AI Safety Debate with Roko

Twitter Space to Clarify Why I’m Opting for Effective Acceleration (with Caveats) 

Even as GPT-3/4 has semi-mainstreamed it, the AI timelines discourse has become sharply riven between AI safetyists and “effective accelerationists” in the past several months.

In this Twitter space from Feb 27, 2023, two representatives of each side – Roko Mijic and myself, respectively – attempt to reconcile these philosophical differences.

I am going to write up my arguments in more detail sometime later, but they boil to the following:

  • Bootstrapping God-like intelligence as a single agent is probably very difficult to unfeasible. Alignment is hard and this would apply to a malevolent AGI’s agents as well, who will develop their own values and interests.
  • Maximizers are irrational and will be outcompeted by more rational agents, assuming they are given space to actually flourish and develop.
  • There are a multitude of risks inherent in creating a singleton (for that is what is needed) to manage AI safety in a coherent global fashion. These risks involve:
    • Lost opportunities in productivity gains and poverty alleviation, which results in real damage to welfare on account of theoretical blog posts.
    • Strongly reduced chances of achieving radical life extension.
    • Long-term sector capture and AI safety’s transformation into a quasi-religious cult, as occurred with civilian applications of nuclear power and explosions.
    • The AI sector’s transformation into the noospheric equivalent of a monoculture ecosystem, which is inherently more fragile to shocks and probably voids any dubious benefits of restrictive AI regimes.
    • Potential stagnation and even retreat in rates of technological growth, due to long-term dysgenic trends.
    • This period will be one in which existential risks of other kinds will be in play and not necessarily at a constant background rate.
  • The very fact we’re experiencing these observer moments suggest that they are extensively recalled or simulated in posthuman worlds, which suggests we are on the cusp of a good singularity.

The Z of History: 13 Months of Commentary

As the first post on my podcast blog at Substack, I would like to take the opportunity to highlight my discussion with Noah Carl the prospects of each side in the Ukraine war.

My basic thesis is summarized in this thread:

Most everything I said there still applies as of March 20, with the exception that I’m now somewhat more bearish about the prospects of Ukrainian offensive success, which just goes to further confirm the “long stalemate” thesis that can only be broken by an OOM-scale increase in NATO supplies or the implementation of a war economy to and by Ukraine and Russia, respectively.

Regarding the Ukraine War, other media appearances on this topic I have done this year to date include an interview and a round table with Mazarin Geopolitics, the latter including a couple of pro-Ukrainian Russians with rosy expectations of Ukrotriumph that have yet to materialize, and more recently, a discussion with the podcasting duo Russians with Attitude, where I conversely took on the role of dampening Z expectations.

My recently refurbished website akarlin.com now has an AK’s Podcasts archive and page that will be progressively filled out over the coming weeks.

New podcasts will appear at akarlin.com as well as here on “Shock and Disbelief” if they are sufficiently important.

Military-Technical Decommunization

Vladimir Putin: “We Are Ready to Show What Real Decommunization Would Mean for Ukraine”

Since my article last week predicting the imminent “Regathering of the Russian Lands”, the prospect of a large-scale Russian invasion has gone from ambiguous to extremely likely (90% on Metaculus). Personally, I think it’s a foregone conclusion, with operations beginning either tonight or tomorrow night, with the most interesting and important questions now being the speed of the Ukrainian collapse, the future borders and internal organization of Russian Empire 2.0, and the ramifications of the return of history on the international order.

February 22, 2022 will indeed enter history as the day when Vladimir Putin decided to become a Great Man of history. In an hour long speech, he basically recounted his magisterial July 2021 article on the historical unity of Russians and Ukrainians, officially endorsing the nationalist position that Russia is the “world’s largest divided big nation”. He stated that the modern Ukrainian state can be rightfully called “Vladimir Lenin’s Ukraine”, asserted that its statehood was developed by the Bolsheviks, and noted the irony in Ukrainian nationalists toppling statues to their father. “You want decommunization? Very well, this suits us just fine. But why stop halfway? We are ready to show what real decommunizations would mean for Ukraine.

