🔮 Inequality; the artificial in AI; African leapfrogging; why blockchain is hard; botnets, corvids & ethics++ #166
|Azeem Azhar||May 20, 2018|
Dept of the near future
🌡️ Inequality is back. The 9.9% is the "new American aristocracy".Intergenerational wealth movement in the US is lower than almost every developed country. I’ve argued elsewhere that one curious side-effect of better mobility (digital and physical) can result in stronger trends towards assortative mating and homophily which may also result in socioeconomic gridlock. (Elsewhere, the new World Inequality Report notes that inequality is rising in nearly every part of the globe. Much of this is driven by greater privately-owned capital and the concentration of that ownership.)
🔒 On technology, politics and globalism: territorial data regulations are diverging with more governments requiring local storage of data. The number of nations implementing these intra fauces terra has tripled since 2016 to 84. Some estimates suggest it might reduce national GDP by 3.4%. (Perhaps unsurprisingly, authoritarian regimes fiddle their GDP growth statistics significantly, new satellite analysis suggests.)
🌍 Leapfrogging: Super survey of how the leapfrog model of innovation is helping socio-economic development in Africa. Can it scale?
Agile ethics: a manifesto for working ethically at speed. (Super interesting.)
🧠 AI guru one: Profile of Jurgen Schmidhuber, the less-well-known godfather of modern neural nets. (GREAT READ)
AI guru two: Gary Marcus on how much Google’s recent Duplex demo highlights the scale of research still required in the broad domain of artificial intelligence (and how deep learning alone won’t get us there.)
In order to make Exponential View viable over the long term, we need to grow the community. In fact, my models suggest I need to triple its size.
The only way to do this is with your support.
The best way you can do this is to forward this email to seven or eight friends/peers/contacts with a personal recommendation they subscribe and point them to the link http://www.exvw.co
The second best way is to write a short review on a blogging platform like Medium or LinkedIn and share it.
Please take a few moments to do that. (And please don’t rely on other readers to do this; this will only work if you help.)
Dept of the A means "artificial"
When Google demo'd Duplex, its new chatbot thing, I was pretty impressed with the technical complexity of what they had achieved. (Yes, some people argue that the dialogue may have been staged or, at least heavily edited and Gary Marcus, above, was cautious)
I was less impressed by the ethics of the demonstration. One, reading is just awkward. A multibillion-dollar firm’s CEO (comp. in 2016 $200m) proudly demonstrating a chat system “fooling” two presumably lowly-paid receptionists at high street stores. There was a power imbalance in that demonstration which set me a little on edge. Perhaps the demo should have been a surprise call to one of Google’s founders?
There are real problems with bots masquerading as people. We’ve seen clumsy bots on Twitter and Facebook confuse and befuddle people for close to a decade. I used to use Andrew/Amy, an impressive scheduling bot. But I stopped using it when I realised that humans who I cared about were spending their times crafting thoughtful messages to a script. Wasting their time and attention. (I’m not sure the code behind Amy now signifies that messages it generates are not from a human to prevent this, perhaps it does.)
I’m not an unalloyed fan of the rabid anthropomorphisation of today’s AI tools. These are tools, spreadsheets, hammers, flints, with a bit more verve and fairy-lights. They are impressive. They use better maths to act less deterministically than the dumb cogs and wheels of the past. They can deliver us some quite remarkable benefits, as long-term readers of this missive will know. But today, they are tools. They are improved by the application of scientific method and good engineering. We use words like “training” because those processes are analogous to the way we train conscious, biological entities. But it is only an analogy.
They do not have agency. Even the largest (non)self-promoters (looking at you, Sophia and Atlas) have no agency, nothing that resembles personhood. They are tools, crudely engineered toys from the vantage point of the 22nd century, however much they might impress we troglodytes of 2018. The golden lab puppy we all wished we had has more agency, a better sense of self.
