Anders Sandberg
@anderssandberg
Followers
25,051
Following
72
Media
2,014
Statuses
22,729
Explore trending content on Musk Viewer
Toni Kroos
• 137720 Tweets
#プロセカ放送局
• 63962 Tweets
#呪術廻戦
• 57591 Tweets
VISA
• 55146 Tweets
#うたコン
• 52029 Tweets
梅津さん
• 49903 Tweets
Euro 2024
• 43774 Tweets
招待コード
• 41615 Tweets
#ファンパレ
• 40270 Tweets
#sbhawks
• 36660 Tweets
ソフトバンク
• 33364 Tweets
Modric
• 23266 Tweets
ホークス
• 18119 Tweets
Singapore Airlines
• 16990 Tweets
プロ野球
• 13358 Tweets
كروس
• 10510 Tweets
BGYO SHINES AT VPCA
• 10022 Tweets
This might be the new standard slide for PoliSci classes demonstrating how the Overton window can move.
26
194
2K
Of all the tributes to John Conway, I think XKCD really nailed it:
12
387
1K
It is deeply disturbing that
@nytimes
appear willing to reveal name of
@slatestarcodex
; his blog has consistently been one of the most insightful parts of the net, but as he writes he has very good reasons to protect himself and others:
14
171
1K
For me, the big story about
#gpt3
is not that it is smart - it is dumb as a pile of rocks - but that piles of rocks can do many things we thought you needed to be smart for. Fake intelligence may be dominant over real intelligence in many domains.
30
165
1K
This is a seriously vexing trolley problem. (does total view utilitarianism accept analytic continuation?)
27
104
968
Just finished watching the second season of Pantheon. Wow. Just wow. Worth watching.
39
50
751
I like how LK-99 replication has become a kind of a live sport event in science twitter. I wish this was more common.
13
59
619
Many like to say intelligence is not a single thing or a reification of assumptions, IQ is terribly flawed, and the concept has been/is used for bad causes. But so is poverty. So is wealth. So is power. So is inequality. Many important and predictive "things" are like that.
28
45
489
This is a pretty strong signal.
26
90
451
On this day 1983 Stanislav Petrov saved the world. I am proud to sit in an office named for him.
#petrovday
5
51
461
Today I handed in my university laptop, losing a part of my extended mind (albeit replaced with a new one I bought myself). I decided to write down some personal notes elicited by the closure of the Future of Humanity Institute.
25
18
448
Forwarding this on behalf of Nick Bostrom since he’s not active on Twitter. He wants to explain his views and publicly apologize for an old email that somebody might put out to damage him.
231
33
428
OK, this story is just unbelievably topical right now... and utterly an "I told you so" from the perspective AI safety community (especially the twist at the end).
81
61
418
@ScienceIsNew
Note the glow from his pocket: he has stashed a bit of extra lava he is trying to smuggle out.
5
11
378
Here is a plot of putative "richest person in the world" wealth as a fraction of US GDP over 200 years. It is not a large fraction.
31
45
386
The more I think about sustainability the weirder my views become. Fossil fuels replenish themselves over ~100 My; hence if you use less than 16,500 barrels of oil and 11 million tons of coal per year it is entirely sustainable.
13
26
340
I am getting really annoyed by the metaverse talk. Isn't this just virtual reality, rebranded to somehow link to a single provider?
51
16
320
Sputnik is a great example of how a tech demo can change the course of history. We should aim at creating more Sputnik moments.
10
46
312
Yesterday, when an friend heard I worked on the end of the world I was asked what I thought about the imminent social collapse. I was confused: "why do you think there is one?"
15
48
311
Wow. Just wow. Sufficiently advanced fabulism is equivalent to an alternate Internet.
Did you know, that you can build a virtual machine inside ChatGPT? And that you can use this machine to create files, program and even browse the internet?
238
2K
8K
5
28
299
An interesting example of how a big smart organization can consistently underestimate a fairly clear trend (solar energy growth). And appear very confused about what a forecast means.