[Read more…]

Regathering of the Russian Lands

Already in 1990 I wrote that Russia could desire the union of only the three Slavic republics [Russia, Ukraine, Belarus] and Kazakhstan, while all the other republics should be let go. It would be desirable if [a resulting Russian Union] could be formed into a unitary state, not into a fragile, artificial confederation with a huge supra-national bureaucracy. – Alexander Solzhenitsyn.

The Empire, Long Divided, Must Unite

There is a good chance that the coming week will either see the culmination of the biggest and most expensive military bluff in world history, or a speed run towards Russian Empire 2.0, with Putin launching a multi-pronged assault invasion of Ukraine to take back Kiev (“the mother of Russian cities”) and the historical provinces of Novorossiya.

There is debate over which of these two scenarios will pan out. The Metaculus predictions market has given the war scenario a 50/50 probability since around mid-January, spiking to 60-70% in the past few days. This happens to coincide with the public assessments of several military analysts: Michael Kofman and Rob Lee were notably early on the ball, as were some of this blog’s commenters, e.g. Annatar. The chorus of skeptics is diverse, but includes Western journalists and Russian liberals who tend to believe Putin’s Russia is too much of a cynical kleptocracy to dare go against the West so brazenly (e.g. Oliver Carroll, Leonid Volkov); Western Russophiles who are all too aware of and disillusioned with hysterical media fabrications about Russia, and are applying faulty pattern matching (e.g. Michael Tracey); and Ukrainian activists who have spent the last eight years hyperventilating about “Russian aggression” and have been reduced to shock and disbelief now that the real thing is staring in their face.

For the record, my own position is that the war scenario was ~50% probable since early January, might be as high as 85% now, and it will likely happen soon (detailed Predictions at the end).

My reasons for these bold calls can be sorted into four major bins:

  1. Troops Tell the Story: What we have observed over the past few months are all completely consistent with invasion planning.

  2. Game Theory: Russia’s impossible ultimatums to NATO have pre-committed it to military operations in Ukraine.

  3. Window of Opportunity: The economic, political, and world strategic conjuncture for permanently solving the Ukraine Question has never been as favorable since at least 2014, and may never materialize again.

  4. The Nationalist Turn: “Gathering the Russian Lands’ is consistent with opinions and values that Putin has voiced since at least the late 2000s, with the philosophers, writers, and statesmen whom he has cited most often and enthusiastically (e.g. Ilyin, Solzhenitsyn, Denikin), and more broadly, with the “Nationalist Turn” that I have identified the Russian state as having taken from the late 2010s.

I will discuss each of these separately.

[Read more…]

Review: Wheel of Time S01

Wheel of Time S01 (2022)

The Rafeverse isn’t a different turning of the Wheel as Rafe and Sanderson have claimed, nor even a Turning in which the Dark One won as some have suggested here (if that had happened, he would have been free in all worlds, at all times), but a Mirror World or World That Might Be.

The distinguishing feature of these Mirror Worlds is that while they are possible worlds, their appearance and sustained existence is improbable in the extreme. Stronk women taking down Trollocs with a pocket knife commando-style, while a blademaster can’t kill a single one. Globalized cosmopolitan age levels of ethnic heterogeneity in podunk villages that haven’t received more than a couple of peddlers per year for a millennium. Social mores of a late liberal society persisting after an apocalyptic total war and 3,000 years of upheavals and decivilization. “Darkfriends” managing to erase mention of the Eye of the World from Tar Valon’s libraries

Causality in this world is broken, with all its attendant effects on world self-consistency. Incidentally, this also explains the very low IQ of the characters in the show. Intelligence is only adaptive in worlds governed by consistent rules that can be figured out and then exploited for a competitive advantage. In a world in which an Aes Sedai can’t stop Whitecloaks from burning her at the stake while a bunch of untrained wilders destroy an entire Trolloc horde, or in which a village Wisdom can follow an Aes Sedai’s “tell” which her own Warder cannot, there is no significant payoff to intelligence, hence it was never selected for there. In this sense, Lan is actually rational and smart for not wasting his time training any of the boys in how to use their weapons, this is not how XP is actually gained in this world. He, at least, is fully cognizant of how his world works, and navigates it efficiently.

