hackinthebochs

I would say most people do in fact need a life of purpose and fulfillment through some aim larger than themselves that has intrinsic value, can reinforce self worth, etc. I think very few people will find purpose and fulfillment through and endless stream of games, movies and parties. The question is what do those people lacking purpose do with their time as they try to find it, or how do they cope with a life of purposelessness?

We could also just stop research into LLMs. Or keep them air gapped. Tell me how that's going?

For any real-world action required to allow an AI to escape some manner of containment, there will always be a person willing to do it out of hubris/ignorance/nihilism.

>without an actual understanding of what they're doing.

At what point do you start to question your assumptions that are causing you so much cognitive dissonance?

But to answer your question: to predict the next token really well you just have to model the world. Think of it like this, a simple statistical model might say "when token A is seen respond with token B". The next step will add conditions, "...respond with token B unless X has been seen, then respond with Y". Add a few billion more of these contexual clauses and you have a sequence of logical rules that indirectly model the relevant processes in the world.

The stacked transformer paradigm picks out points in circuit design space. It is very possible this architecture has no inherent limitations on what it can compute in principle.

Yes, they have some alien failure modes. But that should be expect, they are an alien intelligence. I might be willing to grant that a lack of ability to reflect on its own level of knowledge is a demerit to it being generally intelligent. But then again it is largely an artifact of training. I suspect if there were a guessing penalty during pretraining they would develop or more readily communicate the strength/reliability of their knowledge.

But it's not relevant as a metric to gauge distance to human intelligence. Humans see individual letters, LLMs do not. If I asked you the relative activation of the cones in your retina as I showed you some solid color image, you couldn't do it. You simply do not have cognitive access to that information. But that says nothing about your intelligence.

A more accurate test would be to give it a list of words (or anything represented as a single token) and ask it how many times that token appeared. I'm sure they have no trouble at that task.

Tokens are the most basic input unit of an LLM. But tokens don't generally correspond to words or letters, rather sub-word sequences. So Strawberry might be broken up into two tokens 'straw' and 'berry'. It has trouble distinguishing features that are "sub-token" like specific letter sequences because it doesn't see letter sequences but just the token as a single atomic unit. 'Straw' and 'r' are two tokens but an LLM is entirely blind to the fact that 'straw' has one 'r' in it.

As an analogy, I might ask you to identify the relative activations of each of the three cone types on your retina as I present some solid color image to your eyes. But of course you can't do this, you simply do not have cognitive access to that information. Individual color experiences are your basic vision tokens.

>You said “Given causal closure, any causal influence is explained by prior physical events, thus bypassing any relevance of this special property.”

You're taking the sentence out of context and missing the work it is doing. Given some special property, if it doesn't have causal influence, then it is explanatorily impotent. If it does have causal influence then it is just a standard physical property subject to laws and such, i.e. non-special.

> For example, a wire’s conductivity is produced by its physical microstructure. Its conductivity helps explain why current flows through it. The fact that the conductivity itself has a physical explanation does not “bypass” its relevance.

Yes, conductivity is just a physical property (in my quote you left out the parenthetical about being identical to some physical event). It is explanatorily relevant because it is definable within the explanatory regime of physical dynamics.

>Likewise, suppose some biochemical property of neurons is essential to consciousness.

This is the meat of the issue so I will focus on this. The question is what kind of property could a biochemical property be that is not abstract and is crucial for consciousness? My argument is that there are no suitable properties to play the necessary role of grounding consciousness in the "real world" while denying it in a virtual/simulated world. The remainder of the space of candidate properties are non-physical, unscientific, and ultimately wishful thinking.

Given the idea that biology/chemistry reduces to physics, this strongly constrains the space of candidates for a biochemical property to ground consciousness. The content of physics is structure, dynamics, conservation laws, fields, probability, etc. The equations of physics don't describe consciousness, but the behavior and constraints on the entities quantified over. Note that as the equations of physics quantify over relevant entities, they are abstract. To say a biochemical property is crucial for consciousness is just to say that a physical property, broadly construed, is crucial for consciousness. Candidate properties can be divided into two sets: the structure, dynamics, etc of the content of physical theories (we can call these relational properties), and any non-relational/intrinsic properties that are not the content of physical theories but are relevant to the realization of physical events. These latter properties are not the subject matter of physics and we are epistemically blind to whatever they may be.

