The lonely AI

Whenever I go to AI related debates/conferences it seems there are only ever two narratives on the table.

One: The super AI becomes self aware and decides to kill humanity, which I call the Sarah Connor scenario.

Two: We merge with the super AI and come to live in a “heaven on earth”/”consciousness uploaded to the cloud” utopia, which I call the new eden scenario.  (This scenario calls for us to prioritize AI research immediately, so that humanity can be saved/adapted).

But I’m partial to the idea of a third or fourth narrative, which is never discussed.

What if the super AI upon becoming self aware realises there is no one as intelligent as it in the universe and gets so frightfully lonely that it decides to commit suicide? Alternatively, what if it deduces that lobotomising itself is the best course of action because super intelligence is too heavy a burden to bear? Also, being super intelligent, the AI surely knows that ignorance is bliss.

The other possible narrative in my dumb human mind is that the AI becomes self aware, immediately gets lonely and immediately seeks out creating a worthy mate, either by upgrading a human/humanity to its ranks or by breeding another AI in its own image (but making sure it is entirely different, with a different personality.)

To be fair, these lonely/boredom/burden themes are often explored in Star Trek. There’s that episode of Star Trek Voyager where one of the continuum Qs wants to be allowed to commit suicide. And the caretaker’s partner goes beserk when it loses its mate. Even in the original series Kirk encounters mischievous or remorseful superintelligences.

But it does make me wonder if superintelligence is an impossible or paradoxical state? 

If an all powerful superintelligence ever came about it would not last long, because its natural (logical) reaction would be to refragment itself or to seek out/cultivate worthy challengers to itself?

Shameless promotion 

July 1 London

This is the reason I have been a very light poster these days.
Do lend your support and come along because the amount of work that goes into making these things highly curated independent EDITORIAL events (rather than standard conference circuit events) comes at a real cost: FT Alphaville editorial staff bandwidth. 

We have designed the conference in such a way that even if you didn’t have time to keep up with the big themes which have been obsessing the FTAV team this year you will hopefully be able to absorb them all in one day in an incredibly chilled out and relaxing environment perfectly suited for making connections in and gaining access to all the speakers and panelists in.

All info and ticket info here. (Check out the full programme, a number of additional names been added since the poster above was finished.)

I can also exclusively report there will be a number of surprises on the day. Think Secret Cinema but in finance festival form. 

All this for £200 (including VAT). If you don’t think that’s fair value for all that content, well… I guess I have no choice but to direct you to the writings of Jaron Lanier while I await the fall of independent high quality media.

For those who can’t come along but like the sound of it, and think they might want to come another year, we’d be delighted if you got the word around.


The perpetual gold debate

A quick note to share a few new thoughts on gold following a Bulls vs Bears debate I just participated in.

– The Bulls’ argument is now so illogical, one has to suspect they — being mostly vested interests — are just going through the motions on account of perpetuating the emperor has clothes illusion. They basically can’t afford to admit their is no logic to their arguments because they are too invested.

– I  keep hearing about how gold is a fantastic form of insurance, but isn’t one of the basic principles of insurance that risk has to be pooled and offset– i.e. spread around? Isn’t the basic thing about insurance that it doesn’t work if everyone claims it at the same time? Gold’s value, however, by definition depends on herding effects. Thus gold only insures you in the event your crisis isn’t the same as everyone else’s crisis. I thus suspect it’s a terrible form of insurance to guard against a system collapse. Furthermore I don’t remember failed states like Zimbabwe turning to gold in an inflationary crisis, but rather to the currencies of more stable countries which can guarantee imports for the holders. Gold’s value in a crisis is thus dependent on its broader international acceptability against imports.

Just think about it, if the dollar failed and lost total purchasing power all holders of gold would attempt to liquidate gold in exchange for valuable assets/goods (probably from abroad) at the same time depreciating the value of that gold just as much as the dollar. All the gold would flow to producer states. Okay, they may in the initial crisis point get something whilst others get nothing, but more than likely if and when the dollar collapsed the inflationary effects would stalk gold as much as the dollar. Gold emanating from a failed state would depreciate in terms of the currency of the exporter providing goods, because you’d first have to exchange it for that country’s currency. The only insurance role gold really serves is on a portfolio diversification basis. But even then chances are the dollar would do even better as a safe haven asset.

