hal đź‘ľ Profile picture
atm: genomics, automation, semiconductors and pandemics no need to qualify your replies, I assume good faith
Jan 27, 2021 • 5 tweets • 2 min read
don't understand why this infuriates folks. it is absolutely true that London is the most connected city on the planet, by far, and that this has made UK response to covid harder. surely this can be pointed out without it meaning one is "not holding the government to account" here is some reasonable data showing just how much more connected London is rome2rio.com/labs/global-co…
Jan 27, 2021 • 4 tweets • 1 min read
how many connectivity corps never launched because Loon existed, $s never invested in lowering the cost of net connections (debatable whether good or ill), because GOOG's market per trapped knowledge at X? And all for yrs of innovation with no market medium.com/loon-for-all/c… trusting companies with massive market power to deliver innovation (outside of war time) has never been a good idea. it isn't now and probably will never be
Jan 27, 2021 • 6 tweets • 2 min read
good, well sourced piece by @CaseyNewton. I find it strange though that, in talking about dangerous activity in private spaces, there is so little discussion of traditional, agent-style policing. The emphasis is all on corps to police themselves theverge.com/22249391/signa… what i mean by this, for instance, is tooling for police and other law enforcement groups to manage scaled infiltration of groups they suspect to be dangerous. one could also talk about adding abuse reporting functions, which neither iMessage nor signal have
Sep 10, 2020 • 6 tweets • 2 min read
some niche stuff for export control nerds that didn't make the piece. the most recent changes to the rules, on Aug 17th, the ones that have cut Huawei completely, came about after Qualcomm approached the the Department of Commerce, asking for a license to sell chips to Huawei The previous round of rules, published in May, had cut Huawei off from everything except chips designed and supported by teams outside the USA (but explicitly not those designed by HiSilicon). MediaTek, a big Taiwanese designer, was gonna be a big beneficiary
Apr 16, 2020 • 6 tweets • 1 min read
i wrote about contact tracing apps this week. their value, if any, is to optimise a large testing regime. more interesting is the massive sensor network Apple and Google can whip up just by agreeing on a protocol. it exposes their power. 3.5bn phones suddenly proximity tracking your phone feels like yours, and in many ways it is. but apple and google control it over long timescales, can and do shift the ecosystem to their advantage. i know this is obvious, but the covid bluetooth protocol highlights it for me
Apr 9, 2020 • 9 tweets • 2 min read
This week I wrote about what factories in China are doing to stay running during the pandemic. The increased digitisation, which is the bedrock for remote and autonomous operation, is likely here to stay after covid economist.com/briefing/2020/… Most of the automation in high-tech supply chains at the moment comes before “final assembly”. That is still done by large numbers of workers. Because companies like Apple need to produce millions of a new device every year, the flexibility of human labour is hard to beat
Mar 27, 2020 • 8 tweets • 2 min read
I am seeing a lot of anger about Palantir working on NHS data. That is fair, Palantir has a horrible reputation, and works on projects that run completely directly against my beliefs. But there are a few reasons to at least be open to a company like this wrangling NHS data: The first is that Palantir has huge experience handling sensitive data. If you are used to working with American intelligence, you have the security measures in place to keep NHS data secure
Feb 29, 2020 • 5 tweets • 1 min read
The data protection debate is stuck in a rut. We are all focused on data itself, arguing about who can collect what, its monetary value, what metaphors to use. but to me the most important factor is not data itself, but control of channels through which to record human behaviour in other words, we are all thinking about data like a commodity, but we should be thinking about it more like real estate. access to "surface area" through which to record valuable data is a zero-sum game. there are only so many sensors that can be deployed