start of 2022. Boundaries have been redrawn, directions reset and organisations reformed. As China steers itself towards a new century where higher-quality growth is prioritised over fast growth, 2021 marked the end of an era.
The regulations introduced by the Chinese government in 2021 aimed to create balanced growth by unbalancing incumbent private actors while trying to stimulate future growth by altering the status quo. They’re eggs on a plate, but do they an omelette make?
Excuse my attempts to wax poetics, I never intended to become a Chinese policy analyst or put my development degree to use. But the 2020s is when we expect the unexpected, so let’s try to make sense of it all. Read More
Cryptocurrency Is a Giant Ponzi Scheme
Cryptocurrency is a scam.
All of it, full stop — not just the latest pump-and-dump “shitcoin” schemes, in which fraudsters hype a little-known cryptocurrency before dumping it in unison, or “rug pulls,” in which a new cryptocurrency’s developers abandon the project and run off with investor funds. All cryptocurrency and the industry as a whole are built atop market manipulation without which they could not exist at scale.
This should surprise no one who understands how cryptocurrency works. Blockchains are, at their core, simply append-only spreadsheets maintained across decentralized “peer-to-peer” networks, not unlike those used for torrenting pirated files. Just as torrents allow users to share files directly, cryptocurrency blockchains allow users to maintain a shared ledger of financial transactions without the need of a central server or managing authority. Users are thus able to make direct online transactions with one another as if they were trading cash.
This, we are told, is revolutionary. But making unmediated online transactions securely in a trustless environment in this way is not without costs. Cryptocurrency blockchains generally don’t allow previously verified transactions to be deleted or altered. The data is immutable. Updates are added by chaining a new “block” of transaction data to the chain of existing blocks.
But to ensure the integrity of the blockchain, the network needs a way to trust that new blocks are accurate. Popular cryptocurrencies like Bitcoin, Ethereum, and Dogecoin all employ a “proof of work” consensus method for verifying updates to the blockchain. Without getting overly technical, this mechanism allows blockchain users — known as “miners” in this context — to compete for the right to verify and add the next block by being the first to solve an incredibly complex math puzzle. Read More
Meta’s ‘data2vec’ is a step toward One Neural Network to Rule Them All
The race is on to create one neural network that can process multiple kinds of data — a more-general artificial intelligence that doesn’t discriminate about types of data but instead can crunch them all within the same basic structure.
The genre of multi-modality, as these neural networks are called, is seeing a flurry of activity in which different data, such as image, text, and speech audio, are passed through the same algorithm to produce a score on different tests such as image recognition, natural language understanding, or speech detection.
And these ambidextrous networks are racking up scores on benchmark tests of AI. The latest achievement is what’s called “data2vec,” developed by researchers at the AI division of Meta (parent of Facebook, Instagram, and WhatsApp).
The point, as Meta researcher Alexei Baevski, Wei-Ning Hsu, Qiantong Xu, Arun Babu, Jiatao Gu, and Michael Auli reveal in a blog post, is to approach something more like the general learning ability that the human mind seems to encompass. Read More
Rick's Cafe AI 