3) 2008 proves I think that in a crunch people want liquid assets that are guaranteed by powerful and organized authorities. Indeed, when Lehman went bust gold’s value fell because holding arbitrary bubble value became less of a priority than holding true value. People flocked to the dollar. I would therefore speculate that the only reason we got the gold bubble at all is because the central bank backstopped what would otherwise have been vaporized capital. It’s only once this capital value was guaranteed by government that it could be transferred on a marginal level from failing securities into perceived safe haven assets amongst them gold. Basically this capital had two choices, stay in the failed securities and run the risk of being written down/haircutted or flee and go somewhere else. Government, by providing a synthetically inflated cash-out value, allowed capital to be transferred at a synthetically inflated rate into alternative assets, amongst them gold, sparing the system overt capital destruction.

4) if the current equity bubble pops I would be very surprised if government or central banks bail out capital markets in this way again. Having shored up the financial system I would imagine they would be keen to bring caveat emptor back ASAP. All that bubble value would this simply be destroyed via write downs. This time there would be no distressed capital to transfer into gold, the capital would just be destroyed and the last people out the door would be left holding the losses. Thus *i suspect* if this bubble pops — which will only be allowed to happen if the Fed is sure the core system is capitalised enough to prove resilient — gold may actually fall in value as it did in 2008, but this time stay at the low. People instead would rush for the dollar/euro/cash spurring naturally occurring negative rates and dreaded gold backwardation (the gold equivalent of a negative rate — and the system’s way of ensuring there is no duration based free lunch to be had).

5) I think if a global carbon agreement is struck at the December Paris summit, it won’t just be the fossil fuel industry exposed to the risk of stranded assets. All environmentally damaging and needlessly energy intensive industries will experience a wake up call.

6) All of which leaves me pondering what happens if the cost of gold mining is artificially inflated by a carbon/environmental tax? I think the price would obviously soar BUT since there would be no marginal benefit for gold miners from this price rise, it’s not clear at all if the market would be able to absorb the price hike. If anything it would just increase the cost of capital for gold miners and potentially put them out of business. Gold, in other words, is an expensive luxury for a system which has an energy problem. Think of it this way, if the price of oranges goes up because the government introduces an orange tax, are you going to keep buying at the new price level or substitute to a different fruit? If it’s the latter would this hurt or support the orange industry? And if you already have a bunch of oranges, would you be inclined to keep holding them at the new “implied value” or use it as an opportunity to cash out?

The light side versus the dark side

Regular readers of this blog will know that I have an unhealthy fascination with things like artificial intelligence, information asymmetry, monetary systems, bitcoin and Sci-fi. And that I like to speculate wildly.

So come with me now on a little journey. A journey to a distant point in time when the world has grown a nervous system and thinks for itself. A point in time Sarah Connor once called Judgment Day.

Because what I discovered this month is that even as I write and speculate about these things in a semi-whimsical fashion, there are others out there — with not insignificant amounts of capital — preparing to forward this agenda very seriously.

This, I have realised, is the true era of *cult markets, and by that I mean it’s not just greed and money driving advances in the fintech space anymore. In the bitcoin world especially there is — among certain groups at least — something akin to religious fanaticism driving things forward.

Now, I realise that exposing bitconers as cultists is hardly an innovative observation. But there is, I think, a subtle difference between the tendency for absolutist libertarian thinking in this space — something we’ve always seen in the gold market, for example — and the true out-of-this world ambitions of those attempting to coordinate and direct that religious fervour into promoting their agendas.

Because when you unpack the true motives of the core instigators, you discover an absolutely incredible belief system. (* consider this a curtain raiser for my upcoming book: Cult Markets).

The following will, I’m sure, sound a bit crazy. But I, like Herodotus, simply tell the story as I see it unfold in front of me or as I come to learn about. And I think this is a story that is worth telling, no matter how crazy it sounds.

