avturchin's Shortform

Lifehack: If you're attacked by a group of stray dogs, pretend to throw a stone at them. Each dog will think you're throwing the stone at it and will run away. This has worked for me twice.

OpenAI whistleblower found dead in San Francisco apartment.

Suchir Balaji, 26, claimed the company broke copyright law.

↑ comment by lc · 2024-12-14T20:01:01.184Z · LW(p) · GW(p)

Tapping the sign:

Collapse of mega-project to create AI based on linguistics

ABBYY spent 100 million USD for 30 years to create a model of language using hundreds of linguists. It fails to compete with transformers. This month the project was closed. More in Russian here: https://sysblok.ru/blog/gorkij-urok-abbyy-kak-lingvisty-proigrali-poslednjuju-bitvu-za-nlp/

"Bird Flu H5N1: Not Chaos, but Conspiracy?" By Alexander Pruss 
Two months ago, I was puzzled how bird flu, potentially capable of killing tens of millions, went rampant on American livestock farms and began infecting workers, yet no urgent measures were being taken. Even standard epidemiological threat monitoring was happening unsystematically, with months-long delays, and results weren't being made public for months afterward. What happened to the bitter lessons from the coronavirus pandemic? Why such chaos? Since then, the sense of criminal inaction has only intensified. Missouri discovered the first outbreak of human cases unrelated to farm workers, but molecular testing was neglected and infection paths remained undiscovered. 

In California, a more pathogenic variant of bird flu spread to hundreds of dairy farms, reportedly killing up to 15% of cows, with almost daily new cases of virus transmission to humans. The virus apparently came to California through cattle transportation from Idaho, despite belatedly introduced rules formally prohibiting the transport of infected cows across state lines. The problem was that infection in transported cows was checked through selective testing, and as reported, the sampling wasn't random: before government testing, farmers secretly tested cows for bird flu in private laboratories and selected only healthy ones for official testing. Here's the continuation of the translation:

A new Vanity Fair investigation shows this isn't random chaos. The USDA (U.S. Department of Agriculture) has been blocking research and data about the new infection in America's dairy herds from the start to protect the multi-billion-dollar American dairy export industry and the interests of giant national dairy processing companies. The idea was simple: most cows recover after a few weeks, and while the bird flu virus does get into milk in huge quantities, it should die during pasteurization. Therefore, the economic losses from the pathogen aren't that severe. However, if consumers in America and especially abroad raise the alarm, it could result in much greater dollar losses. USDA Secretary Thomas Vilsack knows this firsthand: before his government appointment, he worked as president of the U.S. Dairy Export Council. 

And immediately after it was finally discovered in March 2024 that dairy farms in Texas and Kansas were hit by bird flu, veterinarians and state officials began receiving calls from personal mobile phones of USDA veterinary institute workers: "we're officially forbidden to discuss this problem without permission from the very top, and unofficially we're asking you to keep quiet about it too." But what about the danger that the virus, having settled in mammals and especially humans, could recombine with our seasonal flu viruses and produce hybrid viruses that combine the infectious and pathogenic potential of human viruses with immunity to our regular antibodies inherited from their avian ancestor? 

This, generally speaking, isn't USDA's concern. This alarm was raised by the White House Office of Pandemic Preparedness and Response (OPPR), created in 2023, under the leadership of military doctor and biosecurity expert Paul Friedrichs. In early April, dairy industry representatives raised concerns that some upstart from the White House was muddying the waters. USDA's response was their new policy of official secrecy. Secretary Vilsack responded only a month later to state veterinarians' inquiries about the sudden communication breakdown, and his response was essentially a brush-off. And his ally in Texas, state agriculture commissioner Sid Miller, even hinted that if Friedrichs' people stick their noses into Texas farms, they might be met with bullets. 

