7,940 karma · joined January 24, 2012
What things that you have never seen do you not believe in?
The fact that the pilots denied that they had shut the switch (one asking the other why they had done so and the other denying it), and that they restarted the engines should be taken into account. Ok, murder suicide is definitely on the table but I would want to see some other reasons for believing that this is so.
Still provides power today.
I don't see why that's necessary for the application... so how about the default is for service_role not to be given to something that's insecure?
How are the different components of these prompts contributing to the result... what happens if one word is changed? what about two words?
I can't quite get this - surely until we have an execution on the proper hardware we can't accept that it works? There are engineering problems to resolve before we can be confident - perhaps they can be easily resolved, but so far they haven't.
I would be very curious to learn what the barriers to a demonstration of Shores on an arbitrary 8bit prime are...
I think that the trained transformer has fixed weights and therefore cannot learn.
I think learning is one aspect of reasoning, and is demonstrated by challenges like navigation or puzzle solving where learning that one route to a solution is impossible is important.
I also think that the single forward pass of the model means that cyclic reasoning isn't feasible and that conditioning output by asking the model to "think" even when that thinking is done on the single forward pass means that logical processes are ruled out. The model isn't thinking in that case, the probabilities of the final part of the output are conditioned by requiring a longer initial output.
Also a financial leverage story. Rendering the true market value of these towers would create balance sheet issues at quite a few institutions. No one is keen to rock that boat.
American suburban architecture is pretty bad, but at least it's disposable. Generations to come won't have to live in these things.
You can taste a beer, laugh so much it hurts, come to know how something works.
The mechanism by which they work prohibits reasoning.
This is easy to see if you look at a transformer architecture and think through what each step is doing.
The amazing thing is that they produce coherent speech, but they literally can't reason.
If you run an LLM with optimization turned on on a NVIDIA GPU then you can get non-deterministic results.
But, this is a choice.
If it's in a book AI can do it.
Asking if it's in the book, maybe not.
Not in the book, definitely not.
LLMs have real problems with both of those things, they are unreliable in the sense that they produce text with factual inaccuracies and they are uncontrollable in that it's really hard to predict their behaviour from their past behaviour.
Big open challenges in AI - I am sure that there will be solutions, but there aren't right now. I think it's very like self driving cars. Implementing a demo and showing it on the road is one thing, but then it takes twenty years before the slow roll out really gets underway.
Other people are not sensible - look at all the junkies shooting who knows what into who knows where. A system that can advise cooks to put glue on pizza is probably not the best mechanism for providing healthcare advice to at least the section of the population who are likely to try it.
On the other hand many people seem unwilling to accept the reports of others that they are conscious and have freedom of will and freedom to act. At the same time these people do not live as if others were not conscious and bereft of free will. They do not watch other people murdering their children and state "well they had no choice". No they demand that the murderers are punished for their terrible choice. They build systems of intervention to prevent some choices and promote others.
It's not orthogonal, it's the motivating force for our actions and changes our universe. It's the heart of the matter, and although it's easy to look away and focus on other parts of the problems of intelligence at some point we have to turn and face it.
We're talking the physical version right? I don't have any counter examples that I can describe, but I could hold that that's because human language, perception and cognition cannot capture the mechanisms that are necessary to produce them.
But I won't as that's cheating.
Instead I would say that although I can't disprove PCT it's not proven either, and unlike other unproven things like P!=NP this is about physical systems. Some people think that all of physical reality is discrete (quantized), if they are right then PCT could be true. However, I don't think this is so as I think that it means that you have to consider time as unreal, and I think that's basically as crazy as denying consciousness and free will. I know that a lot of physicists are very clever, but those of them that have lost the sense to differentiate between a system for describing parts of the universe and a system that defines the workings of the universe as we cannot comprehend it are not good at parties in my experience.
>For this to be relevant, you'd need to show that there are possible sensory inputs that can't be simulated to a point where the "brain" in question - be it natural or artificial - can't tell the difference.
I dunno what you mean by "relevant" here - you seem to be denying that there is a difference between reality and unreality? Like a Super Cartesian idea where you say that not only is the mind separate from the body but that the existence of bodies or indeed the universe that they are instantiated in is irrelevant and doesn't matter?
Wild. Kinda fun, but wild.
I stand by my point though, computing functions about how molecules interact with each other and lead to the propagation of signals along neural pathways to generate qualia is only the same as tasting beer if the qualia are real. I don't see that there is any account of how computation can create a feeling of reality or what it is like to. At some point you have to hit the bottom and actually have an experience.
For example learning from experience (which LLMs cannot do because they cannot experience anything and they cannot learn) is clearly an attribute of an intelligent machine.
LLMs can tell you about the taste of a beer, but we know that they have never tasted a beer. Flight simulators can't take you to Australia, no matter how well they simulate the experience.