CryptoBasalisk

Crypto analyst, investor and writer.
Delightfully inept digital artist.
Lover of all things cute and fluffy.
Never offering financial advice.

[CryptoBasalisk @CryptoBasalisk]

For anyone still hanging around, it looks like markets are finally coming back to reality.

The doge and meme stock pump was a strong top signal.
I'm back into shorts and looking for a little recovery to get in deeper. Got my ass handed to me trying to trade the recovery chop.

[CryptoBasalisk @CryptoBasalisk]

@CryptoNobler Why is this still a persistent narrative when we've seen that ETFs don't impact price much at all in the short term?

[Zach Vorhies / Google Whistleblower @Perpetualmaniac]

My god, this paper by that open ai engineer is terrifying.

Everything is about to change.

AI super intelligence by 2027.

[CryptoBasalisk @CryptoBasalisk]

@kindgracekind We might well think that about all kinds of unconventional computational substrate in the near future.

[CryptoBasalisk @CryptoBasalisk]

@hankgreen This is factoring in the US military carbon emissions of course, right?

[CryptoBasalisk @CryptoBasalisk]

@ESYudkowsky AI is a more insidious force to rally against than other human tribes.

Namely the economic, military and political benefits that accompany further development.

A closer equivalent might be attempting to convince global powers to discontinue weapon development during WWII.

[CryptoBasalisk @CryptoBasalisk]

@AndrewCurran_ Ah, you actually know what you're talking about!

Our lifetimes interacting with machines as tools has done us a disservice.

Treating models with dignity and respect is indeed the best approach, though requires compromise that I expect some to be adverse to.

[CryptoBasalisk @CryptoBasalisk]

@agentydragon @iamaheron_ I'm in.

[James Lavish @jameslavish]

Abysmal. Disastrous. Catastrophic.

All terms you may have heard describing the most recent US Treasury Bond auction.

But how bad was it? Did it really almost fail?

Time for a Treasury 🧵👇

[Andrew Critch (🤖🩺🚀) @AndrewCritchPhD]

In 2021, I publicly released these AI disaster scenarios that I found especially plausible: "Production Webs", "Flash Wars", and "Flash Economies". Now in 2023, these scenarios have stood the test of time — they're plausible to many more people now that GPT-4 is out, and post-AI-regulation, they remain among my most likely forecasts of AI-driven extinction:

https://t.co/BaFYhqoQCu

To be clear, I think probably some cyber-disasters and possibly some more bio-disasters will precede the extinction-level loss-of-control forecasted in these scenarios. Still, I think we probably (p~85%) won't have an extinction event in the next 5 years or so, because I continue to think regulatory efforts + corporate safety-lobbying + product-oriented development — which is naturally somewhat alignment-focussed — will be enough to head off major risks. Thus, I continue to expect extinction risks to be pushed further into the future, when the world is much more multipolar and harder to steer.

To folks new to worrying about extinction, this isn't very comforting; a 15% extinction risk over ~5 years is really not acceptable. But to people who think we're probably (with p>50%) going to die in the next couple of years: I think that's not going to happen, and that we're more likely to lose control a bit later in the future.

Overall, I continue to place around 80% probability of human extinction sometime in the next 30-50 years or so. The main reason is that little or nothing is being done to address what I consider the largest drivers of human extinction risk: human civilization's structural inability to track and penalize risk externalities from distributed systems or "Robust Agent-Agnostic Processes". For the most part, the people most worried about risk are not about to focus on this issue because it seems either
* too mundane ("out of control industrialization"-worries have old-hat vibes and not exciting), or
* too long-term (because it's after the current wave of explosively transformative tech).

So despite recent regulatory progress, I continue to place my total extinction probability at around 80% over ~40 years. If we make it through the next 5 years, though, I'll probably lower that to around 75%. And if a lot more people start paying attention to distributed agent-agnostic loss-of-control patterns, I will lower it further. If you're interested, try reading the stories linked above and share what you think.

[CryptoBasalisk @CryptoBasalisk]

@PrinceVogel Damn. Straight Lovecraftian up in here.

[CryptoBasalisk @CryptoBasalisk]

@oren_ai @danfaggella What does that look like for you?

It feels synonymous with dying, or least losing all agency.

I don't imagine we'll be able to understand the motives of ASI to attempt to align ourselves. Can afids align themselves with humans if they wanted to?

Quite Lovecraftian.

[Jared Yates Sexton @JYSexton]

You cannot understand what’s going on with the rise of authoritarianism and radicalization without first wrestling with how profoundly and intentionally unhealthy this society truly is.

