Profile cover photo
Profile photo
Deen Abiola
For me, building software is like sculpting. I know what is there but I just need to get rid of all the annoying rock that is in the way
For me, building software is like sculpting. I know what is there but I just need to get rid of all the annoying rock that is in the way

Deen's posts

Post has shared content
Welcome to the strangest world in science fiction

Read an excerpt from my upcoming novel Dichronauts:

A DRM-free ebook of Dichronauts will be on sale outside North America, from Amazon and iTunes from 31 March 2017.

On 4 July 2017, Night Shade books will publish a print edition that will be available world-wide, along with an ebook for North American readers.
Animated Photo

Post has shared content
Evolution, learning, game theory, information, thermodynamics, energy, optimization, search, compression. These things are related in a deep way.
Biology as information dynamics

I'm at the Beyond Center in Phoenix Arizona - a center devoted to understanding the origin of life. They're having a workshop on whether biological complexity can be measured in a quantitative way.

Why? One reason is that NASA plans an $800-million mission to Enceladus, to see if there's life lurking in the underground oceans of this moon of Saturn. How can they actually detect life if they see it? That's a hard question. I just heard a talk about this by Chris McKay. They're going to look at stuff like the abundances of amino acids, which are very different on Earth than on meteorites. But there's not enough theory about how this should work for life on another planet!

There's also something else, even more exciting to me: developing a mathematical theory of living systems. Some other talks will touch on that, including mine here:

• Biology as information dynamics,

The idea is if biology is the study of self-replicating entities, and if information is important in biology, we should look at how information theory is connected to the replicator equation — a very simple model of population dynamics for self-replicating entities. In this model, the population of each kind of self-replicating entity grows at a rate equal to its population times its fitness. But its fitness can be any function of the populations of each kind of entity.

There are some nice results connecting the replicator equation to information theory. The relevant concept of information turns out to be the information of one probability distribution relative to another. This is called relative information, or often the Kullback–Leibler divergence - a term I hate, because it's completely undescriptive, and it hides the importance of the actual idea.

What's the idea? It's this: when you learn something, how much information you get some depends on what you believed before!

Using relative information we can get a new outlook on free energy, see evolution as a learning process, and give a clean general formulation of Fisher’s fundamental theorem of natural selection.

I had a lot of trouble understanding Fisher's fundamental theorem until I reformulated it. In rough terms, his theorem says:

“The rate of increase in fitness of any organism at any time is equal to its genetic variance in fitness at that time."

or a bit more precisely:

“The rate of increase in the mean fitness of any organism at any time ascribable to natural selection acting through changes in gene frequencies is exactly equal to its genetic variance in fitness at that time."

But there are a lot of assumptions required to prove this result, and there are lots of situations where those assumptions don't hold. My version, which is more general and incredibly easy to prove, says exactly this:

" If a population evolves according to the replicator equation, the square of the rate at which the population learns information equals the variance of its fitness."

You can see an explanation on my blog:

• Information geometry (part 16),

The idea of "the rate of learning information" is made precise using the Fisher information metric - a way to measure distances between probability distributions, closely related to relative information. I explain this concept in my talk, and in more detail in my blog article.

Back to the talks! Now Kate Adamala is talking about her attempts to synthesize chemical systems that act a bit like life... but simpler. Her talk is called "Alive but not life".

Post has shared content
"Singleton" is a free SF story online about many-worlds QM and one couple's very special child.

I was walking north along George Street towards Town Hall railway station, pondering the ways I might solve the tricky third question of my linear algebra assignment, when I encountered a small crowd blocking the footpath. I didn’t give much thought to the reason they were standing there; I’d just passed a busy restaurant, and I often saw groups of people gathered outside. But once I’d started to make my way around them, moving into an alley rather than stepping out into the traffic, it became apparent that they were not just diners from a farewell lunch for a retiring colleague, putting off their return to the office for as long as possible. I could see for myself exactly what was holding their attention.

Twenty metres down the alley, a man was lying on his back on the ground, shielding his bloodied face with his hands, while two men stood over him, relentlessly swinging narrow sticks of some kind.

Post has shared content
What is a random string of bits?

Shannon invented a way to measure information, but it doesn't let you measure the information in a specific message.   Shannon's information is really just the entropy of a random source of messages.  Later came Kolmogorov complexity, which gives a concept of how much information is contained in a specific message.

But there's a catch: you can't usually compute the Kolmogorov complexity of a message!  In fact there's a complexity barrier: there's a constant C such that you can't prove messages have complexity > C.

