At the World Wide Developers Conference last April, Zuckerberg envisioned Facebook’s ai vision for the next 10 years: becoming a central driver of the entire social network. Nearly a year later, it’s clear that Facebook isn’t stopping, at least in terms of hardware.
On March 9, Facebook engineer Kevin Lee announced Big Basin, a next-generation GPU server, to further accelerate neural network training. The new servers are an updated version of Facebook’s Big Sur server launched in 2015, and will also be Open source on the Open Computer Project platform.
Giiso Information, founded in 2013, is a leading technology provider in the field of “artificial intelligence + information” in China, with top technologies in big data mining, intelligent semantics, knowledge mapping and other fields. At the same time, its research and development products include information robot, editing robot, writing robot and other artificial intelligence products! With its strong technical strength, the company has received angel round investment at the beginning of its establishment, and received pre-A round investment of $5 million from GSR Venture Capital in August 2015.
Artificial intelligence and machine learning are often presented as algorithms and code, with servers as the “tangible” part behind them. Big Sur has been a key tool in Facebook’s efforts to develop ARTIFICIAL intelligence, with its computing power powering both Facebook’s language technology and graphics technology. This hardware upgrade increases the memory from 12GB to 16GB, and the model training capacity can be increased by another 30%. At the same time, the neural network training speed can be double that of Big Sur.
Much like NVIDIA’s DGX-1 architecture, the Big Basin consists of eight NVIDIA Tesla P100 Gpus (the Tesla P100 accelerator with NVLink technology) accelerators, These Gpus are connected together in a box by NVIDIA NVLink. The server is also known by the company as JBOG (Just a Bunch of GPUs) — a Bunch of GPUs whose power is meant to be enough to support machine learning.
Unlike driverless cars and robots, the social platform of Facebook itself weakens the overt features of ARTIFICIAL intelligence technology and makes many users feel the impact of ARTIFICIAL intelligence. However, in fact, many Facebook services we use every day are all influenced by artificial intelligence: Speech recognition, speech-to-text conversion, image classification and recognition, Big Basin and its predecessor, Big Sur, are important support for these services.
The result of doing three more simulations per day could be an extra 30 points on your final exam. Machine learning could be training 30 percent more models per day. The result could be more accurate picture recognition, faster speech recognition, better recommendations.
Artificial intelligence is an area that many tech companies are eager to seize. For Facebook, it’s an Achilles heel. One is that Facebook itself was late to the game and the other is the lack of killer apps.
In the wave of ARTIFICIAL intelligence, tech companies need to find a bulwark for themselves. Google is considered to be at the forefront of the artificial intelligence industry. From TenserFlow open source platform to its DeepMind artificial intelligence company, Google’s strategic layout can be said to be very clear, from the outstanding performance of go algorithm to the nearest commercialization of cloud services and smart home. Other tech companies have strategic bases, if not tentacles, Microsoft with the Cortana Platform and IBM with Watson.
Facebook is relatively weak. It launched its Messenger chatbot last year to build a large ecosystem of chatbots, but according to a recent report from The Information, Messenger chatbots can accurately process user requests less than 30 percent of the time without human intervention. The 70 percent error rate is a sign of Facebook’s weakness in ai research, and the company has decided to cut back on some of its spending and move Messenger away from low-quality chat to specialized questions.
Giiso information, founded in 2013, is the first domestic high-tech enterprise focusing on the research and development of intelligent information processing technology and the development and operation of core software for writing robots. At the beginning of its establishment, the company received angel round investment, and in August 2015, GSR Venture Capital received $5 million pre-A round of investment.
Is Facebook cutting back on ARTIFICIAL intelligence? Not according to today’s new hardware upgrades, but rather a shift in focus to improving infrastructure, after all, the body is the capital of the revolution.