This is one of the more hopeful processes happening on Earth right now - because it may give rise to a culture of people with something like security mindset, who try to break things, instead of imagining how wonderfully they'll work. lesswrong.com/posts/aPeJE8bS
Follow
Daniel Martin
@DanGMartin1
Here to chat about physics, compute, design, learning, and to have a good time. Software
Daniel Martin’s Tweets
Re chatbots: Hollywood already figured out how a free society can balance the harms from new media cheaply blasting out antisocial filth and smut vs the necessity of free speech. You just label the offensive stuff and let people see what they want. 1/
2
3
25
Show this thread
Most posts by are insightful, but this one is unusually insightful, even for him.
overcomingbias.com/2023/02/why-is
4
10
64
My only exceptions are sentient computers and cute UX in Japan
Quote Tweet
1
Driving up local gradients can be more poison than medicine. And not because they’re local maxima.
Quote Tweet
Interesting video. There is a thing that happens where a designer sees something happening they don't like, and says, "I will prevent that by adding X1". Then other stuff happens, "I will require the player to X2", "I will make this less viable because of hindrance X3," twitter.com/VeriitasGames/…
Show this thread
Quotes from four high level managers, re why they don't use prediction markets:
overcomingbias.com/2023/01/predic
10
14
52
The pressure to consider some lyrics ‘the original ones’ has good rationale, but sticking to it too tightly holds us back from worthwhile explorations of the lyrical space of that song.
The same goes for style transfer.
Show this thread
Today, once a pop artist records a track, we still remix, but how often does the original author (or others) record with new lyrics?
1
Show this thread
Earlier conceptions of music, along with newer traditions like blues and rap, benefited from being loose and pastiche.
1
Show this thread
I expect generative models to improve our relationship to music by moving us away from the discrete model of a song.
1
3
Show this thread
Is this different to today though 🤣
Quote Tweet
1
Ah, a name for my current view on qualia.
Quote Tweet
Some things illusionists do not claim to be illusory:
-the material universe
-cilantro
-the Kardashians
-introspectable mental states and processes
-pain
-consciousness
...
One thing illusionists do claim to be illusory:
-qualia
I agree that this is very interesting, but it’s not due to Moravec. It’s because of the way it fetches answers.
Quote Tweet
"AI can accurately describe the human condition, can't count to 5" was not on my bingo card. Looks like Moravec strikes again twitter.com/reconfigurthin…
1
1
In an effort to prepare children for the real world, we’ve built a huge wall between the real world and school to permanently separate the two.
Quote Tweet
Thx to Scott Aaronson, GPT outputs will soon be watermarked w/ a random seed, making it much harder to submit your GPT-written homework without getting caught
He doesn’t give too many details about how it works, but I suspect its possible to bypass using a clever decoding strat
Show this thread
2:19
139.6K views
2
4
26
Show this thread
One route into the Deutschian worldview is understanding how conventional epistemology overextends itself.
See also: “Doctors advise there is no evidence for […]”.
Quote Tweet
Here are some numbers: 1,1,2,3,5,8.
How can we predict the next number?
First of all, let's try to guess what kind of pattern could generate the numbers above.
1+1=2 1+2=3 etc. It is the Fibonacci sequence. So we predict that the next number will be 5+8 = 13.
Show this thread
3
10
10
66
1
A) We must halt the spread of misinformation
B) The solution to bad speech is more speech
- A4.2%
- B41.7%
- Both/Neither/Other45.8%
- See results8.3%
24 votesFinal results
1
2
I'd like to find a good crash course in neurochemistry.
Something relatively simple. "Here's a list of all the major neuromodulators (dopamine, serotonin, epinephrine, acetylcholine, etc.) and the list of common effects of each on different psychological dimensions."
1
5
23
Show this thread
This contains a small island’s worth of knowledge on the practice of software (and on how to think!) grugbrain.dev
8
"The consequences of my investigations was that [...] even if we wished to do it, we basically couldn't."
That's good news! Those papers were spooky.
1
2
Show this thread
"All of the negative scenarios just don't seem... I [became] the world expert on 'bubbles of nothing', which is an extremely catastrophic possibility of vacuum decay, in which a bubble appears [in which] spacetime ceases to exist, and that expands outwards at the speed of light."
1
1
Show this thread
"On this question I think physics basically only has good news, with the sole exception of the possibility of heat death, which itself could possibly be transcended [...]"
1
2
Show this thread
The latest twitter chaos is fascinating. Some notes from the philosophy armchair. Free speech/Exit, Voice, Loyalty/open institutions/ephemera.
2
6
47
Show this thread
Yup. This is a clue that something is up!
Quote Tweet
The fact that the age of the Earth and the age of the universe are even remotely similar numbers is honestly pretty weird.
Show this thread
1
I got the sense the whole world wanted to enjoy the same thing, so I watched football for once, and loved it. The sun shines on Messi! ☀️⚽️
Bookstores are even more craven than TikTok, in a sense: books compete there on the processing timescale of ~1s.
not sure why the OP got suspended, I’m out of the loop
Show this thread
However I view Roko's thread as misunderstanding the nature of LLM cognition, so it won't lead to "Em cities".
1
1
Show this thread
This is a good insight: LLM capabilities, like ems, seem to exist in a family tree of 'in silico' artifacts.
Spend $1bn to train one model, save a copy. Train it on different data, save another copy. If you're not happy, restore from backup and retrain.
Quote Tweet
I think #GPT-style LLMs are going to be functionally similar to @robinhanson's ems.
There will never be any scanning or slicing involved though. Backprop just extracts the information and makes an isomorphic network in silico.
But still, that means Robin was right.
Show this thread
1
4
Show this thread
I do feel some cringe in this style of tweet, in that it’s pointed towards the numinous. As we learn from history, it’s better to be open to the profound rather than pointed towards it.
Show this thread
In 1999, Bowie described the internet as an ‘alien life form’. Arguably, this was first realised in The Great Weirding of 2016-‘19.
The past doesn’t furnish a guide to the future. Towards which weirdness is our society now pointed?
2
2
Show this thread
A view from the bottom up. The opposite of Seeing Like A State.
Quote Tweet
Last year, a group of geeks convened on Discord in response to a Twitter thread about California’s woes with the vaccine rollout. They very quickly found themselves running a critical piece of national public health infrastructure.
Here is our story:
worksinprogress.co/issue/the-stor
Show this thread
1
1
9
People are easily fooled by verbalism. You can avoid fooling yourself just by naming things more precisely:
“The AI” —> Language model
“web3” —> Decentralized immutable ledger
2
3
25
The uncritical response of to ChatGPT should serve as a wake up call to doomerism. (Tagging his account because I respect him.)
AGI may destroy civilisation, as humans themselves might. But there is a force - reason - that pushes in the opposite direction.
It would be neat if Twitter had a trending topic minimap so I could move away from ChatGPT for a bit and explore some other spaces.
14
12
337