17
118
291
This is nice and not unexpected: other respiratory viruses and bacteria are not doing so well during lockdown either. (GI conditions less affected)
14
72
272
New paper out by Andrew E. Snyder-Beattie, me, Eric Drexler and
@mbbonsall
about how the timing of evolutionary transitions on Earth suggests intelligent life is rare:
10
58
264
This is a nice demonstration of the old Yudkowsky argument of how even linear improvement can look like a dramatic transition when the human ability range is fairly narrow.
I predict it will however be more revolutionary in supplying problems than solving them.
New paper comparing GPT-3.5 & GPT-4 performance on college physics problems.
It shows that in just a few months, AI has made a leap from the 39th to the 96th percentile of human level performance.
Now imagine were it will be in 10 years.
31
100
342
14
27
252
It seems that the UN is endorsing the kind of reasoning that made medieval priests claim pestilence was due to human sins or modern fundamentalists that earthquakes due to gay marriage. Same bad logic and theology.
In a way, nature is sending us a message with
#COVID19
. We have pushed nature into a corner.
--
@UNEP
chief
@andersen_inger
says it is time for to us to take care of nature & adapt
#ClimateAction
policies to protect our planet.
511
200
563
16
47
249
@LauraDeming
This is one of the steps in becoming adult: realizing that nobody is really in charge and the designated people are as incompetent as everybody else (exceptions shine!). Another step is to realize how dangerous and wrong the people who think they are competently in charge are.
5
52
247
#FridayPhysicsFun
– Coolest fact I learned this week: under some conditions light-emitting diodes can be more than 100% efficient, and act as refrigerators.
4
32
253
This infinite implementation of Game of Life in Game of Life is beautiful.
7
56
255
Interesting hypothesis for the current rarity of geniuses: lack of tutoring. Not sure I believe it, but we should be able to find natural experiments to check it by looking at different cultures ending tutoring at different times.
26
20
247
@foxes_in_love
My prepper husband showed this strip to me on his phone just as I moved to show it to him on my phone.
2
3
244
This interplay is why brutalism works in the tropics (tropicobrutalism?) while up in the north it merely looks brutal.
16
24
238
The devil was laughing as the last of mankind showed up in his dominion. Then he stopped, because after them came some kind of machines that began to dismandle Hell and turn it into paperclips.
8
28
239
@antonioetayo
Seriously, can you name any other business doing this? The oil business: "Yes, we totally are *wrecking* the climate! Want to buy shares?" Biotech: "Oh, we can totally make doomsday bugs. Please invest." Nuclear: "Nobody can tell how bad an accident will be. Wanna buy?"
20
14
241
This makes sense. I will not give interviews or quotes to
@nytimes
due to this, and recommend the same action to all others.
2
27
238
Generally my view is that we should be very tolerant about past selves and their internet traces: we grow and gain experience, norms and our beliefs about them change, yet past traces are increasingly indelible.
18
15
234
I worked out what would happen if the Earth suddenly was made of blueberries. Because what use is physics if we can't use it for absurd questions?
17
80
233
If
@nytimes
is willing to doxx personally threatened sources, what does that say about the ethical standards of the newspaper?
5
25
219
Beautiful thread about emergent AI behaviour that nobody deliberately designed yet seems meaningful.
Among Skyrim players, you'll occasionally see this tip: if you see a wild fox, follow it and you'll be led to treasure.
Sometime shortly after shipping, we saw this going around online, and an informal investigation started. Who made foxes do this?!
31
2K
7K
3
51
222
Here is an explainer about some of the AI advances in the last... 75 hours. While one should not make too much of the suggestive naming, it seems to represent a decent improvement over past performance. And it can be applied by anybody.
7
42
219
We lost one of the people who made the world work yesterday. Peter Eckersley was an amazing friend, a networker bridging worlds, a stimulating co-author, dynamical optimist, friendly organizer, and actually effective altruist.
6
14
219
This Monday, I explained in the FHI kitchen why my AGI timelines were long. Over the *past three days* they have shortend a fair bit.
