Proof of Immortality III

[jond]

Dave,

- I would say that it depends upon all three...

- The "self" I'm talking about is difficult to effectively communicate/define. The best I can do, perhaps, is it's the self concept that reincarnationists think keeps returning.
- Whether thing, process or illusion, science must say that it (as a totally specific thing, process or illusion) must be brought into existence by laws of physics, chemistry and biology. And either, we could replicate this thing, process or illusion by replicating the PC&B, or we couldn't -- it being a second level emergent property that is simply unique, however many times we replicate the PC&B. Toon suggests (as I understand Toon) that the specific time would have to be included in order to replicate a specific self...

Jabba: do you understand that the self, according to the scientific paradigm, is not an entity that could exist separately from our bodies, but rather it is a process that is shaped and determined by every experience a person has over the course of their lives? The only way to replicate it would be to replicate each and every experience that person has had, which is of course impossible. This been explained so many times, and yet you comtinue to refuse to acknowledge your understanding.

 

[JayUtah]

The "self" I'm talking about is difficult to effectively communicate/define.

That isn't really true. The "self" you're talking about has gone through several attempts on your part to define, only to be refuted each time. You don't comment on the refutations; you just quietly issue more speculation. It's hard for you to describe because in four years you haven't managed to come up with a description that transcends the emergent properties of the brain. So now you just punt anytime you're asked.

The best I can do, perhaps, is it's the self concept that reincarnationists think keeps returning.

Then if this is your "reasonable alternative" you need to show it's reasonable. Simply pointing to a hypothesis doesn't make it a reasonable hypothesis. And if this is the hypothesis to which you are attaching numerical estimates of probability, then you need to show the numerical reasoning behind the estimate. Simply pulling numbers out of your kiester isn't a rationale.

Just don't, when you can't supply any of the needed rigor, back away and say you aren't married to the reincarnationist formulation. Either endorse it and construct an appropriate proof, or formulate your own hypothesis and construct an appropriate proof.

And either, we could replicate this thing, process or illusion by replicating the PC&B, or we couldn't --

No.

We've already been through this half a dozen times, using the Mt Ranier analogy you don't like. Replication is not a requirement of a process that arises from a complex natural system. Do not paste requirements on your critics that they do not actually espouse.

 

[jond]

Jabba: I really recommend this book. It's really well written, easy to read and understand, and the author is a leading neuro scientist, well respected in the field. It may not change your beliefs (I don't think anyone or anything could) but you will at least gain some understanding of where the scientific field stands today.
https://www.amazon.com/Tell-Tale-Br...id=1472057048&sr=1-1&keywords=tell+tale+brain

 

[godless dave]

Just to save time, let me stress that "replicate" means "make a copy of". When you make a copy of something, you now have two of that thing. Not one.

 

[Loss Leader]

This is some second-level nonsense from Jabba today: all sorts of numbers and equations and metaphysical dualism. He's really stepped up his usual nonsense. This is almost Amega Wand level crud.

 

[Monza]

9.1.2. Now I must estimate (roughly) the prior probability (rounded off to 3 decimal places) of each more specific possibility (hypothesis), given ~H.
9.1.2.1. That only some of us have but one finite life: .000
9.1.2.2. That we each have numerous finite lives: [highlight].200[/highlight].
9.1.2.3. That only some of us have numerous finite lives: .000
9.1.2.4. That we each have an infinity of finite lives; [highlight].200[/highlight]
9.1.2.5. That only some of us have an infinity of finite lives: .000
9.1.2.6. That we each have an infinite life: [highlight].200[/highlight].
9.1.2.7. That only some of us have an infinite life: .000
9.1.2.8. That time isn’t what we think it is (to be explained): [highlight].200[/highlight]
9.1.2.9. Some other explanation: [highlight].200[/highlight]

When you round off a number, it means that you have measured or calculated something to a finer precision. For example, to round off to three decimal places implies that you have calculated these probabilities to more than three decimal places. It seems odd that they all come out to a nice round number with all those zeros. It is not impossible (maybe you actually calculated these values at 0.20023, 0.19972, etc.) though I suspect that is not the case. I think you just made up these numbers then added some zeros on the end to give the illusion of precision and to obfuscate where these numbers came from.

 

[aleCcowaN]

It just sounded scientifickilly.

Was the concoction above Jabba's attempt to Bayes?

 

[JayUtah]

It just sounded scientifickilly.

Was the concoction above Jabba's attempt to Bayes?

Yes. Since he asserts Bayes lets you use "subjective beliefs" as priors, he takes that to mean he can just make up any number and have the result mean something.

 

[aleCcowaN]

Yes. Since he asserts Bayes lets you use "subjective beliefs" as priors, he takes that to mean he can just make up any number and have the result mean something.

Jabba's is a nightmare, that's why it's so difficult to be sure it's Bayes. He uses overlapping partitions, non exhaustive partitions and he cuts a boundless cake wherever he wants and assigns an arbitrary flavour and weight to each part.

