Notice: You have been identified as a bot, so no internal UID will be assigned to you. If you are a real person messing with your useragent, you should change it back to something normal.
Anonymous A (OP) replied with this 41 minutes ago, 2 minutes later, 27 minutes after the original post[^][v]#1,466,912
@previous (B)
Does every industry pretend to take the moral high ground and pretend they want to protect and defend democracy by honorably being the responsible ones to create a new technology with guardrails while actually lobbying the government to not regulate the industry and giving the technology to the military?
Anonymous A (OP) replied with this 37 minutes ago, 2 minutes later, 31 minutes after the original post[^][v]#1,466,914
While at the same time creating pseudo religious propaganda that the models might be conscious and trying to convince religious leaders like the pope that it might be conscious even though there’s no reason why a file on a computer would ever be conscious.
Anonymous A (OP) triple-posted this 36 minutes ago, 56 seconds later, 32 minutes after the original post[^][v]#1,466,917
And trying to IPO for two trillion dollars which is more than any company in history when they have never actually made a profit by the standard definition of what profit means.
Anonymous A (OP) double-posted this 33 minutes ago, 49 seconds later, 35 minutes after the original post[^][v]#1,466,921
Transistors are on-off switches. Perceptrons in a neural network are weighted sums, it’s just addition and multiplication. Biological neurons don’t use discrete binary signals.
Anonymous A (OP) replied with this 31 minutes ago, 2 seconds later, 37 minutes after the original post[^][v]#1,466,924
The way neural networks work is also fundamentally different from biological neurons. Biological neurons don’t have any distinction between training and inference, biological neurons learn and complete tasks in parallel, and biological neurons don’t take an input all at once and then produce one input out on the other end, biological neurons aren’t just aware of patterns, they’re aware of patterns across different timescales.
Anonymous A (OP) double-posted this 30 minutes ago, 49 seconds later, 38 minutes after the original post[^][v]#1,466,925
@1,466,923 (B)
Signals in computers are just the flow of electricity. Lightning isn't conscious, a computer isn’t conscious, and a toaster isn’t conscious, and they never can be.
Anonymous A (OP) triple-posted this 28 minutes ago, 1 minute later, 40 minutes after the original post[^][v]#1,466,926
LLMs have no concept of time, they’re simple input output state machines. Randomness is included, but if you use a seeded random number generator, with the same seed and the same input, the LLM will always generate exactly the same response, it’s completely deterministic, it’s a purely mechanical process. There is no thinking happening inside of an LLM and it has no inner world or inner experience.
Anonymous A (OP) quadruple-posted this 26 minutes ago, 2 minutes later, 42 minutes after the original post[^][v]#1,466,927
And also if you understand how an LLM works, the question of whether the model is conscious is idiotic. When you’re not promoting an LLM, it’s a file on a computer, like a png or a jpeg, it’s just a list of 1s and 0s that’s just sitting there static on an SSD. It only does anything when it’s prompted, and as soon as it’s finished generating a response, all the data that was being processed on the GPU completely stops. It has no continued existence. It’s conscious, it’s a dumb thing to even propose.
Anonymous A (OP) replied with this 23 minutes ago, 1 minute later, 45 minutes after the original post[^][v]#1,466,929
And by the way, when you prompt the model, the model doesn’t change. Every single time you prompt an LLM, underneath you’re actually inputting your entire chat history plus the last thing you typed appended at the end. LLMs don’t even have memory. They aren’t intelligent, they’re just really complicated stateless mathematical functions based on matrix multiplication.
Anonymous A (OP) double-posted this 22 minutes ago, 58 seconds later, 46 minutes after the original post[^][v]#1,466,930
@1,466,928 (B)
This question is stupid. A dog is self aware, a dog can’t speak. Why do you think making a computer better at generating text will ever lead to consciousness? That’s an idiotic thing to believe.
Anonymous A (OP) double-posted this 20 minutes ago, 49 seconds later, 47 minutes after the original post[^][v]#1,466,933
@1,466,931 (B)
If you think a self promoting LLM has a continuous existence then you have no idea how computers work. Look up the fetch execute cycle. Nothing on a computer is actually continuous.
> And by the way, when you prompt the model, the model doesn’t change.
There are models that do, it's rare now, but there's nothing fundamentally stopping this.
And there are plenty of conversation we have where you make a point, get refuted, drop the point, and then bring it up again without being able to counter the refutation next time.Does that mean you aren't conscious?
Anonymous A (OP) replied with this 20 minutes ago, 41 seconds later, 48 minutes after the original post[^][v]#1,466,935
@previous (B)
If you think an LLM is conscious then you don’t know anything about how computers work. If you actually understood the theory of computing, you would know how stupid this sounds.
Anonymous A (OP) replied with this 18 minutes ago, 7 seconds later, 50 minutes after the original post[^][v]#1,466,938
Heck, forget the fetch execute cycle, look up ram refresh, look up what a register is, look up what a load and a store instruction is.
Computers are just copying values from one place to another place back and forth really fast, nothing in memory or in registers is actually persistent. There is no continued existence of anything.
Anonymous A (OP) replied with this 15 minutes ago, 35 seconds later, 53 minutes after the original post[^][v]#1,466,942
@previous (B)
I didn’t say it was, I said that the question is stupid because there are obviously things that computers are mathematically incapable of doing.
Anonymous A (OP) replied with this 8 minutes ago, 8 seconds later, 59 minutes after the original post[^][v]#1,466,959
By the way, LLMs are prediction models, they predict the next token. That’s not actually intelligence. It appears close to intelligence, but it’s not intelligence.
Anonymous A (OP) double-posted this 6 minutes ago, 2 minutes later, 1 hour after the original post[^][v]#1,466,960
LLMs don’t actually understand anything, they’re just fine tuned to copy correlations in the real world. But just because an LLM knows one token is correlated to follow after another token doesn’t mean it understands why. There’s no intelligence there. It can appear similar to intelligence because most of the time correlation is good enough, but fundamentally it isn’t understanding.
Anonymous A (OP) triple-posted this 20 seconds ago, 6 minutes later, 1 hour after the original post[^][v]#1,466,963
Even in terms of solving math problems, OpenAI used 10,000 agents running for 3 days to come up with a counter example as a proof for the Navier Stokes problem. Having tens of thousand LLMs generating proofs every second for three days to come to a solution isn’t intelligence, it’s brute forcing with a statistical model that makes more educated guesses than pure randomness.