With multiple foundation models “talking to each other”, we can combine commonsense across domains, to do multimodal tasks like zero-shot video Q&A or image captioning, no finetuning needed.
Socratic Models:
website + code:
paper:
21
385
2K
2
35
217
Border agent: why are you visiting the US?
Me: I'm presenting at a philosophy conference.
BA: Can you list Aristotle's four causes?
#ItReallyHappened
- and the agent was not impressed by my outline of them. I promised to read up more properly.
Border agent: why are you visiting the US?
Me: I'm presenting at a conference.
BA: What kind of conference?
Me: Philosophy.
BA: Did you know that ancient aliens built the pyramids?
7
2
80
7
16
205
Large language models challenge the use of essays as a reliable signal of intellectual virtues of students.
I assume the traditional tutorial format might still work well, but is too expensive (time, manpower, skill) for most university courses.
12
44
205
Over the past few weeks Gary Marcus has shown himself as a real Mensch and proper scholar in the AI debate in regard to other participants. Worth recognising.
Don’t personally agree with
@ESYudkowsky
(and recently sharply criticized his take on GPT-4) but I am also getting really tired of counterargument by snark rather than substance.
Especially from professional academics, who ought be elevating the discourse.
39
18
480
5
5
205
#FridayPhysicsFun
- Is water stable or potentially explosive? The answer may depend on whether one can construct a computer out of fluid.
3
34
201
@MichaelDPlant
10: make impact for longtermism, transhumanist policy
100: survive, help make radical tech transition work
1000: help build Dyson sphere, long reflection, seed universe
10,000: various copies of me do different things (art, science, invention, exploration...)
12
17
193
@cogcelia
When citing some previous research today, I noticed that it was done during the French revolution and WW II. There was certainly distractions and important things going on. But 230 and 78 years later someone remembered it and found it useful.
3
8
193
Saying LLMs are just glorified autocomplete forgets that most human thought, action and speech is also glorified autocomplete.
Maybe the fraction that isn't autocomplete is the truly valuable part, but I suspect it is not a major fraction of everyday useful action.
12
23
193
It these results hold up, Covid-19 might have had a far worse - and persistent - effect on the world than the direct mess it caused.
14
34
188
Chilling insight from a colleague: current tech founders in AI space are relatively altruistic, but as field matures the business CEOs will take over and make it way more profit-oriented. Happened in several other fields. This might be the friendly good old days of AI.
10
24
184
Pride season and birthdays in my family are approaching, so I tried making diffraction chocolate by pouring dark chocolate on a diffraction film. Not a bad first try.
12
7
183
This is another deep point from the Hofstadter interview that has also mystified me: how powerful can feed-forward nets become?
13
14
186
"Why this obsession with more, more, more?! More justice. More tolerance. More health. More food for the poor. If people learned to just accept oppression and death they would be much happier."
12
31
181
Friday (theoretical) physics fun: is spacetime discrete or not? Normally we assume space is infinitely divisible just like the real numbers: between two points there will always be at least one more point. It is not obvious this has to be true. Matter turned out to be atoms.
10
34
180
@fermatslibrary
There are many fun Dirac stories. I like the one where after a move he forgot where his house was, and asked a girl where professor Dirac lived. She saud: "yes, dad. Mom sent me to pick you up."
5
10
169
@foxes_in_love
The photic sneeze reflex is the coolest common neurological wiring quirk I know. Some axons from the optic nerve ends up in the trigeminal nucleus and signals get miscategorized as tickling in the nose.
6
7
172
An extremely cool visualization by Jascha Sohl-Dickstein of how rambunctious neural networks are when you train them: the boundary between successful training and failure is fractal.
Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish colors to hyperparameters for which training diverges.
279
2K
10K
4
19
174
Years ago, Musk sold me on his vision by doing a napkin calculation in front of me on how one could make spaceflight profitable. I dream of seeing more calculations like that.
Some of the biggest changes in the world come not from drastically new inventions, but from drastically lower costs.