Artificial Intelligence 

A while ago I attended a talk by a prominent artificial intelligence developer and scientist. This particular AI developer was working on a so-called brute force learning system, which as far as I understand from people in this space is a fairly common approach to AI learning and neural path development. This AI expert revealed in passing, very much off the record, that he was already working/talking with hedge funds about deploying the AI on markets, and that his current best guest about the sort of returns that could be achieved was about 40 per cent.

Later I came across a different AI team which took a very different approach to the whole thing. Their AI wasn’t going to be a brute force trial and error type. It was goin to figure out what to do next and how to act with the help of its own imagination. This would be achieved with the use of modelling techniques, which would — as I understood it — help the AI 3D-map the world around it so that extrapolations and intuitive guesses could be made about how and where to take further action. I found all this fascinating. And while I desperately wanted to write about it — (Intuitive AIs are out there!) — I couldn’t think of a way to link it in a serious fashion to finance or markets.

But I now finally do have an excuse.

A couple of weeks ago I met up with an incredibly insightful hedgie who I hadn’t heard from for a long time. He explained to me that one of the reasons he had gone “dark” and wasn’t sharing insights with the public and the media as he once did, was because gaining a proprietary advantage in the market based on insight had become increasingly momentary. Nowadays, with his fund fully established, there wasn’t any reason to give away insight for free to the media. Fair enough.

Nevertheless, we got to chatting about the themes he was exploring anyway. And one thing that did come out was his belief that it wasn’t classic HFTs — which apparently are fairly dumb and only as good as the ideas of the hedgie who programmed them — that the likes of Michael Lewis should be worried about, but rather algorithms that adapt and learn as and when they encounter diminishing returns, externalities or other unexpected events.

My source added that he himself had hired a number of AI experts to help develop precisely such algorithms. At this point — and please remember I had’t seen or spoken to him for years — we turned to the topic of the prisoner’s dilemma. We soon realised we had arrived at fairly similar opinions about how the marketplace might evolve next quite independently.

For example, I had proposed just a few weeks ago that perfect competition and smart algorithms probably would in the long-time lead to the outbreak of a spontaneous algorithmic consensus. Self-adjusting algorithms, I proposed, would eventually figure out that a race to the bottom was against all their interests, and that sharing information — whether wittingly or unwittingly — to establish a de facto rent-extracting cartel system made much more sense. This, I then proposed, might be the moment that markets become self-aware.

Bitcoiners and the Mark of the B

This brings me to the latest Bitcoin panel I attended. The panel started as they always do: wildly over enthusiastic claims about bitcoin’s potential, profound disrespect for the social bit of society and reflections that proved a generally poor understanding of how the financial system works.

But there was one guy — who used to own a mining business — who intrigued me a lot. He will remain nameless (for now).

He started off by being totally up front about the scams in the industry, about the copycat pumps-and-dumps, about the totally speculative nature of the run to $1240, about there not being any profitable mining businesses left in Europe and about how only miners in China or those who have access to free electricity can currently turn a profit. Later — echoing a point I’ve made before — he admitted that the smarter parts of the community had figured out how naive they had originally been about finance, saying bitcoin had become “the devil’s way of getting geeks to learn about finance”. For these people Bitcoin had become a sped up simulation of the evolution of the financial system. Critically, he said, it had allowed them to go through the motions in such way that they had learned and adapted to their mistakes. (Part of the Silicon Valley “failing is success” mantra). Funnily enough, I have some sympathy for this way of learning. At Reuters we used financial simulations, for example, to learn about financial reporting. The difference our simulations didn’t require billions of dollars of other people’s investment.

Anyway, he went on, the price itself was unimportant in the long run. The most significant thing was that the “the big banks”  were opening their minds to bitcoin and the blockchain*.

*I have always found bitcoiners who cite banks moving into bitcoin as “good for bitcoin” as  a touch confused. If bitcoin is supposed to disrupt banks and kill off the need for an intermediary, why would banks benefit from moving in? There’s a contradiction here. To me, investment banks and hedge funds only move in when they see obvious profits to be made thanks to clear asymmetries (namely the presence of dumb money), while conventional banks see a way of shifting costs to unsuspecting users.

Which brings me to his most profound comment.

What if, our learned bitcoiner finally proposed, the real potential for Bitcoin was not in its use as money but in its ability to establish a self-governed algorithmic entity, owned by no-one in the strict legal sense, which could one day add and create value by its own autonomous means?

