A smart person once gave me their hot take, it's better to double down on hedonistic pathways you are sure of and exploit that.
This makes sense, since time is limited and exploration is a high variance strategy with too many false positives.
Jhanas are still tempting though.
Aditya
@adityaarpitha
Alignment researcher. Passionate about AI safety. We should know how to safely align an AGI before building one.
PhD student
Aditya’s Tweets
Imagine you were convinced of a cancer diagnosis. You have 3 years at best to live.
The common reaction is to re-evaluate priorities and die with no regrets. Try out new experiences you'd pushed for later, explore new directions?
Tick the items on your bucket list off?
4
1
Show this thread
The artist first needs to experience That. Then he will naturally be consumed by a need to communicate it to others. Art is his attempt to express his interpretation.
The implicit has enough degrees of freedom to do this expression justice.
There is so much room at the top.
Show this thread
Art is shining a torch onto the platonic ideal and manifesting a coarse approximation of That using colours, sound or flavor.
We are all natural artists. We stunt ourselves when we limit our output into inadequate channels, trying to be precise, rigid and explicit with say words
2
2
Show this thread
I wonder if analysis has been done on who flies around the world. I think it's a small group of people who take most of the flights.
The world is truly a global village when elites tour the world like most people take buses on the weekend.
3
It started with sampling a wide variety of options in Yodobashi, Tokyo and getting my Final B3.
Then in London a few weeks ago I got my Shure 846 SE gen 2.
It's psychedelic, the clarity of the music makes my heart ache, I want to cry and dance. It's worth the obscene price tag
Show this thread
This statement from David Spivak's latest talk (youtube.com/watch?v=894CGg) deeply resonates with me.
I too have never understood what the 'artificial' in intelligence really stands for: nature made us, and via us, it made neural networks and GPT-3.
7
13
67
Show this thread
I think it is wild that we can have very smart people look at the same evidence and reach probabilities from 0 all the way to nearly 100 for serious outcomes like extinction. The lack of consensus points to a need for deconfusion, getting on the same philosophical framework.
Quote Tweet
Had an insightful conversation with @geoffreyhinton about AI and catastrophic risks. Two thoughts we want to share:
(i) It's important that AI scientists reach consensus on risks-similar to climate scientists, who have rough consensus on climate change-to shape good policy.…
6:10
227K views
1
4
How high? LSD and psilocybin bind to TrkB receptors a THOUSAND times greater than standard antidepressants like Prozac! Even 1,000x higher than ketamine!!! (13/19)
8
7
92
Show this thread
Anyone know what fun stuff to do in London would be? I tried bouldering, and visited museums. Now looking for comedy clubs, theatres, music festivals, etc
5
4
It is a kind of arrogance, a narrow mindedness that the first aliens which notice our planet would first notice us humans...
In fact to them, through certain framings, it would be blindingly obvious that our planet is a hellscape built on top of suffering of many simpler beings.
Quote Tweet
Short (maybe true?) Story
qualiacomputing.com/2020/02/23/qua
no one doubts that hallucinations are integral to the functioning of *image* models.
text is not fundamentally different. we've just done better at appreciating image models for creating things that don't exist yet, instead of trying to turn them into glorified databases.
6
13
157
Show this thread
Again it's his optimistic hope that there are more breakthroughs needed. He admits to uncertainty. Then he should be playing for all the possible worlds he could be in, managing risk.
I like that people are speaking out their reasoning, makes it clearer alignment is neglected.
Show this thread
It comes down to agency or autonomy, as Mark calls it. He is betting on it not being present in models even as we optimize for "intelligence".
Interesting to see that the goalpost has been moved to ASI instead of AGI.
1
1
Show this thread
Granted this is to be expected as despite having the word "omniscient" in all our vocabularies, we never had any *active* entity--living or artificial--that actually had an approximate version of that.. 6/
1
1
3
Show this thread
My tools are those of a mathematician, but my aesthetic is that of an early natural scientist.