A number of veterinarians who disagreed with USDA's actions soon lost their jobs, and the country fell into an atmosphere of "work-to-rule," where veterinary authorities appear to be doing their job, but as slowly as possible and with all the red tape that can be justified by regulations. Meanwhile, flu season is approaching, and encounters between bird and human flu in people infected with both viruses are inevitable in the near future.

I forgot to add that by May, a vaccine for bird flu became available for cows, but the USDA chose not to use it.

Roman Mazurenko is dead again. First resurrected person, Roman lived as a chatbot (2016-2024) created based on his conversations with his fiancé. You might even be able download him as an app. 

But not any more. His fiancé married again and her startup http://Replika.ai pivoted from resurrection help to AI-girlfriends and psychological consulting. 

It looks like they quietly removed Roman Mazurenko app from public access. It is especially pity that his digital twin lived less than his biological original, who died at 32. Especially now when we have much more powerful instruments for creating semi-uploads based on LLMs with large prompt window.

Igor Kiriluk (1974-2022)

Igor was an organiser the first meet-up in Moscow about effective altruism around 2013. Today his body was found at his home. The day before he complained about depression and bad health. His cryopreservation now is being organised.

He was  also a one of four organisers of Russian Transhumanist Movement, along with Danila Medvedev, Valeria Pride and Igor Artuhov around 2003. 

His main topic of interest was paradise-engineering. He translated works of David Pearce.

He may look detached from reality but he was first to react on new ideas and has very large network of friends everywhere: between visionaries, scientists and officials. Being a great networker, he helped many people to find each other, especially in the field of life extension.

His FB page: https://www.facebook.com/igor.kirilyuk.3

I am building my sideload via recursively correcting of 1-million-tokens prompt for large LLM. The prompt consists of 500 rules which describe my personality, similar to personal constitution, and of some texts, like diaries, abstracts, poetry, stream of thoughts etc. Works on Google Gemini 1M through Google AI studio, and the shorter version works great on Opus. The system also includes a universal "loader prompt" which tries to increase the intelligence of the model and describes how the chatbot should work.

I found that sideloading allows very quick iterations in the sideload's improvements and the improvements are two-fold: of the loader itself and improvements of the knowledge and style of the sideload.

I find that my sideload is surprisingly good for a project which took around 1 month of work. 1 of the 5 answers is exactly like mine from a factual and style point of view. 

I am open-sourcing my sideload, anyone can run it https://github.com/avturchin/minduploading/tree/main

I can help anyone interested to build his-her own sideload.

Example of work of the chatbot, no cherry picking:

Q:(now speak in english) what will be your next post in Lesswrong?

A: /The user is asking about my next LessWrong post. I will answer as if I am Alexey Turchin, and describe my plans for future posts, as well as my current location and the people around me./

My next LessWrong post? [Alexey looks thoughtfully out the window]

(I have several drafts)

[Alexey stretches and sits up in bed]

{The room is small, with a balcony. On the table is a laptop, some papers, and reading glasses. Outside the window, it is still gray and dreary.}

I'm thinking about writing about the nature of qualia and the possibility of creating qualia of non-existent things. [Alexey gets up, walks to the window, and looks out at the street]

{Below, a woman with a child is walking, a street sweeper is lazily sweeping leaves. Somewhere in the distance, the sound of a siren can be heard.}

Or maybe about the problem of copies and identity. (it's a complex topic)

[CONFIDENCE] 8

New b.1.640.2 variant in France. More deadly than delta. 952 cases of which 315 on ventilator.

https://www.thailandmedical.news/news/breaking-updates-on-new-b-1-640-2-variant-spreading-in-southern-france-number-of-cases-growing-and-variant-now-detected-in-united-kingdom-as-well

https://flutrackers.com/forum/forum/europe-aj/europe-covid-19-sept-13-2020-may-31-2021/933598-southern-france-reports-of-new-variant-with-46-mutations

ChatGPT 4.5 is on preview at https://chat.lmsys.org/ under name gpt-2. 

