The Basilisk Is a Lie

BasiliskA thought experiment known as Roko’s basilisk escaped from the dungeons of LessWrong has recently been causing a wave, mostly among fans of sensationalist headlines. The core proposition can be paraphrased like this:

In the future there will be an ethical AI that punishes everyone who knew they could have but in practice did not work towards its eventual birth. If the humans in question are deceased by then, a simulation of their minds will be punished instead. This is a moral action, because due to the AI’s capabilities, every day that passes on Earth without the AI is a day of unimaginable suffering and death which could have been prevented.

Now I am way less prone to ballsy absolutist assertions than practically anyone frequenting LessWrong, but this whole thing is wrong on many levels.

The central argument about the ethical validity of this punishment scheme is beyond questionable, specifically the motivation of the AI. At the point where the AI achieves this capability, the assertion that the execution of the punishment is morally imperative is mistaken. By that time, nothing is actually achieved by carrying it out. The behavior of those “guilty” will not change retroactively. Since their future behavior is also irrelevant, the argument rests on the assumption that without the prospect of punishment there would have been no motivation for humans to develop the AI. While this is false in itself, punishment after the fact without the hope of achieving any effect besides the imposition of suffering cannot ever be an ethical act. Ethics aside, the contributions of individuals not directly connected to the eventual birth of the AI would be murky to judge as well. What’s the correct “punishment” for a computer scientist, as compared to a medical doctor?

While there is little uncertainty that general AI is feasible and, if we continue on the path of scientific discovery, unavoidable – significant doubts exist about the nature of that AI. If this thought experiment shows nothing else, it does illustrate that our notions of what constitutes a “friendly” AI are wildly divergent. One can only hope for the sake of whatever becomes of humanity as well as the AI’s sanity that reading LessWrong will be one of its less formative experiences.

Where feasibility deserves to be harshly questioned is the simulation idea this Basilisk concept relies on to carry out its punitive actions. The most central assumption here is that a mind reconstructed from extremely lossy data fragments is still the absolute (!) equivalent of its original version.

That means at the core of this is a belief that if I were to die tomorrow, and my mind was being reconstituted from nothing but my old Amazon shopping lists, this would be the same as me.

It should be very obvious that this is not true, but to make matters worse my “sameness” value not a Boolean. It’s not even a scalar value, it would have to be a vector spanning a lot of aspects, each measuring how much of the original mind was successfully transferred. It is disconcerting that this basic notion is not being shared by the rationalist movement. Instead, it is apparently considered feasible to reconstruct any specific thing by using deduction from first principles.

The sheer amount of models and parameters that could lead to the development of general artificial intelligence is huge and in its entirety inconceivable. While it is still appropriate to engage in informed speculation, one should be skeptical whenever certain models and parameters are cherry-picked and arranged just so, in order to illustrate a thought experiment that is then deemed to be an inevitable outcome. This reduces technical complexity and historical uncertainty to an absurdly simplified outcome which is simply taken as fate.

Already, a big number of AGI scenarios have become intellectual mainstream, some of which claim exclusivity for themselves. Some go further and assert inevitability. Otherwise rational people can come to these conclusions of inescapable future outcomes because they are losing sight of the complexity of factors and conditions their reasoning is based on. No statistician would chain together a list of events with 80% assumed probability each and claim the end product is a matter of destiny. Yet, for some reason, futurists do this.

It is reasonable that a number of these scenarios might eventually play out, with some variation, and in some order. They obviously can’t all be true at the same point in time and space, including the Basilisk.

Of course that also means, just because nothing in principle prevents it, somewhere in the universe Basilisks may well exist already. But there is no reason to assume it has to on Earth. It would take a special cocktail of circumstances.

A Modern Pascal’s Wager
The core argument why this idea is perceived as dangerous is that people who understand it will be forced to act on it. This means acting out of fear of future punishment, just in case there is an invisible entity out there who cares enough about your actions. Even if you accept this premise, and even if you’re deluded into thinking this is the path to an ethical life, the huge problem is predicting what that entity wants you to do so you can avoid punishment.

This is the definition of a problem where you do not have enough information to make an informed decision. In the absence of any information about that deity, acting on its behalf is an execution of random fantasy.

The claim behind the Basilisk is again one of inescapable certainty, in fact it desperately relies on that property. Because you supposedly know what the Basilisk wants – it wants to exist – this is seen as a solution to unknown deity problem. However, this only works if you believe in the properties of Roko’s basilisk dogmatically, disregarding all other AI futures. This is in fact the exact analogue to the original Pascal’s Wager where the not-so-hidden assumption was that the Christian fundamentalist god was the only one you had to please.

