@MC_chibill @Navier_Tokes @lucasmeijer 'Pro vaccine papers are big pharma propaganda! Read who funds that shit!' See how you're just the same as the people who are rotting society?
English
zzz
20 posts
English
Everybody who thinks ai is conscious has to do a mandatory from scratch transformer implementation. There are only floats and multiplications.
English
@MC_chibill @Navier_Tokes @lucasmeijer You're right that LLMs are trained to predict the next token. You're wrong that no ideas are being represented in the weights and activations of MLP layers. Here's a scientific paper clearly demonstrating this. transformer-circuits.pub/2024/scaling-m…
Are you claiming the paper is fake?
English
@bigwalrus18 @Navier_Tokes @lucasmeijer The evidence is literally every single document on how ML and NN works. You can keep acting like a MAGA idiot that reject the truth.
English
@MC_chibill @Navier_Tokes @lucasmeijer We clearly both know you are wrong now. Are you gonna apologize and stop acting like a retard?
English
English
@MC_chibill @Navier_Tokes @lucasmeijer You know more than him and every other researcher working on this stuff?
English
@MC_chibill @Navier_Tokes @lucasmeijer 'Mech interp aims to understand the internal workings of neural networks by analyzing the mechanisms present in their computations.'
Why would this be an ongoing field of research if we already know everything about how neural networks work according to you?
English
@bigwalrus18 @Navier_Tokes @lucasmeijer Literally just trying to figure out exactly how the probabilities are encoded. This is like basic intro shit. Are you that dumb?
English
@MC_chibill @Navier_Tokes @lucasmeijer youtube.com/shorts/zKM-msk…
Here, nobel prize winner. You disagree?
YouTube
English
English
English
@bigwalrus18 @Navier_Tokes @lucasmeijer You train the full NN on the. Ext token, encoding its probabikity inti all layers. For the love of god, LEARN HOW LLMS WORK.
English
@MC_chibill @Navier_Tokes @lucasmeijer Someone should tell Anthropic MC_chibill on twitter has got it all figured out. They are clearly wrong since activations only encode next token probabilities right? Go enlighten those stupid researchers!!
Anthropic@AnthropicAI
New Anthropic research: Natural Language Autoencoders. Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude’s thoughts, but not in a language we can read. Here, we train Claude to translate its activations into human-readable text.
English
@bigwalrus18 @Navier_Tokes @lucasmeijer Maybe understand the vwry basics how NN and LLMs work before commenting on the interent. The NNs in an LLM encode the probabilities of the next token to decide what the next token should based on the previous input tokens.
English
English
English
It is an LLM.
If you know how LLMs work, this explains it succinctly and thoroughly. If you do not, you should not be opining on AI consciousness at all.
Cristian Garcia@cgarciae88
claude is most likely not conscious but I haven't read a single post explaining why not
English
@MC_chibill @Navier_Tokes @lucasmeijer Lmfao wtf does 'probabilities' being stored in weights even mean?? How can you talk about basics when 'weights encoding learned ideas' is literally the whole point of deep learning?😂😂
transformer-circuits.pub/2024/scaling-m…
Read that and educate yourself before being so confidently wrong.
English
@bigwalrus18 @Navier_Tokes @lucasmeijer There is no "ideas" represented in the weights, probabilites are. Learn the basic first.
English
English
English
@MC_chibill @Navier_Tokes @lucasmeijer Nope! People obviously know the computations that are being done but they don't know exactly how ideas are being represented in the weights and activations. You can read here. Stop being so arrogant.
en.wikipedia.org/wiki/Mechanist…
colah.github.io/about.html
youtu.be/riniamTdUSo?si…
YouTube
English
@bigwalrus18 @Navier_Tokes @rikarends @lucasmeijer Literally everyone that works on AI like OpenAI, Anthropic, Google, hell even I know how it works. Its very very fucking simple. The fact that your to stupid to understand it means nothing.
English
English
@bigwalrus18 @Navier_Tokes @rikarends @lucasmeijer We know exactly how the "complex" (nothing complex about token prediction. LMAO) behavior works.
English
@esefesen @ZVeryn96103 @reset_by_peer Do you think if we simulated your brain on silicon with a super fast rate of inputs and outputs it would be conscious?
English
@bigwalrus18 @ZVeryn96103 @reset_by_peer It's clear to me AI as it currently exists is an imitation of a conscious being, not an actual one as it doesn't have any standalone behavior that wasn't explicitly programmed in. A human with no human education would still mostly behave like a human.
English
English
@ZVeryn96103 @reset_by_peer This such an arrogant asshole take. If you’ve spent like at least 30 minutes with an animal, you know they have souls, that they’re conscious of everything. And they’re smart. So fuck off with this bullshit.
English
English
@ZVeryn96103 @reset_by_peer Subjective experience, qualia, however you want to call it, thats my definition of consciousness. Now prove AI has that
English
@MC_chibill @Navier_Tokes @rikarends @lucasmeijer Of course we know what computations are being done but we don’t actually know how complex behavior arises from those computations. There’s a whole field dedicated to this topic called Mechanistic Interpretability. You are being ignorant.
English
@Navier_Tokes @rikarends @lucasmeijer We know how LLMs work as a whooe at every level. They are reallt fucking simple. 3Blue1Brown as more videos on the subject but considering you cant even grasp the inital one I sent, I dont have ant hope you could be alstart to understand the others.
English