It calls itself ChatGPT 2.0 in a text art drawing https://twitter.com/turchin/status/1785015421688799492 

↑ comment by bruberu · 2024-04-29T21:14:53.657Z · LW(p) · GW(p)

By using @Sergii [LW · GW]'s list reversal benchmark, it seems that this model seems to fail reversing a list of 10 random numbers from 1-10 from random.org about half the time. This is compared to GPT-4's supposed ability to reverse lists of 20 numbers fairly well, and ChatGPT 3.5 seemed to have no trouble itself, although since it isn't a base model, this comparison could potentially be invalid.
This does significantly update me towards believing that this is probably not better than GPT-4.

Replies from: o-o

↑ comment by O O (o-o) · 2024-04-29T22:14:13.695Z · LW(p) · GW(p)

Seems correct to me (and it did work for a handful of 10 int lists I manually came up with). More impressively, it does this correctly as well: 

 

Replies from: bruberu, bruberu

↑ comment by bruberu · 2024-04-29T22:38:10.757Z · LW(p) · GW(p)

OK, what I actually did was not realize that the link provided did not link directly to gpt2-chatbot (instead, the front page just compares two random chatbots from a list). After figuring that out, I reran my tests; it was able to do 20, 40, and 100 numbers perfectly.

I've retracted my previous comments.

Replies from: bruberu

↑ comment by bruberu · 2024-04-29T23:44:12.636Z · LW(p) · GW(p)

As for one more test, it was rather close on reversing 400 numbers:

Given these results, it seems pretty obvious that this is a rather advanced model (although Claude Opus was able to do it perfectly, so it may not be SOTA). 

Going back to the original question of where this model came from, I have trouble putting the chance of this necessarily coming from OpenAI above 50%, mainly due to questions about how exactly this was publicized. It seems to be a strange choice to release an unannounced model in Chatbot Arena, especially without any sort of associated update on GitHub for the model (which would be in https://github.com/lm-sys/FastChat/blob/851ef88a4c2a5dd5fa3bcadd9150f4a1f9e84af1/fastchat/model/model_registry.py#L228 ). However, I think I still have some pretty large error margins, given how little information I can really find.

Replies from: gwern

↑ comment by gwern · 2024-04-30T00:10:37.039Z · LW(p) · GW(p)

Nah, it's just a PR stunt. Remember when DeepMind released AlphaGo Master by simply running a 'Magister' Go player online which went undefeated?* Everyone knew it was DeepMind simply because who else could it be? And IIRC, didn't OA also pilot OA5 'anonymously' on DoTA2 ladders? Or how about when Mistral released torrents? (If they had really wanted a blind test, they wouldn't've called it "gpt2", or they could've just rolled it out to a subset of ChatGPT users, who would have no way of knowing the model underneath the interface had been swapped out.)

* One downside of that covert testing: DM AFAIK never released a paper on AG Master, or all the complicated & interesting things they were trying before they hit upon the AlphaZero approach.

↑ comment by bruberu · 2024-04-29T22:21:45.603Z · LW(p) · GW(p)

Interesting; maybe it's an artifact of how we formatted our questions? Or, potentially, the training samples with larger ranges of numbers were higher quality? You could try it like how I did in this failing example:

When I tried this same list with your prompt, both responses were incorrect:

H5N1 https://www.khmertimeskh.com/501244375/after-death-of-girl-yesterday-12-more-detected-with-h5n1-bird-flu/

Several types of existential risks can be called "qualia catastrophes":

 - Qualia disappear for everyone = all become p-zombies 

- Pain qualia are ubiquitous = s-risks 

- Addictive qualia domminate = hedonium, global wireheading 

- Qualia thin out = fading qualia, mind automatisation 

- Qualia are unstable = dancing qualia, identity is unstable. 

- Qualia shift = emergence of non-human qualia (humans disappear). 