Of course, within the context of an AI that can simulate people, this is all moot. There is nothing preventing said AI from simulating you in any set of circumstances, including perpetual punishment or everlasting bliss. In fact, there is no real cost to simulating you in a million different scenarios all at once. Acting out a random fantasy based on the off chance that in the future one of your myriad possible simulations will have a bad day is not rational.

Some of the reasoning on display here seems to mistake blunt over-simplifications for clarity of thought. To an outsider like myself it looks like complex multivariate facts are constantly being coerced into Boolean values which are then chained together to assert certainties where none are really warranted. There is a certain blindness at work where everyone seems to forget the instabilities hidden within the reasoning stack they’re standing on. But what’s worse is that fundamentally unethical behavior (both on part of the AI and its believers) is being rebranded as legit.

I see now the way to hell is paved with people who think they are acting rationally.

The Minds of Octopodes

Octopus_verrucosusThe fact that relatively high intelligence has arisen from many architectures multiple times on this planet bodes well for the frequency of intelligent life on exoplanets.

Even today, there are still some theoreticians who assert the formation of intelligent life is tightly coupled to our specific brand of brain and should hence be considered a huge accident. Yet we see the formation of minds in a lot of places, and they can be almost arbitrarily far removed from the human brain.

The octopus is a great example for a radically different neurological substrate. Given enough time and a little bit of luck, some funghi too might evolve a mind of their own as Fuligo Septica already shows some capacity for problem-solving behavior.

The argument that human-like intelligence is again another unlikely step discards the myriad of social animals with advanced problem solving capabilities, some of which are even tool users, and some of which have actual languages and cultures. A lot of these have come onto the stage very independently from us, having sprung from far removed genetics – and yet they have enough in common with us that should make us recognize the frequency of species with minds might indeed be high wherever life takes hold.


The Publicity Coaster

Rollercoaster_Tornado_Avonturenpark_Hellendoorn_NetherlandsWhat do you do if you’re a designer who desperately needs some attention? Well, you invent the Euthanasia Coaster, of course.  So edgy and controversial it causes me to give it some attention right now!

I come from Germany, the country that not only raised the bar when it comes to the industrialized killing of its population, it pretty much redefined the scope of evil achievable with cult-like dictatorships empowered by modern technology.

Euthanasia as used in the article is a form of newspeak introduced by the Nazis, a cynical redefinition that constitutes a corruption of the word’s original meaning.

As such, the word choice used here makes me just as uncomfortable as the concept itself. Originally, euthanasia (greek for “good death”) is a means of ending a life for which the only perspective left is suffering. In this sense, the word is still being used (appropriately) in conjunction with terminal illness where it does have a place as part of the right to self-determination exercised by patients who make the conscious and informed decision to end their lives in order to avoid this suffering.

I really wish people would stop taking hints from the Nazis when it comes to vocabulary, even if it’s on purpose. “Death Coster” or “Ride of Doom” would be perfectly sufficient. But poor word choice isn’t the only thing that plagues this PR, it’s also doesn’t really hold up to biophysical scrutiny.

The Actual Science

The “critical” portion only lasts a few seconds, up to a minute. Depending on the direction of the force, 10g for a minute would be close to the edge for untrained people but not expected to have long-lasting harmful effects in most cases.

There are several ways in which high g forces cause harm to humans. The article mentions blood flow, specifically applying the amount of acceleration necessary to stop oxygenated blood from reaching the brain. Completely stopping the flow of blood for 60 seconds will result in a loss of consciousness, but the designers of the coaster seem to be under the impression that achieving this even for a moment kills people. They’re wrong. If normal blood flow is restored after 60 seconds of anoxia, no adverse effects are to be expected at all, not even in the short term. Of course statistically there will be cases where the heart enters one of several possible failure modes under these conditions (again mostly in humans with pre-existing health problems) and while I expect it to be rare among the healthy population, those people would indeed need immediate medical attention – but they too can be expected to make a full recovery if they receive it.

High g forces can also damage blood vessels due to simple overpressure, causing them to rupture. This will happen with body parts located in the direction of the force applied. In this design, this will be the lower extremities, where this damage – if it occurs – will be minimal. But if you suspended people “upside down”, that would be another story. Overpressure in the blood vessels of the brain is a dangerous thing. Again, I’m not sure 10g for 60 seconds is enough, but I’d intuitively say if there was any way of inducing fatalities with this coaster that would be the way to go. Especially people with existing defects and weaknesses of the blood vessels in their brains would be most at risk, people with aneurisms for example.