1
5
160
Show this thread
Our monkey brains work on repetition. Gratitude journalling works, positive affirmations work. This is also why self deprecating humor can really hurt your self esteem in large doses irrespective of how obviously untrue the joke is.
Took too long before I stopped being a Jester.
1
1
There is the life you lived thinking there are only a few people compatible with you romantically. Then there is life after you realize how vast the world is and you couldn't possibly connect with all the marvelous people out there with the time you have.
1
3
AI Ethics - "We seem to be capable of embuing AI systems with human values using techniques like RLHF. So the more pressing question is which values we put in them and having watchdogs to make sure corps do the right thing"
AI Alignment - "These methods won't scale to handle AGI"
2
This is why I like twitter's character limit. Makes you really careful with which words you use, makes you concise and pack information into the tweets.
Quote Tweet
constraints make creative work fun
Show this thread
1
Here are three policy proposals from the AI ethics community that we believe would improve safety! 🧵
safe.ai/post/three-pol
11
81
191
Show this thread
There is a time every few months when I listen to this cover by our little gremlin.
youtu.be/cP3eChqUphA
I know about the shark, but considering Senzawa is actually gone. The song hits so differently. It captures the feeling of missing.
The crazy part is this actually kinda worked! Someone should make a movie about this...
Quote Tweet
Realizing that the problem of AI safety was too big for him to solve alone or with a small team, he set out to attract the attention of the next generation of gifted mathematical minds to work on the problem at a larger scale by … writing Harry Potter fan fiction.
Show this thread
1
It's not people who lack imagination that worry me. It's people who are only able to imagine utopia and have perfected the art of avoiding uncomfortable possibilities.The security mindset is about cultivating that skepticism, laying assumptions bare and having epistemic humility
“The trick is to use the drugs once to get there, and maybe spend the next ten years trying to get back there without the drug.”
— Maynard James Keenan
1
2
3
Reminded me of “only way to stop a bad guy with a gun is a good guy with a gun” but even that is a better take because guns are not as powerful "aligned AGI".
It also warmed my heart to see how emotional Connor got in defense of humanity. You can see how deeply he cares.
Quote Tweet
Please watch this unedited clip from today's AI safety debate between @NPCollapse and @JosephJacks_:
Show this thread
3:48
40.8K views
2
3
I guess people want a neat story they can point to in their head like Terminator or Skynet. A concrete thing you can point to as the baddie. When the danger is more amorphous, convergent, and abstract, it is so much harder to communicate it.
Hats off to David for such a good job
Quote Tweet
A clip from my first live TV appearance!
With @AyoSokale on @GMB discussing extinction risk from AI. twitter.com/GMB/status/166…
1
“if someone does crack the code and build a superintelligence … I’d like to make sure that we treat this at least as seriously as we treat, say, nuclear material” - Sam Altman
huh that at least is doing a lot of work there.
1
1
When you go seeking a solution desperately. It is wise to keep your mind open to the possibility that the question might be worth dissolving.
1
How many people who confidently state AI cannot do X would have also said the same about the systems that we have today?
2
1
It's heartening to me when I see such great work. Humanity is coming together there is a response.
It made me tear up. I kinda want us to survive more now cause I wonder what "movie" they will make about The Alignment Problem.
1
I love this guy 😂😂 I am entertained.
Quote Tweet
Replying to @Soph8B and @julio_bree
99% of AI researchers don’t think AGI is going to kill everyone but they are the boring 99%
1
I'm not sure I can finish watching superbad. I am now re-watching it after a long while. Some parts are so cringe (in a good way) that I physically can't handle it. The second hand embarrassment I feel has me pausing the movie, looking away and also squirming in the seat. 🤣
Replying to
Not to be annoying and 1hr is great but at 2hrs / day a lot of people have found that samadhi just stays around all the time, it’s the only big phase shift I know of, afaict it’s like 0-1hr big, 2hrs big, no big diff between 2hrs and 8hrs
13
4
187


