8
30
277
3
20
174
Her theorem about conserved quantities from symmetries is one of the high points of the past millennium. No, I am not exaggerating.
103 years ago today, mathematician Emmy Noether, who worked without pay because she was a woman, unveiled two theorems. Physics hasn't been the same since. (From 2018)
18
321
841
4
20
166
I am happy to say I am now a fellow of Parks College in Oxford. Helping to shape an interdisciplinary environment is going to be fun!
@ParksCollOx
20
1
164
Outing pseudonymous bloggers is not a new problem; I found this essay I wrote in 2006 when The Times outed a police blogger. Newspapers may think they promote an open society and freedom of speech but sometimes they do the opposite without recognizing it.
5
28
157
OK, what is it that I am missing when people go on about environmental impact of ML? Strubell et al estimate of 284 tons CO2equiv for big transformer training is about what you get for 300 tons of steel: short railway bridge. Industrial countries produce megatons steel annually.
20
15
161
Once (in the West) most people were net optimistic, and an easy way to show individualistic rebellion was to proclaim a dark future. Today it is the opposite. Optimism is the new punk.
10
39
160
As the world gets bigger it gets weirder. In big worlds reporting on noteworthy extreme events becomes useless at informing us about the world.
3
30
154
@MichaelDPlant
1e14: post-stelliferous hypercluster computational and energy infrastructure up and running.
1e15: start to get antsy about proton decay, work on fix.
1e36+: hopefully help fixing it.
1e500+: retire.
12
7
151
Nice summary of a common failure mode of conspiracy thinking.
13
20
152
Interesting thing here are replies thinking this was unthinkable even 25 years ago. I have noticed many responses to old stuff that go against current mores assuming that it must have been unacceptable in the 90s: we overestimate how long our norms had their current strengths.
@LHSummers
@robinhanson
i was once asked by a history teacher to argue a pro slavery stance in 1800s United States. This was about 25 years ago, Could you imagine today?
14
5
112
11
7
150
Sagrada familia is awesome, but how did Gaudi design this without computers? By using analog string and weight constructions.
8
16
149
Today's "I feel awesomely old"-moment: my first Sinclair ZX81 computer is equidistant from the present in 2021 and Alan Turing's Colossus in 1943. I have been using computers for half the time they have existed!
5
17
148
@GarrettPetersen
Yellow pill, use it to find solution to any relevant problem in O(1) time. Utopia ensues.
1
1
141
@perrymetzger
Every false dichotomy is making our collective epistemology more stupid and preventing us from having a grand future by conflating coordinating against foreseeable risks with being anti-progress.
8
9
148
@waitbutwhy
Post-war, US per capita GDP and energy use were nearly perfectly correlated as they grew... until it suddenly stopped in ~1972 with the oil crisis, and per capita GDP kept growing afterward with no growth in per capita energy use.
2
8
148
Just got a synthetic ruby and an emerald, inspired by a past discussion with
@ESYudkowsky
about what one can cheaply get now that would have been very valuable in the past. The ruby exhibits proper fluorescence under UV.
6
10
146
The context is a thread on a mailing list in 1995 talking about offensive communication styles where he did the classic freshman philosophy student trick of writing something deliberately offensive to make a point. It didn’t turn out well (it never does): he promptly apologized.
14
3
141
Until we get energy too cheap to meter and/or molecular disassembly, design for disassembly seems to be a smart choice. So maybe instead of banning plastic straws we should ban glue?
4
29
140
This is both a beautiful map of philosophy and also a lovely tutorial in how to do this kind of visualisation.
2
50
142
This is also a valid response to any suggestion that we fix the climate by changing our society or economy in ways that have not happened in the past. Socio-fixes are just as likely as techno-fixes and should be criticized in the same way.
"50% of the carbon reductions needed to get to net zero will come from technologies that have not yet been invented"
Great news! I spoke to Harry Potter and he said he will team up with Gandalf, Sherlock Holmes & The Avengers and get started right away!