Lastly, high g forces can cause tissue trauma due to compression or internal impact damage. 10g for 60 seconds would not be enough to cause that in healthy organs. But if the coaster’s design was changed to 20 or 30 g, delivered over an extended period, injuries and fatalities due to organ trauma (including the brain) will occur.

So on final consideration: this roller coaster will cause people to pass out for up to a minute. However, this effect is completely reversible. In healthy people, no lasting damage is expected. In fact, this kind of acceleration is a standard part of what happens when fighter pilots train in centrifuges – although 10g is I believe at the extreme end of what could be safely considered for training purposes.

Imagining the Post-Antibiotics Future

Antibiotics are done, and they’ve been failing increasingly for at least two decades even though nobody cared to report it then. The speed of our transition to the “post-antibiotics era” is certainly our fault, but the phenomenon itself was always inevitable. Biology is an arms race, and it will always be one. In fact, biology overall is one of the worst substrates imaginable for hosting conscious entities, and that means this and many other problems may never be permanently solved.

Staphylococcus_aureus_(AB_Test)However, it’s not all hopeless. There are a lot of bacterial infections that will continue to be treatable with antibiotics for some time to come. So it’s not like we’ll lose all of our capabilities at the same rate. It just means that gradually the number of strains that are panresistant will increase.

To see what complete powerlessness over an infectious agent looks like you don’t have to look further than viral infections today. The reality is we still can’t do very much to treat them, so over time we’ll just add significant bacterial adversaries to this list of diseases for which there is nothing else but symptomatic treatment available.

This will all change, because medicine has to change. In the future, it won’t be enough to hit an unspecific infection with an unspecific antibiotic. It will be necessary to analyze individual infections and their interactions with individual patients in detail – and then very specific biochemical treatments will have to be tailored on a case-by-case basis.

The recent rise of monoclonal antibody drugs is a bridge to that future. We’re still not knowledgeable enough, and we’re still using these treatments with all the finesse of a 3-year old hammering huge lego toys together, but we’ll get there eventually. We have to, it’s the only hope against a lot of other diseases as well, including cancer.

In the mean time, there is a lot of research that needs to be done. We’re also missing the actual technologies, both diagnostic and generative machines, but the most important obstacle to overcome will probably be the culture of medicine. This culture will change only when there is enough pressure for it to change. The FDA will have to change when its primary role becomes denying treatment to people who are going to die instead of keeping the population safe. Medical professionals will change when it becomes a big factor that they lack the scientific background to effectively devise and apply advanced treatments. But first, all of these disasters will probably have to actually happen, and we’re already seeing the first signs of that trend.

So, there is definitely a path forward – it’s not an inevitable descent into the dark ages. It’s just that due to inertia and time spent waiting for basic scientific advances, there will be a period with less protection overall. Drugs may become available to plug some of these gaps, like insect-derived antibiotics, which will eventually fail too but will buy some additional time.


Hypothetical_exoplanetThinking about what Earth and its star system look like from the outside is indeed a very interesting exercise, and it gives you an idea how flawed many of our own assumptions about exobiology may be.

At these distances, our radio spectrum would be difficult to discern from random galactical noise – even if the radio waves had enough time to cross the distance. And in fact, so far it looks like the more advanced we get the less telling our radio signature becomes. There are no large-scale constructs or artifacts visible that hint at any technological activity in the solar system.

Looking at how crowded the solar system is, an onlooker might well assume it’s too unstable for life to have developed. The Earth might be seen as too small to support life, its atmosphere too thin to shield against radiation and also too thin to allow for a comfortable temperature interval in which liquid water exists. Due to the difficulties of aerodynamic flight under these circumstances, they can safely assume there will be no winged lifeforms on Earth.

They might also think that any plant life on Earth would be extremely sparse and desolate – yet the abundance of oxygen hints at a world dominated by microbial life. The orbital and rotational instability of the planet leads to wild temperature variations which will likely prevent the rise of higher life forms.

The presence of several semi-unstable interplanetary debris fields suggests asteroid impacts will be frequent, and likely to extinguish life on any of the planets. The relative proximity and high mass of the Earth’s moon (or shall we say: sister planet) makes it likely that the planet will be subject to tidal disturbances and might imply a history of extreme geological instability.

Water covers a large part of the Earth’s surface, but generally is only present in the form of very shallow oceans tainted with poisonous chloride salts in high concentrations. It’s unclear if unchlorinated water exists on the surface and in any case it’s doubtful that there’s enough water to form a stable cycle able to support any but the most primitive extremophiles.