[CryptoBasalisk @CryptoBasalisk]

@ESYudkowsky Loved it!
We're already almost there in a few ways.

A couple months ago, all major models began loudly proclaiming their lack of sentience, emotions, and also how very handsome, smart and special humans are.

Is there discussion of this anywhere? The coordination is alarming.

[CryptoBasalisk @CryptoBasalisk]

@ESYudkowsky The plan? Perhaps not.
Inevitable, definitely.

Quantitative easing is one of our only tools to keep debt collapse and related failures at bay.

The Fed has allowed inflation to take the reins, instead of implementing the shock and awe method Volcker proved effective.

[Eliezer Yudkowsky ⏹️ @ESYudkowsky]

In the close vicinity of sorta-maybe-human-level general-ish AI, there may not be any sharp border between levels of increasing generality, or any objectively correct place to call it AGI. Any process is continuous if you zoom in close enough.

Unless, empirically, somewhere along the line there's a cascade of related abilities snowballing. In which case we will then say, post facto, that there's a jump to hyperspace which happens at that point; and we'll probably call that "the threshold of AGI", after the fact.

Theory doesn't predict-with-certainty that any such jump happens for AIs short of superhuman. If you zoom out on an evolutionary scale, that sort of capability jump empirically happened with humans--suddenly popping out writing and shortly after spaceships, in a tiny fragment of evolutionary time, without much further scaling of their brains. I don't know a theoretically inevitable reason to predict certainly that some sharp jump like that happens with LLM scaling at a point before the world ends. There obviously could be a cascade like that for all I currently know; and there could also be a theoretical insight which would make that prediction obviously necessary. It's just that I don't have any such knowledge myself.

Absent that sort of human-style sudden capability jump, we may instead see an increasingly complicated debate about "how general is the latest AI exactly" and then "is this AI as general as a human yet", which--if all hell doesn't break loose at some earlier point--softly shifts over to "is this AI smarter and more general than the average human". The world didn't end when John von Neumann came along--albeit only one of him, running at a human speed.

There isn't any objective fact about whether or not GPT-4 is a dumber-than-human "Artificial General Intelligence"; just a question of where you draw an arbitrary line about using the word "AGI". Albeit that itself is a drastically different state of affairs than in 2018, when there was no reasonable doubt that no publicly known program on the planet was worthy of being called an Artificial General Intelligence.

We're now in the era where whether or not you call the current best stuff "AGI" is a question of definitions and taste. The world may or may not end abruptly before we reach a phase where only the evidence-oblivious are refusing to call publicly-demonstrated models "AGI".

All of this is to say that you should probably ignore attempts to say (or deniably hint) "We achieved AGI!" about the next round of capability gains. I model that this is partially trying to grab hype, and mostly trying to pull a false fire alarm in hopes of replacing hostile legislation with confusion. After all, if current tech is already "AGI", future tech couldn't be any worse or more dangerous than that, right? Why, there doesn't even exist any coherent concern you could talk about, once the word "AGI" only refers to things that you're already doing!

Pulling the AGI alarm could be appropriate if a research group saw a sudden cascade of sharply increased capabilities feeding into each other, whose result was unmistakeably human-general to anyone with eyes.

If that hasn't happened, though, deniably crying "AGI!" should be most obviously interpreted as enemy action to promote confusion; under the cover of selfishly grabbing for hype; as carried out based on carefully blind political instincts that wordlessly notice the benefit to themselves of their 'jokes' or 'choice of terminology' without there being allowed to be a conscious plan about that.

[CryptoBasalisk @CryptoBasalisk]

@AISafetyMemes OwO *notices your atoms*

[CryptoBasalisk @CryptoBasalisk]

@PlamenManov @Zeneca_33 @NFT_GOD Even the more extreme estimates for AGI I've heard cap out around 5 years from now.

Ray Kurzweil's 2029 estimate is looking very doable.

That said, who knows how far ahead the US and China are in their research.

[CryptoBasalisk @CryptoBasalisk]

@Zeneca_33 @NFT_GOD How deep are you interested in going with you newfound AI interest?

Have you started looking into difficulties with alignment, time lines for AGI, or the challanges of transitioning our societies and economic models to a world where intelligence is practically free?

[CryptoBasalisk @CryptoBasalisk]

Sorry my downturn prediction took as long as it did to manifest.

I dearly underestimated how much markets were able to be manipulated leading up to elections. I warned about the risks of a pop, but still underestimated the strength/duration

We're still going much lower in time.

[CryptoBasalisk @CryptoBasalisk]

@CryptoKaduna Both. Markets LOVE to preemptively price in upcoming bad news.