Human intelligence is very specialized, so equating it with general intelligence isn't right
https://x.com/ylecun/status/1846034609521246234?t=NwnwDFNDrd1BOfUdtGE41A&s=19
The ultimate fight in cognitive science: Is the human intelligence, which is shaped by evolution, a collection of special-purpose programs, or a general-purpose blank slate that can be filled with any computations, or combination of both, something in the middle, or something else?
Scaled deep attentional multiagent self play reinforcement learning grounded in physical world simulation with scaled test time compute is all you need for AGI
https://x.com/sgershuni/status/1845384202256109896?t=jPuCs0hf7LUcRZvBQjhtig&s=19
definitions of intelligence, AI (maybe machine intelligence), AGI, ASI, and it's relation to human intelligence - are more specialized (like visual cortex) or mor general or something in the middle or something else?
If everything gets automated, do you support UBI?
Fascinating how you need to find signals in constant daily flow of petabytes of data in CERN.
There seems to be a cultural memetic phase shift where people are becoming more proAI
Accelerate AI for good
White mirror
I was at a meeting of people who are doing AI in healthcare yesterday and lots of them were complaining how pushing some innovation in Czech (European) healthcare system can sometimes even take 10 years because of the infinite layers of regulations and certificates and so on even if the methods are already proven to work by scientific studies
The universe is selforganized emergent harmony at the edge of chaos across scales 💞
We have easily accessible sum of most of human knowledge in a magical box in a pocket and most of the population is using it to rot their brain by watching TikTok dances
Superposition of all possible mathematical frameworks
LangGraph is graphbased neurosymbolic AI over LLMs
[LangGraph](https://www.langchain.com/langgraph)
Autonomous multiagent AI systems will eat everything
"
I wish we had a proper fully empirically tested theory of what makes a physical system sentient.
It has so many ethical implications! I don't want us to create physical systems that feel raw biggest pain imaginable.
Even bigger grey area is synthetic biology, brain organoid computing, hybrids etc. 😭
Or when I look at how we treat other animals...
To what pattern does experience correspond to? Does experience correspond to some algorithm, some function, some physical phenomenon, something else, all, none, or are those different lenses on different parts of the same conscious system? Does substrate independence hold? All the different answers that people propose in neuroscience, philosophy of mind etc. just don't feel satisfying enough. Even if they feel reasonable to some people, you still have to test it, which feels impossible.
And is empirical testing of this is even possible? As science currently can see just neural correlates of various states of our experience, and trusting the subjects not lying about their experience.
Do even the assumptions under that, nonpanpsychist physicalism, even hold?
Maybe the solipsists and illusionists are up to something.
Idk, I feel like giving up in this field for the nth time, I often feel like it's impossible under the standard scientific methodology.
"
I wanna see more implementationist models trying to reverse engineer curiousity, care and creative spirit, other than just some brain region activity correlations that can be found for any phenomena :D
Shapeshift your causal phenomenology between Maars levels of analysis with bidirectional causality between eachother.
"
všichni každej den používáme umělou inteligenci, třeba když použijeme google search
pokud teda pod umělou inteligencí nemyslí jenom jazykový modely finetunutý na chatbota
což mě tak nějak štve denně, že to je jediný co dneska lidi po "umělá inteligence" znají
když je to 0.1% celýho oboru AI
nebo třeba různý typy AI systémů ve zdravotnictví, kde teď vidím nejlepší pozitivní aplikace
Tho when it comes to the current AI wars, I think most people don't hate AI itself as a technology, but the exploitative inequitable system using some types of the dual-use technology in certain contexts
Je zajímavý že na AI je teď největší fixace v kulturní válce, a na ostatní technologie co se v týhle mašinerii používají podobně pro dobro i zlo ne tolik. Asi to bude taky tím jak moc jsou některý části AI nový a jaký mají hlasitý impact.
Nebo se lidi dost bojí kvůli tomu že je to může nahradit např v práci... proto jsem pro např UBI
a pro decentralizaci a demokratizaci technologií, místo další a další koncentraci technologické, ekonomické, politické, kulturní apod. moci
Forces that are bad and forces that are good for all the people are both trying, and we need more good forces
It makes me sad daily that such technology that is also already doing so much good and has even more extreme potential to do enormous good is now being hated because some people use it for evil
And news media tends to amplify the negative more than the positive which then most people see
And it's not just the practical usecases (healthcare usecases ftw I think), like, trying to build intelligence in machines from research perspective feels like the most interesting thing possible to the supercurious hypernerd part of me
My intuitive mind often cannot comprehend why people don't want to understand the equation of the universe and the equation of intelligence, but I'm a nerd hyperfixating on these concrete topics, so that's why. My rational mind can find all kinds of explanations for why is it like that, by for example how evolution and environmental factors shaped everyone differently.
this is survey of methods of mitigating hallucinations specifically in LLMs (which is a subset of deep learning systems) from 10 months ago, so it's already outdated as the progress in this field is moving at the speed of light (like semantic entropy ([Detecting hallucinations in large language models using semantic entropy | Nature](https://www.nature.com/articles/s41586-024-07421-0) advanced retrieval methods (has its own surveys you can search for), agentic systems (has its own surveys you can search for), new knowledge graph and GraphRAG methods (has its own survey you can search for), etc.) [[2401.01313] A Comprehensive Survey of Hallucination Mitigation Techniques in Large Language Models](https://arxiv.org/abs/2401.01313)
AI for healthcare is aligned with both effective altruism (EA) and effective accelerationism (e/acc)
One retweet about how AI will solve everything
One retweet about how AI will destroy everything
One nuanced tweet about how AI will revolutionize everything in both good and bad ways
One retweet about how AI wont change anything
I should make website AllAiSidesCovered dot com.