Crucially, causal closure together with the causal exclusion argument renders the intrinsic properties explanatorily empty with regards to any state reducible to the language of physics, i.e. structure, dynamics, etc. As it turns out, our verbal reports about consciousness can be described by stimulated wave patterns on the air. In other words, these verbal reports can be described in the language of physics. Therefore, by causal closure, they are fully explained by prior events within the subject matter of physics. By causal exclusion, any intrinsic properties of the realizers of these physical events are excluded from explanatory relevance. There is simply no room for "non-abstract biochemical properties" to be explanatorily relevant to verbal reports about consciousness. But if verbal reports about consciousness are to actually be about consciousness, then consciousness must be explanatorily relevant to said verbal reports. Hence consciousness must be definable within the content of the subject matter of physics.

The only saving grace for biological theories are non-causal, metaphysical influences that are somehow relevant to verbal reports. Or epiphenomenalism where consciousness is actually irrelevant to all the things we say about consciousness. Both of these options are unintelligible; they are cures worse than the disease of functionalism. Some like to imagine some kind of emergence, such as "when physical structures take on specific forms, certain states of consciousness manifest". A simulation wouldn't take on this form in the real world and so a simulation wouldn't be conscious. But laws of this nature are absurdly adhoc and don't offer explanations, only brute laws.

A corollary of causal exclusion is that two causally identical systems have identical explanations for identical physical events. If you posit some property of a substrate purported to be relevant for consciousness, it will either have a causal influence or non-causal influence. Given causal closure, any causal influence is explained by prior physical events, thus bypassing any relevance of this special property (or this special property is just identical to some physical event). If the influence of this special property is non-causal, then due to causal exclusion it can't have any explanatory relevance to causally downstream behavior. E.g. it is causally excluded from relevance to speech acts about consciousness. You're left with some kind of non-physical, non-causal influence. In other words, wishful thinking.

Two further points. Modern physics has been described as "topic neutral"[1] in the sense that it doesn't posit the intrinsic nature of the substrate being described, only describes its behavior. Thus physics properly understood is itself multiply realizable: the behavior of physics can be "implemented" in dissimilar substrates with a range of intrinsic properties. But, according to the argument just outlined, these intrinsic properties can't play any explanatory role in the unfolding of physical dynamics (outside of bare implementation).

The other point is that this argument is fully general in the sense that any closed explanatory regime causally excludes phenomena not captured by entities defined within the explanatory regime. The explanatory regime of abstract semantic information is another such closed regime. Thus, for our speech acts, i.e. abstract semantic information, to be about consciousness, consciousness must be captured within the explanatory regime of abstract semantic information.

[1] https://chomsky.info/201401__/

It only takes a finite number of rules to generate a mapping between unbounded strings of finite symbols. The brain doesn't contain an infinite amount of information after all.

Causal integration + causal closure is a wrap for non-abstract theories. The so-called biological theories of consciousness are just wishful thinking.

One's subconscious drives much of your thoughts, yes. But important decisions are conscious mediated if only because consciousness has veto power. A lack of veto is a tacit endorsement of the action. That is control.

The democratization of intelligence is irrelevant if there isn't also a corresponding democratization of raw resources. In the before-times, labor and intelligence were bottlenecks to turning raw resources into wealth. This is what drove the economy and the vast labor pool since the industrial revolution. In the coming AI revolution, labor and intelligence will no longer be bottlenecks. Those that control the raw resources won't need to employ most of the world to maximize the wealth potential of the resources they control. Those who aren't relevant to the standard of living of the new aristocracy are nothing more than a nuisance to be controlled or disposed of. Totalitarianism is the most likely outcome.

It's really the only live option once the issues are properly understood. Causal closure: a physical event is explanatorily exhausted by prior physical events. Causal exclusion: if physical properties fully cause/explain everything we say and do about consciousness, i.e. our behavior purported to be about consciousness, then consciousness cannot be distinct from the physical behavior and also be a cause of that behavior. For example, for our speech acts to be about consciousness, consciousness must be causally involved in their generation. Hence, consciousness is a second-order physical property of the same category as the information carried by the speech acts about consciousness. That is, consciousness is an abstract property of physical organization.

If you're referring to the Libet experiments, they have been wildly overinterpreted. The conceptual problem with the common interpretation is that detecting a readiness potential prior to a report of a conscious intention to act does not entail the readiness potential is distinct from that conscious intention. The readiness potential may in part constitute one's conscious experience and so cannot be separated from it.

This just alters the thought experiment. What the thought experiment offers is a book of opaque rules, just instructions on how to transform symbols into more symbols without any concern for their meaning. What you're offering is a translation dictionary, what does it mean when one encounters X symbol. You're giving him a book on learning Chinese.

>How does a person's brain containing all the facts, information, and skills that it needs to have a fluent conversation in Chinese imply that the person understands Chinese? How does it not!?

