Don’t fight the AI might of Herbertsmithite
As 2021 came to a close, three breakthroughs across AI, astronomy and quantum computing caught Dick’s attention – but only one made him laugh
The human neuron appears to be several orders of magnitude more computationally complex than silicon ones used to emulate it
A house-sized mirror, made from hexagonal tiles of toxic beryllium, is folded up to fit into the rocket
One year ago I wrote about three technical breakthroughs that were announced in late 2020: an mRNA vaccine against Covid-19, Alphafold 2 solving the “proteinfolding problem” and Apple’s M1 chip. My picks for 2021’s trio of late breakthroughs are equally profound – one in neuroscience, one in space exploration and one in quantum computing – but they’re perhaps a little more difficult to explain.
The neuroscience discovery is at least simple to state: the human neuron appears to be several orders of magnitude more computationally complex than silicon ones used to emulate it. Given each neuron’s number of axon connections, this means the whole brain is almost infinitely more complex than any current AI. A team at the Hebrew University of Jerusalem trained an artificial deep neural network to mimic the computations of a simulated biological neuron, and the result suggests the AI net requires between five and eight “hidden layers” to represent the complexity of a single biological neuron.
Deep-learning networks have created an explosion in AI abilities such as driverless cars, Alexa, Google Translate and Deepfold. They all employ several layers – each containing a large number of silicon “neurons” – between their input and output layers, which when trained on millions of examples configure themselves into a box that classifies input into outputs. How many hidden layers are appropriate remains a keen matter of research: more isn’t always better, and the right number tends to be in single figures. Now it turns out that the classificatory power of a single human neuron is equivalent to a many-layered AI net. Yet our brain runs mostly on coffee and digestives, while training a large deep-learning AI net consumes so much power it’s becoming a threat to the climate.
At the other end of the scale, we’ve all thrilled to the colour photos of galaxies and nebulae captured by the Hubble Telescope. It’s so sensitive it can collect light from far-off objects that left not far after the Big Bang and has been travelling for billions of years since. Not close enough to the Big Bang, however, to answer some of the biggest questions in cosmology: thanks to an expanding universe, the oldest objects get red-shifted into the infrared, while the Hubble works with visible light. As I write we’re two weeks away from the launch of the James Webb Space Telescope, which works in the IR and if successful will push that vision back far enough to perhaps rewrite our story about the origins of the universe.
I say “if”, because the Webb uses the most ambitious space engineering ever. A house-sized mirror, made from hexagonal tiles of toxic beryllium, has been polished to the greatest smoothness ever achieved, t h e n f o l d ed u p t o f i t i n t o t h e rocket. In space it has to unfold correctly a n d each tile must be constantly realigned by computers. And it needs to work at close to 0K, so it must travel to the Lagrange Point, where Earth, sun and a moon’s gravity cancel out, then unfurl a vast, flimsy sunshade made of five layers of metallised film. It’s superb engineering, but the best of luck is still required.
Richard Feynman prophesied that one use for quantum computers would be simulating reality at quantum level, and in the final breakthrough I’d like to present a quantum simulator that has solved a real physics problem. A “quantum spin liquid” is magnetisable matter in a novel state in which it can’t take on a fixed polarity because its unusual threefold crystal symmetry makes domain pairing impossible. There’s a mineral, marvellously named Herbertsmithite, which may exhibit this state, but studying quantum entanglement inside a sample is impossible.
Simulating its lattice on a normal computer takes months, but teams at Harvard and MIT recently d i d i t i n reasonable time using a quantum computer that employs neutral atoms rather than the superconductors used by Google and IBM. It isn’t a generalpurpose computer: it builds a lattice of neutral atoms with the right symmetry using laser beams, both to place the atoms and to switch their “Rydberg states” to make them act as qubits. Neutral atom qubits are more robust than superconductors, and this success may refocus attention on the tech.
This last breakthrough tickles me for a different, less exalted reason. Cue Wikipedia: “[Herbertsmithite] is named after the mineralogist Herbert Smith (1872–1953)... [It] is generally found in and around Anarak, Iran, hence its other name, anarakite.” We all know kryptonite can counteract the powers of Superman, so perhaps anarakite might have a similar protective property against online computer super-nerds.