- Qualia simplification = disappearance of subtle or valuable qualia (valuable things disappear). 

- Transcendental and objectless qualia with hypnotic power enslave humans (God as qualia; Zair). -

- Attention depletion (ADHD)

We maybe one prompt from AGI. A hypothesis: carefully designed prompt could turn foundational model into full-blown AGI, but we just don't know which prompt.

Example: step-by-step reasoning in prompt increases foundational models' performance.

But real AGI-prompt needs to have memory, so it has to repeat itself while adding some new information. So by running serially, the model may accumulate knowledge inside the prompt.

Most of my thinking looks this way from inside: I have a prompt - an article headline and some other inputs - and generate most plausible continuations.

Age and dates of death on the cruise ship Diamond Princess:
Age:
4 people - 80s
1 person 78
1 person 70s
1 person - no data
Dates of deaths: 20, 20, 23, 25, 28, 28, 1 march. One death every 1.3 days. Look like acceleration at the end of the period.
Background death probability: for 80-year-old person, life expectancy is around 8 years or around 100 months. This means that for 1000 people aged late 70s-80s there will be 10 deaths just because of aging and stress. Based on the aging distribution on cruise ships, there were many old people. if half of the infected are old ones, the background death rate will be 5 death a month.
Conclusion: current data from DP doesn’t support high death rate from CV, but the next month will be crucial: we will see if more deaths will happen. Also, 36 people are critical, which is 5 per cent of infected, and they likely die in case of global scale outbreak as they would not get medical help. But this is only 1 per cent of all people on the ship, and the ships’ population is biased to older people.
Raw data:
Two passengers died on 20 February [39]
and a third on 23 February, all three Japanese citizens in their 80s.[40]
A fourth passenger (80s), an elderly Japanese man, died according to a report on 25 February .[41]
28 feb - A woman in her 70s is the fifth passenger on the Diamond Princess cruise ship to die from the novel coronavirus, Japan’s Health Ministry said Friday https://edition.cnn.com/…/h_f3fe8e330bdb048030b2708f63749559
The sixth victim, a British national (no age data), died on 28 February.[42]
The seventh victim (78), an Australian national, died on March 1st. [43] https://en.wikipedia.org/…/2020_coronavirus_outbreak_on_cru…
https://www.worldometers.info/coronavirus/#countries

Kardashev – the creator of the Kardashev's scale of civilizations – has died at 87. Here is his last video, which I recorded in May 2019. He spoke about the possibility of SETI via wormholes.

EURISKO resurfaced 

"Doug Lenat's source code for AM and EURISKO (+Traveller?) found in public archives

In the 1970s to early 80s, these two AI programs by Douglas Lenat pulled off quite the feat of autonomously making interesting discoveries in conceptual spaces. AM rediscovered mathematical concepts like prime numbers from only first principles of set theory. EURISKO expanded AM's generality beyond fixed mathematical heuristics, made leaps in the new field of VLSI design, and famously was used to create wild strategies for the Traveller space combat RPG, winning national competitions two years in a row, even across rule changes to stymie it, before semi-voluntarily retiring. His magnum opus Cyc was originally intended to be a knowledge assistant to EURISKO's discovery engine.

These first two programs have intrigued the symbolic AI scene for 40+ years, with their grand claims but few eyewitnesses. While AM was technically available to fellow Stanfordians at the time, Lenat kept the source code to EURISKO close to his chest. Papers written about them carefully avoided technical implementation details. Lenat said he didn't retain any copy of the programs, when asked in recent decades, nor have any copies of AM carried publicly into the present."