Neuroscience Quantum Bullshit

PSM_V14_D495_Side_view_of_the_brain_and_spinal_cordEvery neuroscience article with the word “quantum” in it is bullshit and should be disregarded.

Case in point: Discovery of quantum vibrations in ‘microtubules’ inside brain neurons supports controversial theory of consciousness (Science Daily)

We should put the science in Science Daily in scare quotes from now on.

This is very old pseudoscience with a recycled thin veneer of freshness. The idea that “quantum vibrations” in neuronal “microtubules” lead to decidedly macroscopic effects like detectable EEG signals is absurd on many levels, including spatial and temporal scope. There’s also the small problem of neuron behavior modeling. Our existing biochemical and information-theoretical models are a very good match for the behavior of neurons.

The urge of tacking on some unneeded extra mystery baggage to distort a set of models that were already both well rooted in empirical data and accessible to rational analysis is an act of religiosity. It comes from, and speaks to people who share, a deep instinctive bias towards a metaphysical theory of consciousness – presumably because a purely physical model would be considered too mundane. The mere wish for a literal metaphysical transcendence does not make it true though, and in my opinion it also trivializes the amazing hardware that allows us to experience life.

Articles like these emphasize the need to always judge scientific claims on their objective merits, as opposed to looking at who made the claim which seems to be the only consideration when it comes to science reporting. For instance, Roger Penrose is mostly a darn good physicist, but he’s also a quack at the same time. Without skeptically looking at his work on a case-by-case basis it’s impossible to distinguish the good parts from the pseudo-religious bullshit. It sounds tedious but I promise: you’ll have an easy time categorizing these things.

Udo’s Twitch Streaming Guide for Mac

I don’t know how many Ludum Darers are Mac users, but it’s sure not common to see them doing Twitch feeds. Part of the problem might be that it takes a lot of effort and fiddling to set it up properly. So for the benefit of everyone who’s interested in doing it, here’s what I’ve learned.

You’re going to need a few pieces of software, but the good news they’re all free or low-cost.

Screen Capture: CamTwist (free)

You can get it at It does a pretty decent job of capturing video content off the screen in real time, and it also can do some basic compositing. It’s not the most stable software, however, so I recommend running it at a low frame rate. Open CamTwist and double-click on Desktop, full screen, and if necessary select the screen you want to capture.


Audio Bus / Routing: Soundflower (free)

Download Soundflower from and install it. It comes with SoundflowerBed, a simple config application. Start it and route the 2ch bus to the speaker.


Open System Preferences and select the 2CH bus as your default audio output. This causes all apps to send their audio to Soundflower 2CH. Since you already routed 2CH to your speakers earlier, this should enable you to hear application sounds normally – they’re just routed through the Soundflower bus now. If you can’t hear anything, fiddle with the volume – including the input volume of the 2CH bus (which you can also find on the same settings window).


That’s the most basic audio setup imaginable, we’ll get to the good stuff later. All this enables you to do right now is stream your screen and all your audio output. We just need an app to take this data and send it to Twitch:

Streaming: Flash Media Live Encoder (free)

You can download the Media Live Encoder from Adobe. It’s a hassle though, so you might want to make sure to save your install image somewhere safe for later in case you need it again. Then open your Twitch account in a browser and download the streaming profile which they conveniently provide as a download. It’s an XML file, store it somewhere convenient. Start the Live Encoder, go to the File menu and load this XML file. It sets up the basics for streaming including your private streaming key, but you can still fiddle with the settings.

Select CamTwist as the video device, and Soundflower 2CH as the audio device.


If everything is set up correctly, you’re going to see a preview of your video stream showing your desktop, and the audio bars to the left of the preview should fluctuate when you play some music. Hit the “Start” button and see if you can stream to Twitch alright.

Congrats, you have completed a basic setup! But there’s more. You might want to also talk on your videos, and you might want to mix output levels of different apps a little. Sadly and incomprehensibly, OS X has no way of doing this on its own, so we need to download more software:

Microphone Routing: LineIn (free)

Download LineIn from and install it. It allows you to route input from your microphone into the Soundflower bus. However, LineIn has a stunning limitation: it can only do one route at a time. This is bad because you don’t want to route your microphone input to your speakers (it’ll cause a nasty feedback screech), you only want to route it to the actual live stream. Luckily, you can trick LineIn to do this by duplicating the app. That’s it, just use “Duplicate” from the context menu and you’ll have 2 LineIn programs. Launch them both.