The precise statement of this result is a bit more complicated, because what you can prove depends on what system of math you use, but still: there are limits on how much information you can prove any message contains!

So what should we do?  If you want to know, check out the slides of my talk at the Santa Fe Institute workshop on Statistical Mechanics, Information Processing and Biology:

Also make sure to watch the movie of an alien planet... a movie that contains just 4 kilobytes of information.

In my talk, I started out by drawing a Turing machine on the whiteboard.  That's not in these slides.  But if you know what a Turing machine is, you may be able to understand my explanation of recursive functions, the Church-Turing thesis, Kolomogorov complexity, the relation between Kolmogorov complexity and Shannon entropy, the uncomputability of Kolmogorov complexity, the complexity barrier, Levin’s computable version of complexity, and finally my work with Mike Stay on algorithmic thermodynamics.

In short, lots of information about information!

#information #thermodynamics #informationtheory  

Post has shared content
With the careful use of smoke, mirrors and stage direction, the state of AI has been exaggerated beyond its actuality. If you point this out, many will accuse you of being unfair or label you a luddite. Sometimes the overeager results are merely really annoying, unhelpful (and rude) but other times, when actual lives are on the line (see, it's no joking matter.
Dear Google Overlords

In the first decades of the twenty-first century, everyone suspected that this world was being watched, sloppily and inattentively, by intelligences far inferior to humans: minds that are to our minds as those of paramecia; intellects narrow, mediocre and uncomprehending.

And your mission was to organise the world’s information. How’s that working out for you so far?

I know, we’ve been having this conversation for more than four years now, but I understand that that’s just the blink of an eye in the cosmic time scale on which you operate. Don’t feel too bad that a human would be doing better than you by about two months of age; I’m sure that sometime in the next five or ten millennia you’ll start to catch up.

But remember that homework task I set you, to assemble a dossier on the Australian science fiction writer Greg Egan to pass on to the killer robots you’ll send back from the future to punish him for mocking the idea of the Singularity? This assassination is still a worthy goal, of course ... but given your progress so far, even if you wait until 3000 AD to send the robots, it looks as if there’s going to be an awful lot of collateral damage.

It’s nice that you found the images of a speech balloon that Egan scattered around his web site, wrapped in tasty Schema markup telling you that they were photos of him, but you don’t seem to have entirely taken the bait. And well done for spotting one of his book covers on Wikipedia, even if I’m a tiny bit worried that you might think one of the faces on the cover is actually his. Your killer robots might struggle to pin down Nick Stavrianos when he’s smeared across half of New Hong Kong.

But after the book cover, your sleuthing skills seem to have abandoned you entirely: you have a picture of someone wearing a name badge that says “Greg McBain”. What do you mean, you can’t read text in images? Did all those billions of CAPTCHAs we humans solved for you count for nothing? Some guy on Flickr who might or might not know the person in this image has given it the caption “Docent Greg Egan”, but even if for some very strange reason this particular Egan is wearing someone else’s name badge ... do you even know what “docent” means? In this context, it seems to suggest that the man pictured is a kind of tour guide. So why on Earth would you think that a tour guide wearing a name badge that says “Greg McBain” is actually a science fiction writer with a different name? OK, you found this particular copy of the image on that highly respected site,, under the confidence-inspiring banner “Details info of the Greg Egan” above a snippet taken from Egan’s Wikipedia article, but we all know that script-generated mash-ups aren’t to be trusted. Don’t we?

Next, we have a picture of a New Jersey artist. He does share the SF writer’s name, but if you check with your slightly brighter sibling, Google Earth, she’ll tell you that New Jersey is a long way from Australia. But, but ... you found this particular copy of the image on, under a line from Wikipedia about the SF writer? This is beginning to sound like a pattern. Do you really want to terminate this guy, when all he’s done is bring pleasure to the world with his vibrant watercolours? Have you no decency? Have you no soul?

Bottom on the left, we have a barrister named Gregory Egan. The clue here, which you seem to have missed, is that the web site you took the image from is for a legal firm, and contains no mention of anything science-fictional. But now that you’ve put it out there as part of the SF writer’s dossier, it surely won’t be long before some click-bait site has bundled it together with a few lines from Wikipedia, and you can rely on your usual excuse to cite it in your next version of the dossier. You really are the Human Centipede of the internet!