There’s nothing more valuable to online marketers than a pregnant woman, since she’s about to plonk a wodge of cash decorating a nursery, buying several thousand nappies and investing in entirely unnecessary products in the misguided hope of easing the early days of parenthood.
As I write this, I’m a handful of days from labour – and, thanks to my impossibly massive belly, sat several inches further from my desk than is strictly speaking comfortable – so I’m very much the target of such aggressive behavioural advertising.
Friends of mine have attempted to keep their pregnancies from the internet advertising machine; they’ve all failed. You could install antitracking widgets or use DuckDuckGo or enable Tor when searching for baby-related advice, but try doing that for nine months – and then ensure no-one in your family messages you over Facebook or Gmail about the incoming arrival.
I haven’t bothered trying to escape the clutches of ad networks, as pregnancy has made me excessively lazy, so I let it happen. That means Amazon, which I barely use, constantly queries my due date. Online ads have evolved from pregnancy tests to hypnobirthing classes to “baby boxes”, which I see as a scam to get you to hand over data for “free” samples of various nappies and creams. Instagram has repeatedly shown me ads for everything from preparatory massage oil for my nether regions to stem-cell banking services to sensory-toy collections.
Had I scrolled past such adverts, it might have discouraged Instagram from showing them again, as would tapping the buttons to hide or report the ad; I’ve done that for the most irritating posts. But I’ve also clicked on plenty – not only because I want to buy that product, but because it looks bonkers or I want to laugh at the hysterically high price. I’ve even clicked through to a product to screenshot it and send it to my friends or husband. I’ve forwarded plenty of posts to him followed by the “laughcry” emoji, or with text mocking the item and its price.
Of course, Instagram’s system isn’t quite savvy enough to understand my mockery. It sees only clicks, not hate-clicks. That means the businesses whose ads I’ve interacted with are repeatedly shown to me, a person who thinks their product is hilarious and the prices even funnier. No wonder their boxes of baby toys and weird creams and artisan reusable nappies cost so much, when they’re paying social media networks every time a hater like me casts an eye on their posts.
While this has been a source of amusement for me, it’s not so funny for others. Pregnancy is a sensitive subject and plenty don’t end as we’d hope. The infamous example is
Gillian Brockell, who in 2018 suffered a miscarriage but continued to be stalked online by baby-related products, with no way of stopping the upsetting, triggering ads.
Instagram does at least let you edit your ad preferences. Head into Settings, click Ads, and it will list the topic preferences it has discerned from your behaviour. Mine are: alcohol, parenting, pets and social issues. But while there’s an option to see fewer ads about such topics, it’s impossible to ban them.
You can also see your “ad activity”, which is the posts you’ve clicked on. This is interesting, not least because I don’t remember tapping some of them. They include a cocktail pop-up on the other side of London – I haven’t had a drink in over a year – and a sofa firm I know I have no interest in because
I’m typing this from our new sofa bed, purchased from a rival supplier.
I’m not saying Instagram is lying about my ad interactions. They may be accidental rather than hate-clicks; there’s something in the way my iPhone registers my fat fingers that means I mistakenly navigate in the wrong direction all the time.
Behavioural advertisers have been at this for years. I recall writing about these issues over a decade ago. Yet ad networks still can’t understand that a click doesn’t mean purchasing intent; it may be a mistake or mere mockery. Of course, for an ad platform that charges per click, it makes sense to ignore the indifferent and count every hit as a positive.
That’s why I’m not too bothered with Google, Facebook and the other networks thinking they know something about me. They know I’m pregnant, and from that they can reasonably guess that I may complete gestation and produce a child one of these days. But they still haven’t got a clue what’s going on inside my head, and I imagine they never will.
Instagram’s system isn’t quite savvy enough to understand my mockery. It only sees clicks, not hate-clicks
I haven’t bothered trying to escape the clutches of ad networks, as pregnancy has made me excessively lazy, so I let it happen