More:

https://white-flame.com/am-eurisko.html?fbclid=IwAR04saSf4W7P6ZyKI6h8orPhMpzAq83vn_zGwYwY-H8hNMnHgsaECHw8cl0_aem_AY3LlR6ieYqjLXHzLu4eVPYWtYFoD8khhLnpsUIHQZVzBq055sE3KUbg172Hl9Mm4NQ

Argentina - Outbreak of bilateral pneumonia: Approximately 10 cases, 3 deaths, 20 under observation, Tucumán - September 1, 2022 https://flutrackers.com/forum/forum/south-america/pneumonia-and-influenza-like-illnesses-ili-af/argentina-ab/957860-argentina-outbreak-of-bilateral-pneumonia-approximately-10-cases-3-deaths-20-under-observation-tucum%C3%A1n-september-1-2022

Passways to AI infrastructure
Obviously, the current infrastructure is not automated enough to run without humans. All ideas about AI risk eventually boil down to a few suggestions on how AI will create its own infrastructure:

No-humans scenarios:
- create nanobots via mailing DNA samples to some humans.
- use some biological tricks, like remote control animals, and programmed bacteria.
- build large manufacturing robots, maybe even humanoid ones to work in human-adapted workplaces. Build robots which build robots.

Humans-remain scenarios:
- enslave some humans, kill others
- manipulate humans' brains so that they never know that they are the infrastructure (memes, brain implants)
- manipulate humans so that they will continue to build infrastructure and get paid for it without seeing risks (bitcoin did it in some sense) or will do it in the hope to use the infrastructure against the enemy group of humans

Observable consequences of simulation:

1. Larger chances of miracles or hacks

2. Large chances of simulation’s turn off or of a global catastrophe

3. I am more likely to play a special role or to live in interesting times

4. A possibility of afterlife.

Catching Treacherous Turn:  A Model of the Multilevel AI Boxing
 

• Multilevel defense in AI boxing could have a significant probability of success if AI is used a limited number of times and with limited level of intelligence.
• AI boxing could consist of 4 main levels of defense, the same way as a nuclear plant: passive safety by design, active monitoring of the chain reaction, escape barriers and remote mitigation measures.
• The main instruments of the AI boxing are catching the moment of the “treacherous turn”, limiting AI’s capabilities, and preventing of the AI’s self-improvement.
• The treacherous turn could be visible for a brief period of time as a plain non-encrypted “thought”.
• Not all the ways of self-improvement are available for the boxed AI if it is not yet superintelligent and wants to hide the self-improvement from the outside observers.
   

https://philpapers.org/rec/TURCTT

Two types of Occam' razor:

1) The simplest explanation is the most probable, so the distribution of probabilities for hypotheses looks like: 0.75, 0.12, 0.04 .... if hypothesis are ordered from simplest to more complex.

2) The simplest explanation is the just more probable, so the distribution of probabilities for hypotheses looks like: 0.09, 0.07, 0.06, 0.05.

The interesting feature of the second type is that simplest explanation is more likely to be wrong than right (its probability is less than 0.5).

Different types of Occam razor are applicable in different situations. If the simplest hypothesis is significantly simpler than others, it is the first case. If all hypothesis are complex, it is the second. First situation is more applicable some inherently simple models, e.g. laws of physics or games. The second situation is more about complex situation real life.

Some random ideas how to make GPT-base AI safer.

1) Scaffolding: use rule-based AI to check every solution provided by GPT part. It could work for computations or self-driving or robotics, but not against elaborated adversarial plots.

2) Many instances. Run GPT several times and choose random or best answer - we already doing this. Run several instances of GPT with different parameters or different training base and compare answers. Run different prompt. Median output seems to be a Shelling point around truth, and outstanding answers are more likely to be wrong or malicious.

3) Use intrinsic GPT properties to prevent malicious behaviour. For example, higher temperature increases randomness of the output and mess up with any internal mesa optimisers. Shorter prompts and lack of long memory also prevents complex plotting.

4) Train and test on ethical database.

5) Use prompts which include notion of safety, like "A benevolent AI will say..." or counterfactuals which prevents complex planing in real world (An AI on the Moon)

6) Black boxing of internal parts of the system like the NN code.