Perfectly balanced as all things should be.
This site on my website is pretty close [Future of humanity, AI, sentience, futurology, politics - Burny](https://burnyverse.com/OmniCortex/Future+of+humanity%2C+AI%2C+sentience%2C+futurology%2C+politics)
We should call neurotypicals neuroconvergents
posledních pár dnů jsem lítal po různých českých AI healthcare companies, je dost cool vidět co už se v praxi dělá
dneska jsem např potkal jedny co predikují a preventují heart failure pomocí lepší měřící EKG technologie a neuronkový technologie na zpracovávání těch dat, co už to nasazují do nemocnic 👍 a predikování a preventing jiných diseases pomocí měření a zpracovávání různých parametrů v očích
před dvouma dnama jsem ještě potkal týpky co měří tep víc dlouhodobě než jen na kontrolách (což dává víc dat na predikci problémů), nebo jiný co dělají AI therepeuta
v tomhle vidím jako jedno z největších potenciálů pro dobro pomocí AI
nebo tohle nedávno vyšlo mimo Česko:
BrainLM: A foundation model for brain activity recordings [BrainLM: A foundation model for brain activity recordings | bioRxiv](https://www.biorxiv.org/content/10.1101/2023.09.12.557460v1)
nebo (chatgpt like proste znamena ze oboji pouzivaj transformer architekturu) [96% Accuracy: Harvard Scientists Unveil Revolutionary ChatGPT-Like AI for Cancer Diagnosis](<https://scitechdaily.com/96-accuracy-harvard-scientists-unveil-revolutionary-chatgpt-like-ai-for-cancer-diagnosis/>)
[A pathology foundation model for cancer diagnosis and prognosis prediction | Nature](https://www.nature.com/articles/s41586-024-07894-z)
Hej teď zrovna surfuju co zrovna za health startupy v česku existuje, dost přemýšlím že nějakej joinu v ai talk channelu
Je mi jedno ze budu mit malo, ja chci dělat něco altruističtějšího co jde přímo vidět že má pozitivní dopad pro lidi.
Když jsem by před pár dny na setkání lidí co se hlavně se zabývají AI ve zdravotnictví, tak spousta z nich si stěžovala na to, jak prosazení nějaký technologický inovace v českým (a evropským) zdravotnictví může někdy trvat až 10 let kvůli těm nekonečným layers regulací a certifikátů atd., i když např studie už říkají že to funguje.
Ale trading firmy dost platěj no, teď nedáno jsem potkal jednoho týpka co dělal machine learning v CERNUu co se rozhodl že leavne research a zbohatne a založil si startup na machine learning predikování trhu.
Ale bylo mega cool ho poslouchat jak v CERNu hledají malý signály v denním flow petabytes dat.
To je jeden subset health tech firem, těch typů firem je hodně. Nejvíc cool mě příjdou firmy co např dělají něco jako foundational machine learning model na např predikci heart failure z víc high resolution dat než z EKG.
Obecně mě příjdou cool různý entity co dělají specializovaný foundational modely na biologii, fyziku, matiku
I feel like so many compsci (non)enjoyers don't care if they enjoy the work or if it fullfills them or if it helps the world, as long as it pays them billions, a mám pocit že to pro mental health často není dobrý
Tak aspoň v Česku mám pocit, že si buď vybíráš mezi zajímavou a fullilling prací nebo finanční buducností.
Ideal system would offer both in one much more often <:Sadge:961746847834202132>
The last few days I've been flying around various Czech AI healthcare companies at events, it's pretty cool to see what's already being done in practice!
For example I met some who predict and prevent heart failure using better heartbeat measuring technology measuring in higher resolution and deep learning technology to process the data, and they are already deploying it in hospitals. 👍
Or other others predict and prevent other diseases by measuring and processing various parameters in eyes.
I also met some guys who measure heart rate more long term than just at check-ups (which gives more data to predict problems)
Or others who do AI psychotherapy.
Outside of Czechia there's BrainLM, a foundation model for brain activity recordings that can predict futur brain states and predict various conditions such as PTSD and anxiety, or foundational models for cancer detection, or of course AlphaFold predicting proteins and helping in drug design etc.
I see AI in healthcare as one of the biggest potentials for doing good with AI!
I want to feed all these 86 billion babies in my brain with proper nutrients and care
Birth of a neuron
https://x.com/ScienceGuys_/status/1848314237287813166?t=D7JCSBL2nOAvAULGLYSRRg&s=19
"
We've always defined ourselves by the ability to overcome the impossible.