Notice that the man while having memorized the rulebook can't convey any fact of his childhood in Chinese. He can only convey facts encoded in the rulebook, perhaps of a hypothetical childhood. The point is there are different ways information is stored in the brain; one way is through rote memory, the other is something like semantic memory. The rulebook encoded as rote memory doesn't endow his semantic memory contact with the meaning of the symbols. Hence the man doesn't understand Chinese.

This doesn't render the Chinese argument successful. The system's reply still stands when understood properly. I like to call it the virtual mind objection. Physically executing the rules in the rulebook just instantiates a virtual mind. Whether the man is reading the rules from the book or just pulling them from memory is irrelevant.

>So why would it be different for the brain? If nuclear energy can’t arise from abstract relationships, why would consciousness or anything else?

Because consciousness isn't something the brain produces like a hormone or an excited particle. Consciousness isn't an output of the brain that is somehow separate from it, it is the manner in which the brain processes sensory information and makes decisions to produce output.

The causal context in which brain dynamics occurs is orthogonal to the relevant properties regarding consciousness. Consciousness is a state characterized by certain kinds of information processing, namely that which carries relevant semantics of the organism encoded in a way that maximizes effective control by the executive center. Consciousness is the rich semantic space in which the executive center engages to affect highly competent control of the organism. It is substrate independent information processing.

Pair programming is the extroverts answer to the solitariness of software development. They love it while everyone else is hating their existence. It works for short stints to solve specific problems but anything beyond that is just gratuitous.

I really don't know where you're coming from. You seem to be imagining that the law requires some kind of cryptographically validated age attestation. The law requires nothing of the sort. It's just an age field associated with the account. There are many ways to get around it, none of which the law demands protection against. It's simply a standardized way for parents to ensure a modicum of reasonably effective age-discrimination for online services.

There are no secure enclaves, no verification, and certainly no black markets. I really don't know where you are getting any of this from.

>TX passed a law stating that adult sites like pornhub must verify age.

Yes, when states are left to their own devices to come up with laws, some of them will be more onerous than others. But I don't know what you think you can infer from this. Casual dismissal was warranted.

>Also children can "come upon" devices

Yes, this mechanism isn't perfect, but is a substantial improvement over the status quo.

>and now that's stuck with the verified age set.

This inference is nonsensical.

>If "just let someone report the age" worked, pornhub wouldn't be blocked in TX.

Also nonsense

>but why can't the websites just ask me for my self declared age?

Because children just click through the age gate when it suits them. With this legislation, a parent that purchases the device and creates an account for the child can set the age once and take the decision out of the kids hands. It's a huge improvement without any significant privacy issues. I can't fathom why the tech crowd is having a collective aneurism over this.

>OS isn't the thing displaying content

But the account on the OS is the right place for the single source of truth of properties of the current user.

Why are you people in such a hurry to have facebook et al collecting government IDs?

Parents already have the ability to do this, at least when it comes to porn. The fact is that most parents aren't doing it. Hence the push for legislation that increase the uptake of content blocks for minors.

There's something to be said for realizing which way the winds are blowing and getting ahead of it. The fight against any and all age gating on the internet is already lost. The question is what will the system look like once it comes into being. This is something we have a lot of say over if we just stop being obstinate and start offering solutions. Sticking one's head in the sand will just result in a worse state of affairs in the end.

The breathless fearmongering over an age field on account set up is just completely over-the-top. This is probably the least bad out of all possible ways to implement age checking. The benefit of this is that it can short-circuit support for more onerous age verification. The writing has been on the wall for some time now: the era of completely unrestricted internet is coming to an end. The question is how awful will the new normal be? Legislation like this is a win all around, a complete nothingburger. We should be celebrating it, not fighting it tooth and nail.

What does "hold these companies responsible" look like to you? I see a lot of empty rhetoric without things being spelled out. If you just want these companies to not exist, just say that.

Why exactly are some of you guys advocating for platforms to have increased responsibility for verifying the age of their users? Why exactly do you want platforms to collect government IDs or contract with third party ID brokers? How is that an improvement over the user agent attesting to the age of the user?

Yeah, we really should just storm the facility where Musk is hoarding all the worlds bread and meat.

Wealth in terms of capital doesn't represent material goods, it represents the system's confidence in your ability to direct capital efficiently. But eventually efficient capital bottoms out at consumable goods. Someone like Musk with a lot of capital under his control is contributing to the end goal of unlimited abundance.

While that's true, it's unrelated to hosting code. A reliable code hosting site should be agnostic to content. Once they start making editorial decisions about what projects to host, why should anyone trust them with their project?