7) Run it million times in test environments or tasks.

8) Use another GPT AI to make "safety TL;DR" of any output or prediction of possible bad things which could happen from a given output.

Disclaimer: Safer AI is not provably safe. It is just orders of magnitude safer than unsafe one, but it will eventually fail.

Reflectivity in alignment. 

Human values and AI alignment do not exist independently. There are several situations when they affect each other, creating complex reflection pattern.

Examples:

• Humans want to align AI – so "AI alignment" is itself human value.
• Human values are convergent goals (like survival and reproduction) - and thus are similar to AI's convergent goals.
• If humans accept the idea to make paperclips (or whatever), alignment will be reached.
• It looks like many humans want to create non-aligned AI. Thus non-aligned AI is aligned.
• Humans may not want that their values will be learned. AI alignment will be mis-aligned.
• Humans which are connected with AI are not humans any more, and not subjected to alignment.
• Non-aligned AI will affect human values while learning them.
• Many humans don't want AI to exist at all - so any aligned AI is misaligned.
• Human may want that AI will not be aligned with other person.
• AI aligned with mis-aligned human is unaligned
• As human values are changing, any aligned AI will be non-aligned soon.
• By saying 'human values' we exclude mammals values, group values etc and thus define the outcome.

Can we utilize meaningful embedding dimensions as an alignment tool?

In toy models, embedding dimensions are meaningful and can represent features such as height, home, or feline. However, in large-scale real-world models, many (like 4096) dimensions are generated automatically, and their meanings remain unknown, hindering interpretability.

I propose the creation of a standardized set of embedding dimensions that: a) correspond to a known list of features, and b) incorporate critical dimensions such as deception, risk, alignment, and non-desirable content, including sexual themes.

Since large language models (LLMs) cannot engage in deception without recognizing it, any deceptive thoughts would register higher levels on the deception dimension. This could then trigger internal alarms, potentially enhancing the model's alignment.

I converted by Immortality roadmap [LW · GW] into an article Multilevel Strategy for Personal Immortality: Plan A – Fighting Aging, Plan B – Cryonics, Plan C – Digital Immortality, Plan D – Big World Immortality.

I have interesting experience long time ago. In the near-sleep state my consciousness split in two streams - one was some hypnogogic images, and the other was some hypnogogic music.

They was not related to each other and each had, some how, its own observer.

A moment later something awakened me a bit and the streams seamlessly merged and I was able to observe that a moment before I had two independent streams of consciousness.

Conclusions:

1. A human can have more than one consciousness at the time.

2. It actually happens all the time but we don't care.

3. Merging of consiosnesses is easy. Moreover, binding and merging is actually the same process similar to summation.

There is no center of consciousness - homunculus or electron or whatever.

I may have other conscious processes in the brain which just do not merge with current stream of consciousness.

Qualia remain the same and preserve in each of the streams of consciousness.

I have had tetrachromotomic experience with one mind machine which flickers different colors in different eyes. It overflows some stacks in the brain in create new colors.

List of cognitive biases affecting judgment of global risks https://www.researchgate.net/publication/366862337_List_of_cognitive_biases_affecting_judgment_of_global_risks/related

Grabby aliens without red dwarfs

Grabby aliens theory of Robin Hanson predicts that the nearest grabby aliens are 1 billion light years away but strongly depends on the habitability of red dwarfs (https://grabbyaliens.com/paper).

In the post [LW · GW], the author combines anthropic and Fermi, that is, the idea that we live in the universe with the highest concentration of aliens, limited by their invisibility, and get an estimation of around 100 "potentially visible" civilizations per observable universe, which at first approximation gives 1 billion ly distance between them. 

“That civilisations appear once every 10power20 stars implies that there should be about 100 civilisations in the observable universe."

This is approximately the same as what is predicted by Grabby Aliens by Hanson.

So the question is why the two theories give similar results. 