And we count these moments, when we dared to aim higher, to break barriers, to reach for the stars, to make the unknown known.
We count these moments as our proudest achievements.
But we had lost all that.
We had just forgotten that
we are still pioneers,
we've barely begun,
our greatest accomplishments cannot be behind us,
as our destiny lies above us.
"
We've just barely begun https://x.com/burny_tech/status/1848500897099485362
I think most discussions about reasoning and intelligence is people talking past each other as everyone is using different implicit definitions, more or less technical and formal, or vibes, which desynchronizes everything
The future will be so bright you can't even imagine
We will build it into even brighter form relative to what our current minds can comprehend
Use AI to accelerate carbon capture technology research and development!
Climate change carbon sink is ducked
https://x.com/webdevMason/status/1848374371234242810?t=z2ivfDkSv5CpsXG0Adqysg&s=19
Everytime I see a robot at irl event, he instantly becomes a friend
Unleash your inner truth seeking curiosity and go on an adventure to explore and understand the equation of reality by watching Stanford and MIT lectures on classical mechanics, statistical mechanics, special and general relativity, quantum mechanics, quantum field theory, standard model of particle physics, quantum gravity, and so on, for free on YouTube
Swim in infinite awe, become one with the universe
https://x.com/burny_tech/status/1848896929494642782?t=sUCUDU8vbuJ-fgeVW2MjCQ&s=19
[Unsupported browser](https://open.spotify.com/track/6pWgRkpqVfxnj3WuIcJ7WP?si=bCW0Vo0fS8CLE2gjzr_rfw)
Reverse engineering the equations of intelligence and the universe and using that knowledge to build stuff for the benefit of all sentient beings
"
Not doing alignment anymore?
To taky furt sleduju, a sem tam si s tím hraju.
Ale nevím, jsem conflicted na tolika úrovních z toho jak má všechno v našem systému komplexní implikace a z toho s čím nejvíc vibuju a nejradši bych chtěl dělat hodně věcí zároveň.
With AI steerability I'm starting to sense that accelerating steerability/alignment/interpretabiloty will accelerate the power of the elites over nonelites in the current system as they will use the benefits of that the most, while for example AI for healthcare is more direct (often but not always distributed) good for all...
But I guess how much of a priority any of this is depends on how probable you see various risks of dystopias, catastrophies, but also potential benefits if technology is used for good and how stable is that and so on... I feel like wanting to amplify the good aspects of technology more recently by building these aspects and talking about these aspects...
But you also need interpretability, robustness, etc. of AI in healthcare, so these two are also related in some ways.
I guess I'm also constantly stuck in analysis paralysis and constantly reevaluating my priorities and therefore not doing as much of what I want to.
"
Some people just want to see only the negative and ignore all the positive in the world. We can culturally heal them.
It's interesting how much often people's AI policy reflects their political views
European memeplex around me, I will not submit to your technopessimism
Is decentralization the way to prevent cyberpunk dystopia attractor? https://x.com/DaveShapi/status/1849918556902986073?t=8MReU81BzaLZmSIPVKETjA&s=19
Decentralization is all you need against centralization of power?
[https://youtu.be/qSBLSFfZ0us?si=Ppn938OJi5oTEmNx](https://youtu.be/qSBLSFfZ0us?si=Ppn938OJi5oTEmNx)
https://x.com/burny_tech/status/1850217989540372989?t=sdtC5FivRPcyG291RGzp0A&s=19
Two kinds of people that probably traumatize me the most are completely radical pessimists seeing only the bad in everything and ignoring all the good, and people who's only goal in life is to have infinite money without caring for anyone else in the process
I love Claude Sonnet 3.6 creations in Minecraft!
https://x.com/adonis_singh/status/1850460264358977587
If you feel not so well: It is just a dark cloud passing across the sun of your soul.
AI pro healthcare, vědu, matiku, engineering, technologie zlepšující kvalitu života apod.
a kreativita a companions je taky cool
a artifical life mi příjde taky mega cool
a automatizace boring věcí aby člověk měl víc času na to co chce dělat
nebo upgrading lidský formy
AI for healthcare, science, maths, engineering, developing technologies that improve quality of life, creativity, companionship, automating boring stuff to give people more time to do what they want to do, upgrading the human form, artifical life!
Travelling forwards in time, backwards in time, upwards in time, downwards in time, right in time, left in time, in complex time, in negative complex time, in multiple directions at the same time in a superposition in loops and interfering wave functions
[https://youtu.be/MzkgWDCucNY?si=pXPh7uWSwBxjyOJR](https://youtu.be/MzkgWDCucNY?si=pXPh7uWSwBxjyOJR)
Healthcare for all
Basic needs met for all
Metametamathematical structure including all possible and impossible, computable and noncomputable, implementable and nonimplementable, finite and infinite, etc. structures https://x.com/algekalipso/status/1852015093254648044
I'm curious when will neural celluar automata explode and get more mainstream! https://x.com/ABhargava2000/status/1851774127113384086