I would say that the aesthetics of this world tends to back up this theory. It has a washed out look, lack of attention to detail (costumes that spontaneously clean themselves), empty spaces, near empty sets, inconsistent distances and timelines, scales and measures that have no anchor in objective reality, and extreme warped perspectives, as when our heroes go for a Sunday jaunt into the Blight and Trollocs emerge to attack the Gap a few hundred meters behind them (in a normal world, this would beg the question of how they managed to avoid getting caught up in that flood, but not one in which time and distances “bend” in arbitrary ways as in the improbable Mirror Worlds).

One prediction we can make from this is that if the Mirror World theory is true, then it is an already highly unstable and indeed “fragile” existence, and one that may well unravel completely when balefire is weaponized again and breaks the already seeping chains of causality that hold reality in place beyond some critical tipping point. The likeliest point for that to happen is in connection with certain events at the Stone of Tear, i.e. the presumed end of Season 2.

Instead of holding anger against Rafe and the showrunners, I would suggest instead sparing a thought and extending some compassion towards the benighted denizens of this Mirror World, who live tormented and twisted lives with no understanding of how things are really meant to be, and whose very existence will probably soon end, at least bringing with it the small mercy of a final release from the permanent psychosis in which they are forced to live.

(Original).

WAGMI: The %TOTAL Maximizer

This short story about “AI on blockchain” was originally published at my Substack on November 24, 2021. I subsequently updated it and republished it here, as well as at my Mirror blog Sublime Oblivion, where you can collect it as an NFT.

In the half a year since I wrote it, the concerns implicitly raised in it have become far more acute. Predicted timelines for the appearance of “weak AGI” at Metaculus have compressed sharply in the past few months, as the scaling hypothesis – that is, the concept that banal increases in model size lead to discontinuous increases in capability – has failed to break down. Meanwhile, there are now projects aiming to implement fully on-chain AI. One such project is Giza, which wants to build them on top of a second/third layer on Ethereum that allows for intensive computations while inheriting Ethereum’s security and decentralization.

Why is putting artificial intelligence on chain a good idea?” asks one piece on this topic. Why not indeed. Think of the possibilities! 😉

[Read more…]

Moscow’s Pacification

Moscow’s Murder Rate Now Lower Than “Prestigious” London’s

For the first time possibly since the late Middle Ages (for Britain had embarked on “pacification” – the vertical reduction of homicide rates, by dint of increasing state capacity, genetic selection, or both – centuries earlier than Russia), Moscow will very likely have a lower homicide rate this year (2021) than London. London had 1.5/100k murders in 2018, the last year for which we have population estimates; on current trends, it should finish up at around 1.4/100k this year (possibly more, if Corona-era projections of population decline are accurate). Moscow registered either 1.6/100k homicides [Rosstat] or 1.4/100k homicides [Prosecutor-General] in 2020. In the year to date (to August), the number of homicides has fallen by 21%. Either way, at somewhere around 1.1-1.3/100k, Moscow’s homicide rate is now lower than “prestigious” London’s.

[Read more…]

Open Takes 1

Higher School of Economics, Moscow.

Kicking Off a New Era of Powerful Takes

As announced this Friday, I am leaving The Unz Review.

Nothing is set in stone. I am considering various alternatives, from resurrecting my website as an active blog, to more exotic Web 3.0 options, such as urbit, where a WordPress clone might be ready as early as EOY. However, I suspect that most of my future writings, at least in the medium-term, will be on this Substack.

