This is a weekly newsletter of what I've seen in tech and thought was interesting. I work at Andreessen Horowitz.

See more or subscribe at
🗞 News

Unless you've been a long, long way off the grid, you'll know that Facebook missed its earnings and the stock fell 20% or so, which is apparently the largest absolute decline in market cap ever ($120bn or so). This is a Rorschach Blot: you can see it as about privacy concerns, or Facebook deliberately dialing back superficial engagement for 'time well spent', or Facebook reaching saturation (with billions of users, after all), or people getting bored of the core Facebook experience, or... whatever narrative you prefer (and it will take a few more quarters to know which fits best). Facebook was not going to be a user-growth story for ever. But - do read the transcript. Link

Amazon, meanwhile, continues to violate the narrative by reporting growing profits. AWS is the most visible part of this, but I've argued that Amazon actually has many profitable businesses, with the income for them masked by newer loss-making startup businesses (that is, the company is a bundle). But it's interesting and arguably even perplexing that the company has decided to show more profit rather than reinvesting it, as it did in the past. Link

Google is now offering small, portable versions of its 'TPU' custom machine learning chips, embedded in boards or USB devices. Edge ML: embeddable in points of sale, kiosks or industrial machinery of all kinds, recognizing things that previously needed a mammal brain. Coming soon to Pixel phones, I suspect (note that the newest iPhones have dedicated ML silicon). Link

Qualcomm expects Apple to use cellular modems from 'a competitor' (= Intel) in the next iPhone. Big news in semis. Link

The ACLU did a publicity stunt in which it got Amazon's cloud facial recognition service to falsely identify US congressmen as criminals, by matching their pictures against a mugshot database. Amazon points out that machine learning is a probabilistic technology, that the ACLU set the tool to 80% confidence (not, say, 99%), and that the sample image set was biased, all of which is true but also, of course, the point: if you are not rigorous in thinking about what parameters you use and what bias might be in the data set, then you will get lots of inaccuracies. Machine learning is not magic, the computer can be wrong, and one should not take the results of any such system on trust. Equally, of course, claiming as the ACLU does that this is 'flawed and dangerous' is also to miss the point: it's a tool with probabilistic outcomes that you can use or mis-use, and more importantly understand or misunderstand.  Link

Google is going to roll out public wifi hotspots in Nigeria (following India, Indonesia, Thailand and Mexico). Link

🔮 Blog posts

Eloquent plea for less alarmist hand-waving blather about 'AI'. Link

There are spies in Silicon Valley. Lots of spies. Link

Jean-Louis Gassée - '50 years in tech'. Link

The political education of Silicon Valley. Link

UK Parliamentary report on fake news and Facebook (this isn't going away). Link

Bloomberg profile of the woman who actually runs SpaceX. Link

Fascinating profile of Gwyneth Paltrow's Goop. Link

😮 Cool things of the week

The General Magic movie. Link

📊 Statistics

Uber has provided 10bn cumulative rides. Link 

GSMA report on mobile in sub-Saharan Africa. Link

Online apparel sales are growing at 7% in the USA, and are now 21% of the total. Link
Like this? Subscribe!
Update your details
You signed up to this list at

Copyright © 2013-2018 Benedict Evans, All rights reserved. To stop receiving updates, you can unsubscribe. Is this email not displaying correctly? View it in a browser