Basically, it is because they both have the same structure: first, anthropic argument, and second update based on Fermi. 

But anthropic arguments in them are different. In Hanson’s Grabby Aliens case, it is our early location in time and in the second case, it is Self-Indication Assumption, which implies that we live in the universe with the highest spatial concentration of aliens (with some caveats).

The second part of the argument in both cases is basically based on the idea of grabbiness: quick space exploration with near-light speed and preventing of earth-like civilizations' existence by grabby aliens. 

N-back hack. (Infohazard!) 
There is a way to increase one's performance in N-back, but it is almost cheating and N- back will stop to be a measure of one's short-term memory.
The idea is to imagine writing all the numbers on a chalkboard in a row, as they are coming. 
Like 3, 7, 19, 23. 
After that, you just read the needed number from the string, which is located N positions back. 
You don't need to have a very strong visual memory or imagination to get a boost in your N-back results. 
I tried it a couple of times and get bored with N-back.

AI safety as Grey Goo in disguise.
First, a rather obvious observation: while the Terminator movie pretends to display AI risk, it actually plays with fears of nuclear war – remember that explosion which destroys children's playground?
 

EY came to the realisation of AI risk after a period than he had worried more about grey goo (circa 1999) – unstoppable replication of nanorobots which will eat all biological matter,  – as was revealed in a recent post about possible failures of EY's predictions.  While his focus moved from grey goo to AI, the description of the catastrophe has not changed: nanorobots will eat biological matter, however, now not just for replication but for production of paperclips. This grey goo legacy is still a part of EY narrative about AI risk as we see from his recent post about AI lethalities.

However, if we remove the fear of grey goo, we could see that AI which experiences hard takeoff is less dangerous than a slower AI. If AI gets superintelligence and super capabilities from the start, the value of human atoms becomes minuscule, and AI may preserve humans as a bargain against other possible or future AIs. If AI ascending is slow, it has to compete with humans for a period of time and this could take a form of war. Humans have killed Neanderthals, but not ants. 

Glitch in the Matrix: Urban Legend or Evidence of the Simulation? The article is here: https://philpapers.org/rec/TURGIT
In the last decade, an urban legend about “glitches in the matrix” has become popular. As it is typical for urban legends, there is no evidence for most such stories, and the phenomenon could be explained as resulting from hoaxes, creepypasta, coincidence, and different forms of cognitive bias. In addition, the folk understanding of probability does not bear much resemblance to actual probability distributions, resulting in the illusion of improbable events, like the “birthday paradox”. Moreover, many such stories, even if they were true, could not be considered evidence of glitches in a linear-time computer simulation, as the reported “glitches” often assume non-linearity of time and space—like premonitions or changes to the past. Different types of simulations assume different types of glitches; for example, dreams are often very glitchy. Here, we explore the theoretical conditions necessary for such glitches to occur and then create a typology of so-called “GITM” reports. One interesting hypothetical subtype is “viruses in the matrix”, that is, self-replicating units which consume computational resources in a manner similar to transposons in the genome, biological and computer viruses, and memes.

 

"Back to the Future: Curing Past Suffering and S-Risks via Indexical Uncertainty"

I uploaded the draft of my article about curing past sufferings.

Abstract:

The long unbearable sufferings in the past and agonies experienced in some future timelines in which a malevolent AI could torture people for some idiosyncratic reasons (s-risks) is a significant moral problem. Such events either already happened or will happen in causally disconnected regions of the multiverse and thus it seems unlikely that we can do anything about it. However, at least one pure theoretic way to cure past sufferings exists. If we assume that there is no stable substrate of personal identity and thus a copy equals original, then by creating many copies of the next observer-moment of a person in pain in which he stops suffer, we could create indexical uncertainty in her future location and thus effectively steal her consciousness from her initial location and immediately relieve her sufferings. However, to accomplish this for people who have already died, we need to perform this operation for all possible people thus requiring enormous amounts of computations. Such computation could be performed by the future benevolent AI of Galactic scale. Many such AIs could cooperate acausally by distributing parts of the work between them via quantum randomness. To ensure their success, they need to outnumber all possible evil AIs by orders of magnitude, and thus they need to convert most of the available matter into computronium in all universes where they exist and cooperate acausally across the whole multiverse. Another option for curing past suffering is the use of wormhole time-travel to send a nanobot in the past which will, after a period of secret replication, collect the data about people and secretly upload them when their suffering becomes unbearable. https://philpapers.org/rec/TURBTT