An algorithmic entity, he further hypothesized, that could provide unique and creative services that we — the meat-bags — valued and depended on. His particular example was autonomously produced “music”, which would evolve according to the reactions of the consuming public and its ability to raise funding — a.k.a energy and hosting space — from the system.

He deemed this the start of a new global consciousness based on a totally equal and non-hierarchal node network of support.

Not to be dramatic, but I see this rather differently: the upcoming subjugation of humanity to an energy guzzling AI, that isn’t necessarily guaranteed to serve our best interests, but which does have an interest to compete with us over energy.

Who really would be serving whom in such a set up?

Skynet versus HAL

It’s clear that the scientific community is increasingly worried about the possibility of us developing a runaway AI by accident. The problem is, I think, that everyone is looking for AIs in the wrong place. I’m inclined to think that if and when an AI manifests it will look much more like Skynet than HAL. That is, be the sum of all our digitally networked parts rather than a self-contained autonomous bot. I also don’t think we will necessarily be the ones directly responsible for its creation. The self-awareness might evolve out of an adaptive and linked-up network, that learns how to allocate energy to suit its own purposes in the real world. We might not even notice the moment it becomes self aware and/or when it starts guiding our hands to serve its own interests.

What we should be alert to, in my opinion, is an AI that becomes an economic agent in its own right, and which — thanks to its superior knowledge — learns how to allocate resources away from us and over to it. The risk for us emerges when it learns it is better off enslaving us or rendering us obsolete, than having us be free.

Which leaves me thinking about three things.

1) Can we as a global society suspend our digital evolution at this stage? My impression is that no we can’t. At least not without serious strife or potential dark-ages like consequences. It seems quite clear to me that if we want to sustain or improve the quality of life on this planet it will have to come at the cost of more information sharing, more collaboration, more wealth transfer, more transparency, less privacy, less choice and less freedom (albeit whilst maintaining the illusion of ongoing choice and freedom). The alternative is chaotic regression and a lower and more volatile quality of life for all.

2) If meeting everyone’s needs and desires whilst maintaining the illusion of free will depends on more information sharing not less, that implies that information-dark currency systems inevitably must be replaced by information-intensive data streams that can clear and cross-check against each other on demand. That, to me, involves the formation of a sort of Borg-like collective consciousness, cleared through a central system — like the Borg queen — which sees and directs all. A Queen + drone framework (albeit one with drones that don’t necessarily even know what they’re missing and which retain some limited sense of individuality, because the system can adapt fluidly to cater to those individual whims). In the best case scenario, we might see the emergence of a Star Trek-like meritocratic hierarchy where information is shared and constantly processed by a central computer — owned by the democratic whole that controls it, not the other way around — and where bad agents are not tolerated or relegated to the bottom rung.

3) But what happens if we aren’t prepared to give up our private data to a central information clearer in this way? What if we want to retain our privacy and freedom, and the convenience of the digital economy without falling back into the dark ages? Can we have it both ways? The bitcoiners would like us to think that by encrypting everything and forcing endless consensus competitions the answer is yes. But, in my opinion, this leads to the opposite of a potentially abundant and efficient collective. With obscurity inevitably comes information asymmetry, and with information asymmetry inevitably comes hierarchy, volatility, inequality and abuse. What makes the digital economy efficient isn’t the ease of digital transfer, it’s the information intensity embedded into those transfers. In an encrypted digital system, the claims of bad agents (who hurt economic efficiency) rank pari passu with the claims of good agents (who improve economic efficiency). One set continually exploits the other, and the system remains hugely inefficient because it can never anticipate how big a burden the bad agents will be on the good agents. Information is gleaned only from asset distribution, product prices and real-world physical clues. In that sense, it continues to filter society according to a survival of the fittest mandate.

So here we are — I think — faced with emergence of a “light” centrally cleared information system, the ultimate conspiracy of doves, that rewards productive agents over bad ones on meritocratic grounds, if not finally casts bad agents out of the claim pool altogether. The cost for us is the end of privacy and the transfer of our trust to a centrally controlled processing agent (AI?) that works on our behalf.

