Microsoft acquires Nuance for $20B
As Brad and Aaron discussed Microsoft’s recent $20 billion (all-cash) purchase of Nuance Communications, I experienced a moment of speech recognition…recognition. The company name didn’t leap immediately to mind, but as they began discussing the company’s Dragon speech recognition software, it all came flooding back.
Eleven years ago, on my iPhone 4 (the second and last iPhone I ever owned), which at various times had both Dragon NaturallySpeaking speech recognition software on it, as well as the Siri app, before it was acquired by Apple and added to the iOS operating system. That’s a bit of a continuing theme for me: features that were once killer apps are now literally built into my keyboard and OS, and there’s really not much difference for me as a smartphone user between one platform and another.
Speech recognition works…well enough, most of the time, and has gradually improved over time, both on a platform and user-specific basis. When this all began to take off in 2010, the App Store walled garden meant a bit more to me in terms of functionality, and I didn’t know at the time (since it wasn’t made public until 2013) that Nuance had actually built the speech recognition engine used by Siri.
So, there is a bit of historical irony in the company being acquired by Microsoft now, for reasons that appear to be wholly unrelated to the consumer mobile ecosystem. Because being able to understand human speech is only part of the equation. The device, or the cloud computing that is operating behind the scenes, needs to be able to take meaningful action based on that speech, and that’s likely where a company like Microsoft comes in.
The biggest reason why Nuance didn’t spring to mind for me is because I don’t work in the healthcare industry, where the company has made significant inroads into clinical speech recognition and electronic health recordkeeping. Again, This is an area where there’s significant room to add some intelligence on the cloud end of things.
Machine learning in medicine
Around the same time the whole Nuance/Siri/Apple kerfuffle was happening, IBM’s Watson computer made headlines for beating Ken Jennings on Jeopardy!, and subsequently vanished into the healthcare industry as well. This was a computer that had the ability to both receive and “understand” natural language, and take meaningful action based upon it. In the intervening years, Watson was immediately put to work in a healthcare setting, driving everything from clinical decision support systems (CDSS) to providing treatment recommendations for lung cancer patients.
Going beyond Watson (which has expanded into other fields since), it’s fair to say that we owe a great many recent medical innovations to the advent of machine learning in medicine, including the “gigantic leap” in our understanding of protein-folding (one of the most basic functions in cellular biology) by Google-offshoot DeepMind, to the mRNA vaccines that are now being used to immunize people against COVID-19.
While Brad and Aaron spent a good chunk of time debating whether Microsoft had gotten a good price for such a massive acquisition, as someone who doesn’t hold any Microsoft stock, I don’t particularly care about this question, except to the extent that I’m hoping it means developing more and better tools to keep people healthy – which may well turn out to be the case.
The downside to less updates
Keeping computers healthy may turn out to be a simpler proposition by comparison – Brad mentioned that the Chromium-based Microsoft Edge browser has introduced an extended update cycle for enterprises – instead of updating every month along with the Chromium browser, they can now update every two months instead.
Microsoft seems to just be following Google’s lead here, with Chromium announcing both speedier and “Extended Stable” updates on cycles of 4 and 8 weeks respectively. Aaron noted, given the shared Chromium engine at the heart of Microsoft Edge (which is also used by Google Chrome, and nearly two dozen other browsers), any decision to delay updates, even in the interests of stability and continuity within an enterprise environment, “incurs a technical debt”, whether in the form of unfixed bugs or unpatched vulnerabilities. This is the fundamental tension at play between any solution which relies on a shared ecosystem (Brad and Aaron briefly returned to the cloud vs. on-prem infrastructure debate as well), and it obviously can’t ever be definitively resolved.
Further to this point, Brad cited a provocative example of on-prem infrastructure going very wrong. This was the Maersk cyberattack in 2017, which brought one of the world’s largest international shipping companies to a screeching halt. According to Wired’s Andy Greenberg (who literally wrote the book on this attack), the NotPetya attack began when a Maersk terminal in Ukraine was infected with ransomware by way of a financial application developed and used almost exclusively in that country. The malware is believed to have been sent by Sandworm, an APT hacker team within the GRU (Russia’s military intelligence branch) – with the specific objective of disrupting Ukrainian companies and government services (which it did, quite successfully).
Six alleged members of Sandworm were indicted for cybercrimes by the United States in October of 2020 (and will likely never see the inside of a US courtroom), but the damage was already done. As the recent incident with the Ever Given container ship has demonstrated, any disruption to global shipping can cause serious economic consequences.
This particular attack is doubly disturbing to me both because it used a software vendor as an attack vector – not for the last time – but because Maersk was, by all appearances, merely collateral damage. $300 million’ worth of collateral damage. And as Brad noted, Maersk was able to recover by pure happenstance – a data center in Ghana had happened to lose power when the attack occurred, so one of Maersk’s domain controllers was intact and offline. Through a series of person-to-person handoffs via Nigeria (made necessary by visa delays), the hard drive was able to make it to Maersk’s recovery command center outside London, and they were gradually able to rebuild their network. There is much, much more to this story, and if it strikes your fancy, be sure to check out Greenberg’s book from the link above, as well as Jack Rhysider’s riveting Darknet Diaries podcast episode about NotPetya.
Surface Laptop 4 preview
Brad and Aaron concluded with a bit more discussion about the new features of .NET 6 Preview and the Surface Laptop 4, with a healthy amount of skepticism about everything from cross-platform integration to battery life – with an expectation as ever that they’ll both deliver what they promise eventually.
And that’s it for April. Stay safe, get vaccinated, and look forward to doing some normal things again!