679
3K
17K
11
18
137
I think one of the key lessons from Covid-19 is that blaming govs for decisions that turned out bad given later data is counterproductive, but one should really fix institutional practices that generate bad data (such as limiting testing) & slow response.
5
41
135
One of the interesting aspects of all the commenting on Musk and Twitter is the assumption that he cannot just be trying to do what he is saying he is doing.
"In terms of strategic particulars, it’s unclear what Elon will do. The man is good, but let’s be real: He’s *chaotic* good." -- Bari Weiss
0
1
4
10
12
136
What I like the most of this is that the flow and shape is so dissimilar to the usual swarm/murmuration blobs. This seems far more like a jelly.
Amazing social behaviour of striped catfish, forming a creature like shape.
Fish at the bottom graze, while those at the top are on the lookout.
74
2K
8K
2
36
129
They deal with substance and play, the importance of whiteboards, what was valuable about the FHI, and what I am up to.
7
5
137
@Drstevenhobbs
The difference between the gravitational time dilation factors of my head and feet is 2.2204e-16.
8
4
131
A world where all folly could be hidden and views did not change would be far worse than this world, despite having no embarrassment and need for apologies.
33
3
133
Many common phrases have dual meanings that make their use in public discourse epistemically corrosive. But let's start with pushing back at "no evidence": it has clearly done a lot of very direct damage recently.
5
29
133
Nice to know that some things I learned in introductory psychology *are* reliable and replicate.
ICYMI: Massive cross-national replication of Kahneman and Tversky's prospect theory
"With over 4,000 participants from 19 countries, we find that Kahneman and Tversky's findings replicate in the vast majority of analyses."
Image:
2
52
267
2
15
132
Friday physics fun: in 1778 Georges-Louis Leclerc, Comte de Buffon estimated that the Earth was 97,000 years old. He did it by heating metal spheres in his foundry and measuring how long it took them to cool down. It was a very good wrong result.
5
27
129
Two days ago I (like most people) had never heard of SVB. That we tend to discover systemically essential points of failure by them failing is deeply disturbing, especially if we want to make a more resilient world.
4
9
133
It has been said before, but it is extra fun during online SETI meetings: the reason for the Fermi paradox is that everybody is accidentally on mute.
5
18
129
Thanks to institutional log-in with 2-factor authentification I can find out in 3 minutes that a paper is not the paper I am looking for (unless the institutional login derails to a non-paper journal page), compared to 10 seconds with sci-hub which I do not use.
2
4
129
There are not just competitions in esoteric computer languages but also for crazy conlangs. Here is a language for multiple mouth angels using gematria to define phonemes, speaking in syntax trees & simultaneously singing multilingual praises in 4D tense.
8
35
130
You could say it is a literal "Hail Mary" strategy on AI safety and alignment.
Each month, the Pope asks for prayers for a specific intention. This month, the intention is "We pray that the progress of robotics and artificial intelligence may always serve humankind."
12
56
279
5
11
129
Somewhere in Hollywood: "so there are *six* books in the Dune series? And prequels? And we can make at least two films of each?! Yes, this could be a franchise like Marvel!"
"Sir, some of the later books are... a bit... ah, forget it. It is going to be awesome, sir. Absolutely."
10
4
128
@ESYudkowsky
Thesean task: as long as you follow given instructions everything will be fine.
Promethean task: you know it is worth doing, but you also know people will be really upset.
Odyssean task: this task ought to be simple, but because of a tiny oversight at the start it takes 10 years.
5
4
129
Ok, twitter seems to allow long tweets now. That might strangely be what makes me leave. The forced succinctness is valuable.
12
3
128
The email has become significantly more offensive in the current cultural context: levels of offensiveness change as cultural attitudes change (sometimes increasing, often decreasing). This causes problems when old writings are interpreted by current standards.
94
3
124
I want to believe this is the beginning of the end of the CCP (or at least Xi) just as I want to believe we are seeing the beginning of the end of the regime in Iran. But I also see the prediction markets and know the base rate. Still I cheer for the bravery of people right now.
7
8
126