💡Machine intelligence in practice; the internet, addiction & regulation; privacy, encryption & security; self-driving races; tech v climate change; Barbie's eavesdropping++ #37
|Azeem Azhar||Nov 29, 2015|
A taxonomy of machine intelligence. Should we regulate the internet? The origins of privacy. Encryption and terrorists. The race to self-driving cars and self-driving car races. Women’s funny problem.
Not following me on Twitter? Just do it! 😀
Dept of the near future
💡Panopticons, lasers, alchemists, navigators & more. EV reader, Shivon Zilis, on a taxonomy for machine intelligence in the real world. EXCELLENT
📵 If the internet is so addictive, should we regulate it? “[Major tech companies] ‘have 100 of the smartest statisticians and computer scientists, who went to top schools, whose job it is to break your willpower.’” ESSENTIAL QUESTIONS
😎 The birth of privacy and beyond. A cultural, political and social history of the last 3000-years. **MUST READ. **
👽👾 “[A superintelligent machine] might be extraordinarily effective at achieving its goals might have unexpected negative side effects”. Nate Soares on the value learning problem. TECHNICAL, RECOMMENDED
🏇 Against innovation: reasons not to replace your horse with a car. **HISTORICAL **
👤Nadim Kobessei on encryption and terrorists: “Terrorism isn’t about means, but about ends. It’s not about the technology but about the anger, the ignorance that holds a firm grip over the actor’s mind.” EXCELLENT
📱Messaging apps are much bigger than social networks, and we’ve hardly begun. DATA RICH
Dept of climate change
You can’t have avoided the media coverage of climate change in the run-up to COP-21. I’m sure your favourite media has run great stories on it.
Two interesting takes for peripheral sources:
* 🔥Global growth in CO2 emissions seems to have slowed down. Most interestingly, CO2 growth and GDP growth have seemingly decoupled in the EU. This is a tremendously positive data point. Ultimately, we need GDP growth to be independent of carbon utilisation. (BTW, ignore the first sentence or so in this post, it is a bit puerile. But the rest of the blog post is a reasonable summary of the underlying research.)
* 💡Bret Victor, the UI designer behind the original iPad, asks “What can a technologist do about climate change?” It’s 12,000 words long, highly interactive and worth a read (but on a large screen device.)
Dept of privacy, encryption and security
🚨 The Economist’s special report on encryption and security is superb. ($⇥).
One key takeaway. Any kind of backdoor or weakening of encryption creates a massive financial risk, by making payment systems inherently weaker and susceptible to industrial hacking. Doubly dangerous in the face of the exponential growth of cybercrime. My take: not only might weakening encryption not make us safer, it would almost certainly make us poorer.
In any case, it isn’t clear that the security services are already making use of the signals intelligence and human intelligence they already receive. 👮🏽The Paris terrorists operated in plain sight, using their real names, mainstream travel websites and High St brands, as this exceptionally detailed WSJ report reveals.
Metadata, in particular, metadata which includes network interactions (who you phone, who they phone, and the overall properties of the calling network) can tell you a hell of a lot.
🔮 For example, just analyzing mobile phone records allowed researchers to accurately predict individual income, home and vehicle ownership and other attributes. (See also EV#24 on using Facebook likes to predict personality.)
KD Nuggets has a clear example of a trivial attribute prediction problem: guessing gender from social network profiles.
This also seems to be an area where governments can look for innovation in the tech industry. Germany tracks 420 potential extremists and France about 1,400 (with a similar number of officers to follow them). Facebook tracks 1.2bn people around the world in real-time, can predict how we are likely to behave and indeed trigger us to behave in particular ways. Would seem to suggest that cross-pollinating some of the software, processes and techniques from the internet ad industry to the terror threat might be a fruitful approach. (Of course, data scientist salaries in GCHQ/DSGE might not compete with those in Facebook.)
Dept of self-driving electric cars
An EV reader asked me why I cover self-driving cars so frequently. Actually, they complained that there was too much coverage of self-driving cars in EV. True to form, here is more.
Self-driving cars are the syzygy of progress in artificial intelligence, practical approaches to decarbonising and addressing climate change and rendering cities more pleasant for the 70% of humanity who will live in one by my 78th birthday.
🔌Tesla’s is making massive progress building a global electrification network for vehicles (with a 176% YonY growth of charging points.)
🌩Elon Musk has also signalled his desire to make autonomous driving happen even faster than the existing pace. If you are a super-solid AI/autonomous vehicle specialist, you might want to apply, as Elon will interview you personally. (Even Ford is now talking about a 4-yr time frame for fully-autonomous cars on US streets.)
I love this Google idea for its autonomous vehicles: how to signpost the car’s intent to pedestrians and other cars.
🚗Evolution-E will launch self-driving Formula E Grand Prix races next year. (Humans watching robots racing robots. All at 300km/h.)
Short morsels for dinner parties
🚀Jeff Bezos vs Elon Musk: a thrilling new space race.
🍓Moore’s Law is amazing. The £4 Raspberry Pi is as powerful as laptops from only 10 years ago.
AirBNB placed 250k Americans into accommodation this Thanksgiving, 100 times more than 5 years ago, representing a 150% CAGR.
🙋🏽The plight of the funny female: why do we seem to value humour in men more than in women.
The location of happiness is the precuneus, according to Japanese neuroscientists.
Peter is an AI assistant designed to help you with contracts and other legal paperwork. (Note: Peter, an AI-lawyer has a male name; the AI-bots that take on personal assistant tasks have female names, like Clara, Amy & Julie, by default. Go figure.)
Video: Watch a deep-Q network learn algorithms by example.
Is the new Barbie doll just a ruse to gather marketing data?
Li-fi is 100 times faster than wifi.
🇧🇪 Belgian physicists work out everyone is lying about the downed Russian jet.
🛩Military history nerd alert: the use of drones in the Vietnam war
This week’s Economist was a brilliant read for anyone interested in the content of this newsletter. Aside from the encryption special, there was a survey of climate change, a challenge to Christensen’s view of disruption, the end of the Internet unicorn & more. I felt I’d be short-changing my readers if I just linked to all of those. But if you don’t subscribe, treat yourself to a copy this week.
I started my new role thinking about internet-enabled marketplaces at Schibsted this week. It’s an exciting place and I’m looking to hire product managers, data scientists and developers, so hit me up by email.
Comments on this newsletter to me by email, as usual.