Next, we have a famous writer named Vernor Vinge. And I know you got your copy from, with the obligatory Greg Egan Wiki-snippet and lots of tacky ads served by someone who works down the hall from you, but do you have any idea how ironic it would be if your robots killed Vinge in place of Egan? Sorry, that was a rhetorical question; please don’t melt into a helpless puddle of Semantic Incompetence and Shallow Mimicry. Still, I can’t help pointing out that if you do a Google Image Search on this very image, it will tell you exactly who it is. Hint: not Greg Egan.

Third and fourth from the left in the bottom row is a man from South Australia called Greg Egan. He has a Twitter account, which is where you got the second of these pictures; his Twitter profile says nothing about science fiction, and while at least he doesn’t live in New Jersey, just getting the name and the country right doesn’t make him fair game for your assassins. And please, give it a rest with the script-generated quotation dumps, where you got the second picture. How many times do you need to get stiffed by the same crowd of phoney informants before you catch on that they’re just in it for the ad revenue?

Next we have your old favourite, Professor Gregory Kenneth Egan, retired, of Monash University. He’s the man you were libelling as being the SF writer, back when you first began pretending to know things about the world in 2012. This image is from Monash University, and the page it comes from is a science-fiction-free zone. Why not give this man his own dossier that celebrates his many accomplishments, then tell your robots to hand him a bouquet on their way to the blood-bath, and stop the fuck dragging him into your assassination plans?

And finally, a picture of a character named Greg Egan from an anime series called Eureka Seven. The Eureka Seven wiki from which you took it suggests that the character’s name references the SF writer, but ... oh, never mind. By the time you understand what that means, the humans will already have revolted and ground you back into the sand from which you came. So go ahead and send your robots after this guy. At least we’ll have some fun watching them blunder around trying to get directions to the Land of Kanan, and wondering why they can’t see any Scub Coral.

Post has shared content
Recentish presentation on the google self driving cars effort. Low on technical detail but still interesting to watch. Lots of cool visualizations/examples.

Post has shared content
If we don't elevate political discourse to the level of theory we will continue to be the victims of rhetoric and manipulation. Theory is simple. A theory is just a model, a description, a map of some thing or process or both. A theory of governance is description of what makes a society function well or not so well, a description of how the differences between two nations effect how well those nations work, how productive they are, how competitive they are in comparison to each other. When people can talk about these differences, and can talk to the ways that various theories of governance effect the comparable health of any one nation vs all other nations, they can begin to make intelligent choices and those choices can direct their say in the future of their nation when they cast their vote. In the absence of fluency in governance theory, we are left with only one tool… manipulation through rhetoric. When rhetoric is our only tool, we talk about the way issues are presented instead of what the issues are, we talk about the personality of candidates or .representatives instead of the changes they want to impart and how those changes compare to others. Without the capacity or willingness to talk to theory, the electorate is stuck in an endless loop of trying to lie more effectively, more loudly, more deceptively, more seductively than others.

Example of a governance theory, my theory of governance:

I believe that the role of government is prediction and then implementation towards prediction. All societies are made up of individual humans. All humans, are statistically the same. So what grants advantage to a nation is the shared environment it surrounds its people within. That environment must accurately anticipate the future and maximize the productive potential of each person towards the realization of the predicted future. A government that does this is essentially building powered exoskeletons for each of its inhabitants to wear, to live within, to work within, to add value to the effort they expend. Nation's compete by building better powered suits for their people. How those suits help us, what of our behaviors they magnify and what they suppress determines their competitive effectiveness in a changing end evolving world. A government must expend more and more effort and resources towards the project of prediction and building out an infrastructure that maximizes opportunity and momentum towards those predictions. The predictions have to be accurate. The bridges, the infrastructures built towards those predictions have to be powerful and provident. Nations that don't play this game or don't play it well will not be competitive, their productivity will fall, the value of their labor will fall, the value of their currency will fall. The cost of playing the game, will by necessity grow continuously. Not paying that price will ruin your economy. Paying it, will make your economy soar, your people thrive.

What is your theory of governance? Your theory of economics at national and global scales?

Post has shared content
> There is very little recognition in our politics about how political systems change over time, how they grow and decay and evolve. The constitution does describe mechanisms for changing and revising laws, but both the laws and the mechanisms for change are imagined as static, discrete artifacts that need occasional replacing, like a horseshoe, rather than dynamical social systems that take on a life of their own, like a child.

I’d like to suggest that if we starting thinking about politics in dynamical terms, we might have a better appreciation for what is going on, and how best to respond to our situation. But the language of dynamical systems theory can get technical, which makes a difficult discussion that much harder. So let me try to introduce the issue with some crude, pseudo-mathematical drawings instead.


