Tuesday, August 27, 2019

we are all mechanical turkeys

When Amazon setup the Mechanical Turk, they were making a play on the fake, chess playing robot, and of course, being ironic, in more ways than they realized (or less, given Ed Byrne's excellent observational humour). Like much of the gig economy, in this post-work world in which we find ourselves not living (despite labour saving devices reducing the need for farmers and builders and factory workers to a %age of what they once were), the mech turk represents exploitation by the owners of the means of production - the peoples' labour is only worth anything because of amazon's vast global platform. However, they (and uber and deliveroo etc) don't represent the pinnacle of the achievement of post-truth capitalism. That prize is won by the cloud services that give you nothing for your labour, your likes and reposts and retweets and the rest. Don't mistake that work for a social interaction with your friends/peers. Mostly, it is part of a system that is engineered to deliver things to you so you can train The Algorithm. The Algorithm is not an AI (artificial intelligence). It is simply a machine, learning from our behaviour. It isn't even a very clever machine. It doesn't typically learn to play chess. It just tunes what it delivers next time to everyone so it can earn advertising revenue for The Platform, when it manages to (fairly badly - e.g. 98% accurately) recommend stuff to people who could have just asked their friends.

Why doesn't it pay you for your labour? Because we are all mechanical turkeys. Like the legendary birds who were told about Thanksgiving Feasts being such a great social occasion for all the family, so fattening up, so they can be that feast, voting for Christmas too, we are the fodder. We are not the produce, we are the produce. We are the livestock, that fattens up the Softstock, to maximise shareholder value for the platform owners. It may look like chickenfeed to us, but it is a whole hill of beans to them.

Of course, the side effects are appalling. This is a rubbish way to do non-creative work, consuming vast amounts of energy running The Algorithm. I Think I shall refer to it from now on as the Artificial Dumber Downer. Take the collective intelligence of the human race, and create something less intelligent than the most stupid person in the population. And fool almost all the people, almost all the time while doing it. And take the money and walk to the bank, don't run.

Monday, August 19, 2019


Dave was really worried. the latest batch of SSD he'd received to put in the server farm was turning up way lower capacity than the last one. What was wrong? He zoomed some friends in friendly cloud providers, but found it hard to hear what they were saying. He ran some error diagnostics on the storage systems - weird, the bad blocks were all over the place, then the diagnostic tool crashed. He tried to reboot the SDN controller net and then tried a re-install. Nothing worked.

Luckily, he had an old AM radio tuned to a local rock classic station that played vinyl only. He tried to find a news station, but they all seemed to be down. so was all the internet. He went outside and got on his bike and cycled over to see some friends on stanford campus who have a clue. They were all hunched over a scope looking at the waves off of a probe.

"Hi Dave, look at this" they said pointing at the very wavy line on the old screen. "It looks like we can't see truth and falsity anymore" they quipped - indeed, it seemed as if the laws of physics had subtly altered so that electronics that distinguished "1" and "0" could no longer do so reliably.

"we're not in a binary world anymore, dave" said the trans in charge of the lab, semi-humourously.
"I wonder if this is because of those experiments Q was doing last week", suggested encrico.
"Q? the guy who's been pushing the bounds on eliminating de-coherence effects, probably" asked dave. "yeah, so we think that what her experiment actually does is to rapdily switch which universe we are all in so that the QC she's using is usually right. The problem is it now means that everything else is frequently wrong"

"Well that about wraps it up for computer science, doesn't it, surely?, asked dave.
"oh no - one quantum computer should be sufficient for everything we ever want to do again", quipped TJ. "so like Q - do or die". "no, god doesn't play die".

Saturday, August 03, 2019

polar bot extinction

Some smart folks at Warwick have done some very large scale analysis of digitized
historical texts and more specifically to this post here, polarization in political speeches and surveys showing perhaps unsurprising things like war correlates with falls in GDP and falls in happiness, but also that (at least in the US) there's a marked diversification of political opinion in the disappearing center of politics, while the "leaders" concentrate at the poles, the behaviour by the larger population is more akin to anarchy.

what this suggests is that, over a sequence of political tweets, one might expect a bot to take a consistent position on a bag of topics, but a human to show lack of adherence to a party (pole) position. a very light (LDA is probably overkill) approach to the set of tweets should see the divergence of real people from the political machine.....it would be interesting to compare this with more complex (e.g. deep learning) approaches.

One problem for the party machine is that they cannot significantly dilute or randomize the belief tropes they exhibit for fear of losing cult followers, so one should be able to filter (unfollow/mute, whatever) them permanently, fairly effectively - certainly, anecdotally, this is what i see in the twittersphere. e.g. in recent politics in the UK, accounts that are high probability bots (low lifetime, small follower cadre, auto-timing retweets etc) also bundle together beliefs (e.g. with brexit, goes "NHS health tourism" "the UK is too crowded with immigrants" and cliche/propaganda phrases like "take back control" and "soveriegnty" etc - so need to take each of these sources over some number of successive tweets and compare with known non-bots to see if real people are more diverse - i think they will turn out to be so. if it works for speeches, should work for microblogging and social media too...