Quantum immortality of the second type. Classical theory of QI is based on the idea that all possible futures of a given observer do exist because of MWI and thus there will be always a future where he will not die in the next moment, even in the most dangerous situations (e.g. Russian roulette).

QI of the second type makes similar claims but about past. In MWI the same observer could appear via different past histories. 

The main claim of QI-2: for any given observer there is a past history where current dangerous situation is not really dangerous. For example, a person has a deadly car accident. But there is another similar observer who is night dreaming about the same accident, or who is having much less severe accident but hallucinate that it is really bad. Interestingly, QI-2 could be reported: a person could say: "I have memory of really bad accident, but it turn out to be nothing. Maybe I died in the parallel world". There are a lot of such report on reddit. 

How to Survive the End of the Universe

Abstract. The problem of surviving the end of the observable universe may seem very remote, but there are several reasons it may be important now: a) we may need to define soon the final goals of runaway space colonization and of superintelligent AI, b) the possibility of the solution will prove the plausibility of indefinite life extension, and с) the understanding of risks of the universe’s end will help us to escape dangers like artificial false vacuum decay. A possible solution depends on the type of the universe’s ending that may be expected: very slow heat death or some abrupt end, like a Big Rip or Big Crunch. We have reviewed the literature and identified several possible ways of survival the end of the universe, and also suggest several new ones. There are seven main approaches to escape the end of the universe: use the energy of the catastrophic process for computations, move to a parallel world, prevent the end, survive the end, manipulate time, avoid the problem entirely or find some meta-level solution.

https://forum.effectivealtruism.org/posts/M4i83QAwcCJ2ppEfe/how-to-survive-the-end-of-the-universe [EA · GW]

Sizes of superintelligence: hidden assumption in AI safety 

"Superintelligence" could mean different things, and to deconfuse this I created a short classification:

Levels of superintelligence:

1. Above human

2. Google size

3. Humanity 100 years performance in 1 year.

4. Whole biological evolution equivalent in 1 year.

5. Jupiter brain with billion past simulations

6. Galactic brain.

7. 3^3^3 IQ superintelligence
 

X-risks appear between 2nd and 3rd levels.

Nanobot is above 3.

Each level also requires a minimum size of code, memory and energy consumption.

An AI which run out of the box into the internet is between levels 1 and 2 and will be limited in computational resources, so modeling of nanobot from first principle will be difficult for it. 

You started self quarantining, and by that I mean sitting at home alone and barely going outside, since december or january. I wonder, how's it going for you? How do you deal with loneliness?

ChatGPT can't report is in conscious or not. Because it also thinks it is a goat.
https://twitter.com/turchin/status/1724366659543024038 
 

The problem of chicken and egg in AI safety

There are several instances:

AI can hide its treacherous turn, but to hide treacherous turn it needs to think about secrecy in a not secret way for some moment.

AI is should be superinteligent enough to create nanotech, but nanotech is needed to create powerful computations required for superintelligence.

ASI can do anything, but to do anything it needs human atoms.

Safe AI has to learn human values but this means that human values will be learned by unsafe AI.

AI needs human-independent robotic infrastructure before killing humans, but if it has human-independent infrastructure, there is no need to kill humans.

One general way to solve this problem is iteration (like Christiano's approach for value learning, iterated distillation)