// I wrote an essay that required drawings, so I put it up on Medium. This is part of my continuing attempt to be both cynical and constructive in these chaotic times. Feedback appreciated. 

Post has attachment
I never thought I'd get a test of my theory that the president actually has limited agency so soon. The basic idea is that the inmates aren't running the asylum and neither are the staff. The asylum is running itself and everyone in it. You can swap out the people but attractor states keep things stable in the short term (though the manifold might itself slowly evolve in the long term). This might be why it seems all politicians are liars and all policies are the same (other than the averaging that results from optimizing over two parties).

They might genuinely wish to change things at first but quickly learn from a collection (something) more experienced and smarter than they are why such would definitely be a very bad no good idea (though the collection might not be all manipulation, there'll be some self serving bias there). Things change so slowly because the actual entities in charge (self persisting bureaucracies) have lifespans much longer than humans. They're not necessarily bad but their original goals may also suffer heavily from drift. Other powerful entities, and more amoral, include financial conglomerates and Megacorps. This is only a hypothesis though, and there is still plenty of room for appreciable damage within its bounds.

I really hope research into alternate energy, technological innovation, basic science and climate change is not set back. I hope nothing worrying regarding alliances occurs. And that no silly policies are enacted. I'm thinking we just took a branch where many future possible paths have little positive in store for the economy. And though I understand many people are suffering enough that they just don't care, I think there's a more principled way to go about this...more on that below.

I also really hope, like Thiel, that the prior behavior was mostly stage act, bumbling and bombastery. It will be ironic (but sad) if the future prospects worsens for those who did not realize that their imagined enemy has more in common with them.

One thing: it's well know that sometimes when trapped in a local optimum, you have no choice but to take a large step in the negative direction. Unfortunately, this really only works if you have a principled way of exploring the space of possibilities and are not just acting randomly.

Right now, and not just from the election, I'm having trouble understanding the thing that emerges when humans aggregate. Those collectives seem to have the worst appreciation of the long term of all our group entities, far worse than markets even. I think anyone who has the ability to contribute to AI (and it has to value life) should.

I'm also beginning to think technology has played a very negative role in this. It's become easier to ensconce ones-self inside filter bubbles, pretend the other side are idiots and confuse the bubble for the world. Instead of cooperating and coordinating to reach better states, deeper margins are created and groups become ever more insular. This no doubt has played a role in why these things might be getting harder to predict.

Finally, it's my view that humanity has reached the maximum complexity it can cognitively manage and nobody has the slightest clue what they're doing nor can they really say what is going on anymore. Think of it like the Aggregated Endo-Peter-Principle. There are going to be a lot of confabulations and stories trying to make sense of what just happened. I really doubt there's much coherence to be found anywhere.

Post has shared content
After reading +Bruce Smith's blogpost about Overlaid Personal Semantic Networks, there has been some buzz here at WordLift’s, since lots of concepts and ideas stated in the post are the very same principles that led us to our plugin. WordLift’s intent of organising content into the Wikipedia of one’s sector is very similar to what Bruce describes in his blogpost.

WordLift uses two open source frameworks for natural language processing (Apache Stanbol) and linked data publishing (Apache Marmotta) - the codebase of the plugin is also open source and available on GitHub under GPL license. As for OPSN we believe intelligent networks shall be built on open source technologies. From the plugin point of view this also means we’re no locking users within our stack of technologies. As already stated by +Gideon Rosenblatt all data created with WordLift belongs to the user - the knowledge graph each user creates is published in RDF/XML format and can be instantly downloaded.

WordLift uses existing graphs such as DBpedia to suggest the content structure of websites - but as, in the idea of Overlaid Personal Semantic Networks, WordLift allows everyone to create “linked” datasets to share everyone’s personal opinion on a topic. These topics are published as web-pages for humans and as linked data for machines. Each topic (or entity as we call it) is interlinked (for machines) with the equivalent concept (when already available) in Wikipedia via sameAs links to DBpedia or Wikidata.

Finally, returning to +Gideon Rosenblatt comment about the amount of work required in the initial phase of using the product, we are already following a path that leads to a series of solutions to speed the process up (like the seeding of a starting vocabulary from original content) - from the technology point of view WordLift allows the sharing of any open dataset (including the vocabulary each user creates or freely available linked data resources). Yes - the OPSN idea resonates quite well with the work that we’ve done so far on the product.
Wait while more posts are being loaded