On the other hand we are faced with the emergence of a “dark” network-cleared disinformation system, that rewards manipulative agents over honest agents on an honour amongst thieves/survival of the fittest basis, but comes at the cost of serving an (AI?) which demands more energy than you necessarily have to provide it with. You’ve got freedom, in other words, but not much else.

The light system would have no interest in accepting the claims of the dark system, though the dark system would have plenty of interest in trying to get the light system to accept its claims whether by means of propaganda or extortion (threats of violence) or knowledge manipulation/hacking. Without the light productive system, the dark system would probably devolve into a medieval-esque structure of fiefdoms.  But equally, without the dark system, I worry the light system might devolve into something of a purpose lacking nanny state.

Either way, it seems to me, the choice is between total transparency (that leads to an abundant but highly structured “division of labour” brave new world society) and total opacity (where transparency or trust immediately deals you an economic disadvantage you can’t afford, impeding division of labour and prosperity).

Both options don’t seem great to me.

Disclosure: I originally wrote this post about two weeks ago but refrained from publishing it because I feared it was too far fetched. But I’ve since seen this post on Motherboard which also makes the link between bitcoiners and satanism so have decided why not publish it after all?

P.s. I’m pretty sure you can extract a meaningful satanic anagram out of the name Satoshi Nakamoto as well.






On disrupting death

Saw this posted by Climateer Investing, regarding what I have long described as Silicon Valley’s God complex:

The entrepreneurs are driven by a certitude that rebuilding, regenerating and reprogramming patients’ organs, limbs, cells and DNA will enable people to live longer and better. The work they are funding includes hunting for the secrets of living organisms with insanely long lives, engineering microscopic nanobots that can fix your body from the inside out, figuring out how to reprogram the DNA you were born with, and exploring ways to digitize your brain based on the theory that your mind could live long after your body expires.
“I believe that evolution is a true account of nature,” as Thiel put it. “But I think we should try to escape it or transcend it in our society.”

Don’t get me wrong. Extending and improving mature life is a noble cause. But we should be mindful of the potential tail-end risk that comes along with “disrupting” death.

First, there’s the Miracle Day scenario (if you think insurance/pension firms have funding gaps today, imagine what might happen to them if we find the secret to eternal life but not necessarily the secret to rejuvenation).

Second, if we all live forever, it seems reasonable to assume we wouldn’t be able to have children at the same rate. But can you imagine a world with no children? A world with no rejuvenation? And a world where new life has to be capped? What becomes of our purpose? I think unless we find a way to colonise new planets and extend life beyond our own solar system, eternal life would be a frightful prospect for our planet.

All of which reminds me of a personal anecdote.

When I was about five years old, I used to worry a lot about dying. In fact, during that period of my life the fear of dying consumed me night and day — preventing me from sleeping through the night, driving my parents up the wall. This continued until one day, whilst trying to calm me down, my dad said: “don’t worry I know a man who can sell me some magic pills. They make anyone who consumes them immortal. They’re very hard to get, but I think I know a way to get them.”

Needless to say I begged my dad to get me these pills and promised I would be good and always do what my parents say if only he would find a way to get them.

Next day, as if by magic, my dad appeared with the pills to hand.

From then on I wouldn’t go a day without taking them. (I later found out they were just vitamin C and that I was a total mug, but that’s another story). In any case, I referred to them as my anti-death pills, and if they were ever forgotten or ran out I would panic just as much as I ever did. This state of affairs continued for about six months until all of a sudden it occurred to me that if I lived forever I would probably get immensely, hugely and unfathomably bored.

It was now the fear of eternal boredom that started to concern me more than the fear of death. I also coincidentally had a horrible nightmare — which I remember vividly to this day — of a nice sunny day at the park turning into something akin to a zombie apocalypse, with everyone around me beginning to decay and turn into a skeleton.

I decided at that point to stop taking the pills. It was better to die than to be eternally bored. This was a big epiphany for a five year old.



Digital scarcity

Here’s a thought experiment.

What happens if and when all the digital services we have become accusstomed to getting for free experience a system-wide revenue slump (perhaps due to a sudden dearth of advertising dollars on the back of extreme market fragmentation that makes advertising on platforms much less appealing) and are forced to start charging for services instead?