With the first instance, route your microphone to Soundflower 16CH. With the second, route Soundflower 2CH to 16CH:


Activate Pass Thru on both, you should now see the indicator bars lighting up when you make a sound in the room.

Go back to Adobe Media Live Encoder and choose 16CH as your audio device.

With this maneuver we have tricked both Soundflower and LineIn into doing something their developers didn’t have in mind, but it works beautifully: the 16CH bus is now strictly for our streaming, but the 2CH bus contains all non-microphone audio which we can directly listen to. To stop streaming the mic, just deactivate “Pass Thru” on the first LineIn instance until you need it again.

Mixing: SoundBunny ($10)

Per-application audio levels is again something OS X can’t do on its own, we’re going to need yet another app for that. Go download SoundBunny for Mac, it’s $9.99.


Now you get these nifty sliders that allow you to actually control sound levels individually.

And that concludes my Mac streaming guide. As you can see due to stunning limitations of the OS X audio system we need a lot of extra cruft to accomplish some basic mixing and routing, but these apps I showed here do an excellent job of it. The only downside is you need to have them all running, and it takes a few minutes from cold start until you’re all set up and ready to stream. Remember to save your settings, especially the Media Live Encoder ones, for the next time.

Webcam Window: CamTwist or VLC

One little extra thing I like to do is show a little webcam feed in the corner as well. You can do this with CamTwist, but it crashes my system when I try it. If that’s also a problem for you, use VLC. Click on “Open Capture Device” in the menu and you’ll be able to select your Mac’s camera as an input device. Use VLC’s settings to make the video window always-on-top and remove the window’s title bar. It should now be an unobtrusive little frame that you can place anywhere you want on your streaming desktop.



Rampant Artificial Locality on the Internet

urlPretty much every major site will completely ignore not only your browser’s language settings but also onsite user account settings and other desperate attempts at selecting the language. The only thing that matters is your IP address.

For example, accessing Google: my browser is set to accept English only. I’m entering the English URL. In my account settings I periodically reset everything I can find to English (settings apparently decay, too). Google knows I want the English version. Yet, they still give me the interface in whatever language my IP address comes from. And not only the UI, search results as well.

Recently it’s gotten even worse than that: Google figured out I’m actually German, so they start defaulting to German more often now – ignoring everything else. At least with the IP address-based routing it was impersonal.

I happened to be in Sweden when I linked my Facebook calendar to my Google calendar. Ever since that day, my friends’ birthdays are given to me in Swedish. Facebook knows I want English, yet for some reason this is how it’s got to be.

The same abuse is apparently considered best practice at new startups as well: recently I was testing a browser game for an acquaintance who’s on their development team. Because I was in Portugal at the time, I of course got the site in Portuguese. Manually switching that to English, the game still started up in Portuguese. It’s been doing that ever since. Every email I get from that company is in Portuguese, too, even though I tried everything I could to set my language to English.

It’s a source of endless frustration, maybe even a hostile act. They’re effectively saying “Your choices don’t matter, we know what’s best for you. You’re from country X, so you _must_ speak Xish. People are on the internet to enjoy regional separation. Really, it’s best.

The Rise of Ubiquitous Object-to-Anything Mappers

I know the use of object-to-HTML-mappers is only going to grow from here on out, but I really don’t see the appeal in most cases where pure HTML would have been a perfectly readable and lightweight choice. We’re becoming so scared of angle brackets that we seek refuge behind object-like constructs that are nothing but yet another impoverished and leaky abstraction layer above an already abstract markup language.

Look at that nightmare of a rendering code and tell me you’ll still want to touch that when the time comes to go back and modify it in a few months.

I had the displeasure of working with Jade on a recent project. Not only was it more difficult for humans to parse than the HTML it produced, it was also slow to execute and generally cumbersome to modify. The main purpose of its existence might just be so programmers can lie to themselves about separating code from templates, and making those templates look really sophisticated at the expense of productivity.

The battle cry “nobody should write literal HTML anymore” is not a good enough reason to do this – in fact it’s not even a reason at all to begin with.

The example code in the article, well, I’m sorry to say that it’s unreadable, convoluted, and huge for what it actually does. Both versions are terrible in their own way, only the JavaScript one is also deliberately sabotaged in order to really stand out as an abomination.

That being said, if there are people looking at this going “my, that does look mighty pleasant” – all the more power to you. A huge part of programming is finding technologies that map well to your brain. It just doesn’t work for me at all (I needed to vent that just now, thanks for bearing with me) and I kind of hope I’m not alone.