As such, if you’re interested in following my work, I would suggest you:

The frequency of new posts will drop, as befits what will now be more of a “newsletter” than a “blogging” format, though I will continue posting weekly Open Threads (henceforth, Open Takes) to serve as a focal point for the community that has aggregated around my scribblings. Going forwards, I will also be privileging “effortposts” such as longreads and book reviews, while shorter content will henceforth be relegated to Twitter.

Paying subscribers will receive additional benefits. I will work out the details in 2-3 weeks’ time.

Meanwhile, welcome to the first Open Takes on Powerful Takes!

[Read more…]

The Last Reaction

It was never my mission to pursue “activist” goals so much as to try to accurately understand and explain how the world works, and at best, play some  modest role in informing the debate in those areas that I hoped could make use of some of my insights.

From that perspective, my record of my “Russian reactions” in my past almost seven years at The Unz Review has been a mixed one:

  1. Despite its stellar predictive power, from development economics to the Karabakh War of 2020, the HBD/”cognitive capitalist” worldview is “unhandshakeworthy” as never before, with Wokeness – #BLM, CRT, identity politics – having become America’s secular religion. This will probably create a lot of damage before sobriety returns, which will likely take a quite a while, as much of the “Dissident Right” seems to have decided that the correct response to SJW overreach was one upping them with Qanon and other very powerful theories.
  2. Assiduously as I tried to “explain” Russia, one blogger was never likely to get very far, given that the underlying dynamics were always driven by the exigencies of American domestic politics (“the game was rigged from the start“).  Hence, the sudden disappearance of Russiagate as soon Trump was out of power. Consequently, we have the supreme irony of Russo-American relations being less bad under Biden, under whom strategic priority has shifted to China, than they were under “Putin puppet” Trump.
  3. Conversely, on the “bright” side – at least so far as many Russians are concerned – is that Putin has adopted my own program for Russia, progressively outlined over the past half decade, almost wholesale (hence my recurrent jokes about him reading my blog). My last longread for The Unz Review both marked and celebrated the culmination of that process, and I hope it will serve as a thematically appropriate endpoint to the Russian Reaction blog.

“You win some, you lose some.” :shrug: But I’m satisfied to have been dealt an American Airlines on the particular “debate” that is most germane to my own life choices.

[Read more…]

Open Thread 167

  • In his latest newsletter, Adam Tooze points out that the Chinese government is asserting greater state control over the economy, including the power of Chinese business magnates to “cash out” of their holdings.

In retrospect, this is perhaps the most logical explanation for the crypto crackdown.

  • Diana Fleischman has a good article in Quillette on how the Leftist moral panic against eugenics has given ammunition anti-abortion activists, with apparently six states now banning women against abortion on the basis of congenital disability. Interesting example of how an SJW – rightoid horseshoe, even in matters so small, helps usher us along towards Idiot’s Limbo with some combination of more disabled people, more restrictions on prenatal testing and genetic screening, a reduction in reproductive rights. Noah Carl notes most of the pointing and sputtering it generated came from left-wing progressives.
  • Mark Galeotti – Kremlin Looks to Establish a ‘Techno-Authoritarian’ Power Vertical 2.0. Seems like a move in the Chinese direction of digitalized, indices-based control over regional governance (along the lines of Mishustin’s reforms of the tax sector).
  • Steve Sailer on new FBI stats showing a 29% rise in murders in 2020. Incidentally, the gap between the US and Russia is now possibly larger than at any time since the Revolution.
  • The Guardian lumps Steve in with Jeffrey Epstein. #AJAB
  • Paul Robinson covers a report which calculates that the incidence of Russian military interventions abroad under Putin has actually declined relative to the Yeltsin era.

* Silventoinen, K. et al.(2020). Genetic and environmental variation in educational attainment: an individual-based analysis of 28 twin cohorts. Scientific Reports, 10(1), 12681. (h/t Steve Sailer). Contra Herrnstein/Murray, the heritability of educational attainment may have actually declined during the second half of 20C. Was the idyllic (to some?) picture of old time America in which people who marry across cognitive barriers a mirage?