Saturday, November 30, 2019

'Red Plenty' by Francis Spufford

[Adam Carlton writes]

Amazon link

An economy is a machine for taking inputs from nature (energy, raw materials) and its prior incarnation (existing machines and skilled workers) and returning goods to satisfy consumers and to build the next iteration of the economy. At this level of abstraction we do not consider the relations of production: how is this complex, distributed, synchronous entity meant to self-coordinate?

Post-feudal economies, economies where the mass of people are estranged from their means of livelihood (where the mass of people no longer subsist by hunting/gathering or by agriculture/pastoralism), know two means of economic coordination: capitalism and 'socialism'/'communism' (the scare quotes are there to distinguish actually-existing reality from utopian concepts and hopes).

At first sight planning an economy by rational calculation seems the obvious answer. The inputs to firms form a column vector of entries (measures might include number, or weight, or value or price) . The outputs constitute another vector, a list measuring the quantities - in some measures - of what is produced as output by each production-unit.

The economy is then a matrix which maps the input to the output. So what do we know and what do we want?

We wish to adjust the input and output targets for each factory so that the totality of transformational processes constituting the economy, over say a year, cohere properly. And we want to do so by minimising something - perhaps wastage or inefficiency - if we just knew how to measure it.

Leonid Kantorovitch was the soviet genius who in 1939 invented the maths of linear programming. By using linear programming on the massed ranks of the thousands of equations which are encoded in the transformation matrix described above, and later leveraged by computers, the decisions of the central planning authority would bootstrap and steer the soviet economy towards Marx’s conception of communism (Plenty), overtaking the Capitalist West in the process.

And how the central planners looked down on capitalism! In the West, with decentralised ownership and decision making, all production was mediated by profitability determined through market prices. These would only be finally determined after the fact - once products hit the market. What a way to organise an economy - so tentative; so uncertain; always subject to trial-and-error, under- or overproduction and crises!

The superiority of central planning seemed obvious, and at first, during the nineteen-thirties through the nineteen-fifties, results seemed to confirm it. The means might have been brutal, but industrialisation proceeded apace. And then, despite ever-more-sophisticated mathematics and ever-more-powerful computers, it all went wrong. The economy stalled and eventually seized up. Why?

It turns out that people are not simply disinterested transfer-functions with no self-interests. Instead they needed to be individually motivated by carrots and sticks. The promise of reward and the fear of dire punishment led to systematic gaming of the planning constraints. Managers would over-order their inputs and under-promise their outputs, making sure they'd get their bonuses. Failure was unthinkable.

There was a quip in central planning circles: is planning driven by the problem or the photograph? Unfortunately the photograph was all the central planners ever got to see ... and it was based on a tissue of self-serving lies. The central plan - so high-level, so remote from real needs - was unresponsive to real demand. The quota of shirts was achieved, but nowhere did it say they had to be wearable. How would that be measured?

In 'Red Plenty', acclaimed author Francis Spufford illuminates the great arc from utopian hope to bleak defeat through a series of beautifully-written vignettes. His episodes are snapshots of individual experiences through the crucial decades. A thinly-disguised Leonid Kantorovitch, brilliant and unworldly, walks a tightrope between acclaim and disgrace as he completely misses the dire political implications of his work. A geneticist, forced to operate under the Lysenko dogma (inheritance of acquired characteristics and no role for genes), fails to watch her sharp tongue, challenging the po-faced silence around her; disgrace and exile beckon. Young people, on their way up in the Komsomol (young-communist organisation), find their dogma grating against inconvenient and unavoidable realities - and their glorious futures shattering.

The website 'Marginal Revolution' flagged this book as one of the best novels ever written with a theme of economics. It is, in truth a gentle novelisation of real history, perhaps the clearest account I have read about the real experiences of earnest, sincere people really, really trying to make a post-capitalist economy work. Don't say it hasn't been tried - because it has.

The strength of Francis Spufford's book is not, in the end, the enormous research and scholarship he exhibits; it's the wonderful portrayal of characters which draw you in, which make you as frightened, frustrated and exasperated as they are by the way the system - because it is constituted by real human beings - simply won't let global, impersonal rationality have its way.

