It turns out that today’s AI landscape is born out of fear, envy and fierce capitalist ambition. First, the emails were exposed in a Justice Department antitrust case against Google mentioned pretend Interested in businessMicrosoft executives are expressing their anger and jealousy over Google’s advances in artificial intelligence. That sparked the urgency that led the Windows maker to invest an initial $1 billion in its now essential partner OpenAI.
In a heavily edited 2019 email series titled “Thoughts on OpenAI,” Microsoft CEO Satya Nadella sent a lengthy letter from CTO Kevin Scott to CFO Amy Hood. “A very good email explaining why I want to do this… and why we ensure this to our infrastructure workers,” Nadella wrote.
Scott wrote that he was “deeply concerned” about Google’s rapidly growing AI capabilities. He says he initially rejected the company’s “exciting gameplay games,” perhaps a reference to Google’s AlphaGo models. One of them defeated world champion Go Ke Ji in 2017, an impressive feat at the time. (Later Google models went beyond this model and eliminated the need for human training altogether.)
But Scott said it was “a mistake” to ignore Google’s progress in gaming. “When they take over all the infrastructure they have built [natural language] “Models that we can’t easily reproduce start to take things more seriously,” Scott wrote. “When I dug deep to try to understand the performance gaps between Google and us when it came to modular training, I was very, very concerned.”
Scott talks about how Microsoft struggled to copy Google’s massive BERT model, the AI model that infers the meaning and context of words in a sentence. Scott attributed his competitor’s infrastructural advances, which Microsoft did not.
“It turned out that it was not easy for us to replicate the large BERT model. “Even though we had the model, it took about six months to train the model because our infrastructure wasn’t up to the task,” Microsoft’s CTO wrote: “Google already had BERT for at least six months, so they had the ability to train a 340 million parameter model. It took us a year to put together, a year to figure out how to produce it and push it forward into broader, more interesting models.
He also expressed his admiration and envy for Google’s autocomplete features in Gmail, saying they have become “shockingly good”. He noted that Microsoft “has fallen behind the competition in terms of several years.” [machine learning] Volume.” He commented on the “impressive” growth of OpenAI, DeepMind and Google Brain.
Scott praised Microsoft’s “very smart” people on its machine learning team, but said their ambitions were stymied. “But the deep learning core teams at each of these big teams are very small, and their ambitions are limited, which means that even if we start giving them resources, they still have to go through a learning process to scale their work. ” Scott wrote. “And in terms of machine learning scale, we’re several years behind the competition.”
After Hood insisted that Scott’s concern was “why do I want to do this,” i.e. investing in OpenAI, the company complied with its CEO’s wishes. Microsoft invested $1 billion in the startup led by Sam Altman in 2019 and the rest is a fast-paced story. (It has now invested $13 billion.) It’s a technology that’s doing incredible things, but at the same time it threatens to destroy the job market and gives propagandists their most powerful tool at a time when misinformation is already rampant.