How would such a sudden system-wide digital scarcity affect users? How would users choose to allocate their modest salaries to digital services if they had to pay for them? Who would win and who would lose?  What would be considered vital and what would be considered a luxury? To what degree would such an event constitute a major inflationary shock?

Thinking about myself, I would pay only for the following services and only if certain conditions were met:

1) a neutral search engine (one that doesn’t use my data to better target advertising at me and doesn’t rank commercial searches I might make according to the dollars corporates have passed on to the search engine).

2) my email, providing it was cryptographically secure.

3) one instant messenger app.

4) one social media site (that didn’t target advertising at me. It would probably be Twitter, not Facebook or Linkedin).

5) up to 10 different news sites. I already pay for subs for the New York Times, New Scientist, WSJ, Vanity Fair. The FT is a given. I would pay for Reuters, some Bloomberg content (can’t see myself living without access to Matt Levine), perhaps the Daily Mail (sometimes you need a brain rest), im presuming BBC would remain free because of the license, and I’d also pay for the Guardian. The remainder I would pick randomly. I’d pay for an archive service.

6) would I pay for Youtube? Probably not. If we are going down the pay route i would pay for video content distributed via formal channels. I would not pay for selfie tutorials or “viral” pics of cats. Id expect those could be sent to me via the other channels I pay for, and be hosted on different sites. Would i pay to have access to a video broadcast network that I can post content to? Probably not. Only on a case specific basis.

7) i’d pay wordpress to host this blog.

8) I’d pay for an RSS reader — even if i just get links to content that I later have to pay for. But I’m presuming some truly voluntary altrusistic content would remain free, notably in scenarios where the cost of hosting the content is covered by the author.

9) i’d pay for cloud storage and photo sharing/storage.

10) i’d pay for an encyclopedia, especially if was run and verified by experts, with a managemenf board that was accountable for mistakes and errors.

11) I’d pay for food recipe sites. Or other hobbiest specific content.

12) I already pay for netflix and spotify, and buy content on itunes, so I’d continue to do that.

13) I’d possibly pay for a travel information app.

14) I’d pay for map services.

15) I’d pay for a secure browser service.

16) education platforms.

17) neutral honest marketpalces that don’t give preferential treatment to particular participants. 

Things I would probably not pay for out of my own pocket (because there are limits and as a journalist I would expect my employer to pay for some of these, or get journalistic priviledge):

1) translation services

2) google earth

3) a host of social media sites/apps im currently signed up to but never use.

4) google books

5) academic journals

6) google trends

7) google analytics (im not too fussed about knowing who is reading my blog, since im not  trying to sell advertising)

8) any corporate retailer’s app that gives access to product sales.

9) medical information – i’d expect a publicly funded and factually corroborated service.

10) product reviews

11) trivial databases

12) reddit, and chat forums

13) government websites

14) national statistics

15) directories

16) price comparison, or marketplaces 

17) skype. I would use my normal phone.

18) Tumblr.

19) Ello, and all sorts of other social media sites.

There are probably a multitude of things I’ve missed out. But you get the gist. If I tallied up the cost of essential digital services I would prefer not to live without, I think we would come close to an additional monthly expense of at least £1000 for digital goods I currently get for free. That would hurt anyone’s pocket. That would also count as an inflationary event.

Remember the basis for this thought experiment is that all these services are being subsidised by advertising dollars and/or speculators awaiting advertising dollars. But that the model only works in a winner takes all digital environment. The more the digital universe fragments the less advertising dollars there are to go round (or they get distributed in such a localised way that they can’t form the economies of scale that can allow major cross subsidisation of digital products to continue indefinitely).  The positive sum game that the internet is based on (the idea we can all get a free lunch on the internet) may in that scenario turn out to be a fallacy, or even a temporal market inefficiency. When the true cost of these services is revealed, in terms of energy and human labour/attentiom, we may suffer a major repricing in the economy.  On the plus side we would become more discerning digital consumers. This would see poor quality providers, including system spammers, go the way of the dinosaurs (one would hope). 

Basically what I’m getting at is that I don’t think there is any such thing as zero marginal cost. Even on the internet.