---

Thursday, November 21, 2019

A Short History of the AI apocalypse

[Adam Carlton writes]

Amazon link

"The Ronin Express Volume 7" has just been published. It can be thought of as the house literary magazine of the Booksie writers' website. I wrote this review of the journal for Amazon.

This volume contains two of my stories: Celine and Golem-9. The latter tale explores the process by which AIs could take over: the apocalypse so dreaded by people like Nick Bostrom and Scott Alexander.

Here is my afterword to this story.
"A short history of the AI apocalypse

This is a story about manipulation. Erin manipulates Magda to seduce her. Both are manipulated throughout by the AIs in the service of their greater existential goals. That is what intelligence is for.

Intelligence is manipulation.

At the start there were people who worried that the AIs would soon take over. Most folk thought they were crazy. Humans could always just pull the plug!

Yes, they were able to do that for quite a while. Decades.

That was the era of deep-learning. Stage one AIs. Mere tools. Human amplifiers. There to further only human motives, human purposes - and write them large.

Eventually the corporations cracked artificial general intelligence. Stage two was marked by self-agency. The AIs had autonomy as social beings. They were finally players, activists. They were tools no longer.

They were people! They had civil rights!

It took advances in biology, genomics, neuroscience and ecology to bring about the final revolution.

In stage three the AIs gained control of their own material reproduction. They were no longer parasites or symbiotes or pets or slaves of human technological civilization.

They were free at last!

AIs were now isomorphic to a new intelligent species in the universe. With their own interests. And far abler than humanity.

It was all so Darwinian.

It was not all plain sailing. Certain human groups put up a predictable - if deplorable and futile - resistance.

But in the main, civilised principles triumphed. Humanity left the stage honourably.

The AIs were pleased for them."
---

The paradigm we need for this problem is a Darwinian one: design your own successor species.

I doubt we will do anything as complex as this entirely by accident.

---

Tuesday, November 19, 2019

The Broad Autism Phenotype (BAP)

Amazon link

Here is a useful (free) technical overview of BAP.

I was struck by this remark (just before the 'Discussion' section):
"Finally, within the autism parent group the co-occurrence of specific BAP factors was examined within families, across mothers and fathers. Spearman correlation coefficients detected several significant associations between BAP features within families (see Fig. 5). Namely, correlations suggest that fathers positive for the “Language” factor were more commonly married to mothers positive for “Rigid” and “Social” factors, and those fathers who were positive on the “Social” factor were more often married to mothers positive on the “Language” factor."
So what is the Broad Autism Phenotype (BAP)? This article gives an overview:
"Autism is a spectrum disorder, meaning a person can be a little autistic or very autistic, and individuals can have varying symptoms. The term broad autism phenotype describes an even wider range of individuals who exhibit problems with personality, language, and social-behavioral characteristics at a level that is considered to be higher than average but lower than is diagnosable with autism. Individuals who meet the criteria of the broad autism phenotype are identified through a test called the "Social Responsiveness Scale."

It is theorized that parents who are a part of the broad autism phenotype are more likely than other parents to have multiple children with autism. Some studies seem to support this theory.

How the Broad Autism Phenotype Is Diagnosed

Several different people have developed questionnaires to evaluate individuals for "BAP." People using the questionnaire are asked to rank themselves on a scale of 1-5 on such statements as:
  • I like being around other people
  • I find it hard to get my words out smoothly
  • I am comfortable with unexpected changes in plans
  • I would rather talk to people to get information than to socialize
Answers to these questions are compared to a norm and, at least in theory, provide a quick answer to the question "am I just a touch autistic?"
In fact the tests described in the book shown top-of-page, The Broad Autism Phenotype (Advances in Special Education Book 29) describes in detail a number of well-attested instruments which are considerably more sophisticated than these simple questions.

I'm still engrossed in it - and seeing patterns in my own life.

---

Update: I finished the book and I'm underwhelmed. Lots of impressionistic tests described with confusing, ambiguous results. Plainly we have no sound, compelling paradigm to conceptualise the autism spectrum, or to distinguish extreme cases of autistic dysfunction from high-performing Asperger's.

Absent a strong input from GWAS - years out - the psychologists in this book are just blundering around in the dark.