So in the same vein that we shouldn’t be treating these portions of executable code as people, we shouldn’t be having them masquerade as people (except in certain therapeutic or specialist circumstances). During its demo, Google showed, once again, that the tech company had grasped with the ethical boundaries that it regularly butts up against. It crassly wandered over a line.
That is a line we might one day o'erleap. Science might help us strep over it by helping us deliver a better understanding of a system’s capacity for awareness, agency or consciousness. But it is not a line we need to cross today.
Vyacheslav Polonski's new research highlights how differing attitudes towards trust of AI could lead to a new digital divide: "refusing to accept the advantages offered by AI could place a large group of people at a serious disadvantage...[for example] differential access to job opportunities."
AI is creating an unprecedented demand for computational resources.OpenAI reports that since 2012, the amount of compute required to train the latest neural nets in research has been doubling every 3.5 months. We’re only at the beginning of this AI journey and over the next 10-years we’ll see these intelligent systems embedded everywhere, so demand for compute will continue to grow rapidly.
Why AI will move to the edge: Designers who want to build intelligent capabilities to the edge don’t have the luxury of relying on high-performance GPUs for computational requirements- they need computationally efficient systems that comply with strict power and footprint limitations. (See also: interview with the founder of Cerebras, a somewhat under-the-radar neo-chip company.)
Ocado: stunning profile of the British grocery delivery service which has amongst the most automated warehouses in the world.
Dept of blockchain
As blockchain projects mature, there is an increasing competition between the major platforms like Ethereum, Tezos, EOS, Dfinity et al to become one of the dominant platforms for the coming token economy. In the readable interview, Brian Behlendorf of the Hyperledger project, predicts a consolidation phase across these platforms, possibly imminently.
I'm not so sure it is imminent. Many of these projects are well-capitalised due to ICOs and are still early in their technical build phase, let alone their market development. We've seen how in early days of news markets many early leaders run out of steam during many phases of consolidation or shake out. (Think, for example, of CP/M, the Z80, GeNIE, Prodigy, etc)
In addition, these different implementations have fundamentally different attributes, such as their governance structures and degree of centralization vs decentralization. All of these result these platforms being more or less well-suited for particular types of applications. (A good discussion covering some of this challenges is in this critical view of whether blockchain will decentralise the web.)
💦 Why blockchain is hard: good analysis of the fundamental tension between centralisation and decentralisation.
The bitcoin network approaches the energy consumption of Ireland suggests new research and may hit 0.5% of all global energy consumption by the end of the year. (Lines up with one of my predictions for 2018.)
What you are up to
EV reader, John Gordon, is preparing a must-attend event with one of the pioneers of digital innovation, Jaron Lanier. If you're in London on June 14th, reserve your spot. Readers of EV get a special friends-only discount with code EXPONENTIAL.
Short morsels to make you smart
Profile of Jordan Petersen, the new prophet/profit of the old patriarchy and its entrenched status-relations.
Fascinating analysis of botnet business models. (A 10m-device botnet might cost $16m to set-up but could generate $20m per month from click fraud.)
🍺 Pubs and closing; gyms are opening. How Brits are swapping their pints of lager for litres of sweat. (Millenials, natch.)
WhatsApp is the platform of choice for fake news in India. Some users are subject to 1,000 messages a day.
Dockless bike-sharing appears to encourage multi-modal transit by city residents (at least according to research by Mobike, a dockless bike-sharing company.)
🕵️ Chinese high school students are being monitored for attentiveness using facial recognition.
Primes produce a diffraction pattern, like crystals and unique liquids. 🤓
Crows organize “funerals”, but not to mourn.
💸 UBI doesn’t cause inflation, as previously believed.
Is nature continuous or discrete? On the origins of the atomistic view of the Universe.
Jeremy Lent has written a very thought-provoking critique of Steven Pinker's Enlightenment Now. Long-read.
Please also take a moment to think about my request at the start of the newsletter this week.
We (collectively) need to build a larger audience in order to make this viable and higher value. This can only be done by personal recommendations. The things that will help the very most will be to forward this mail to at least 8 contacts with a recommendation they sign-up at http://exvw.co/