It's just "Tweedledee-dee by 0.2, I carry a 3 and it adds up to, immortality! Yay!"

I suppose being afraid of death can lead to these ruminations, magic rags and other make-believes.

 

[jt512]

This thread is definitely bad for my health. I start giggling non stop when I read Jabba's probabilistic concoctions.

The contrast between his patronizing «The probability (“likelihood”) of E given ~H» and his «now, I must estimate the likelihood of my own current existence given the different specific hypotheses under» with total probability reaching way above 100% even when some overlapping is taken care of, is bone crashingly hilarious.

Jabba is saying that his hypothesis ~H (let's call it K to simplify notation) can be partitioned into n submodels, K₁, K₂, ... , K_n. He has an event E (he exists) and he computes the probability L_i of E under each submodel: L_i=P(E|K_i), i = 1,2,...,n. There is no reason that these n L_i's should add up to any particular number: they are probabilities of an event under different models (ie, events in different conditional sample spaces). Indeed, I have repeatedly tried to explain to Jabba that, for his hypotheses H and ~H, P(E|H)=1 and P(E|~H)=1, because he can only observe the universe that he exists in.

 

[I Am The Scum]

It just sounded scientifickilly.

Was the concoction above Jabba's attempt to Bayes?

I bet he puts on a lab coat every time he types up a new post.

 

[aleCcowaN]

There is no reason that these n L_i's should add up to any particular number

I was thinking in this particular bit

probability...
9.1.3.6. That we each have an infinite life: 1.00
9.1.3.7. That only some of us have an infinite life: .50

What's meaningless in sight of the aberrations in Jabba's """""""math""""""".

 

[jt512]

H

Yes. Since he asserts Bayes lets you use "subjective beliefs" as priors, he takes that to mean he can just make up any number and have the result mean something.

I don't really have a problem with made-up priors. Priors are just an expression of personal belief. There's no hard reason that any two people should agree on them; after all, we each bring different background information to any particular problem.

Furthermore, with a little algebra, we can write Bayes Theorem in terms of the odds of H, like this:

P(H|E) / P(~H|E) = [ P(E|H) / P(E|~H) ] × [ P(H) / P(~H) ] .

This shows that the posterior odds of H can be factored into the likelihood ratio and the prior odds, each factor contributing independently to the posterior odds. Since the data only affects the posterior odds through the likelihood ratio, even people who hold radically different subjective prior odds of a hypothesis should be able to agree on how much their odds should change given new data E. In this sense, Bayesian inference still "works" in the face of disagreements about the priors. But it requires H and ~H to make unambiguous predictions about the data. Thus, lack of consensus on the priors is tolerable, even expected. But lack of consensus on the likelihoods, which implies that the models (hypotheses) have not been adequately specified, is fatal.

 

[jt512]

I was thinking in this particular bit

What's meaningless in sight of the aberrations in Jabba's """""""math""""""".

The partition itself is bizarre. I suspect it was invented as an attempt to satisfy the criticism that ~H is a composite hypothesis.

 

[JayUtah]

...even people who hold radically different subjective prior odds of a hypothesis should be able to agree on how much their odds should change given new data E. In this sense, Bayesian inference still "works" in the face of disagreements about the priors.

Yes, and I called this out a couple times in a previous Jabba thread as one of the useful applications of this sort of reasoning. And as we've previously discussed and agreed, one of the great strengths of this type of modeling is the license to seed it with information that may be quantifiable according to informal expectations.

But lack of consensus on the likelihoods, which implies that the models (hypotheses) have not been adequately specified, is fatal.

Right; we always come back to P(E|H) and P(E|~H) and Jabba's wholesale guesswork for them.

 

[caveman1917]

I don't really have a problem with made-up priors. Priors are just an expression of personal belief. There's no hard reason that any two people should agree on them; after all, we each bring different background information to any particular problem.

Furthermore, with a little algebra, we can write Bayes Theorem in terms of the odds of H, like this:

P(H|E) / P(~H|E) = [ P(E|H) / P(E|~H) ] × [ P(H) / P(~H) ] .

This shows that the posterior odds of H can be factored into the likelihood ratio and the prior odds, each factor contributing independently to the posterior odds. Since the data only affects the posterior odds through the likelihood ratio, even people who hold radically different subjective prior odds of a hypothesis should be able to agree on how much their odds should change given new data E.

What's more, one can write it out further so that we have equal priors and implicit evidence E' such that an application of Bayes' theorem to those equal priors with E' gives the priors we started with. Given that successive applications of Bayes' theorem are an iterated multiplication, and multiplication is commutative, this further means that we can just consider E along with the equal priors first, and have agreement on that first result. We don't have to start with the priors people disagree on, we can change the order of the evidences (implicit or not) however we want.

If people disagree on priors they could just start with equal priors and leave their individual, disagreed upon, information for afterwards.

 

[Jabba]

Jabba is saying that his hypothesis ~H (let's call it K to simplify notation) can be partitioned into n submodels, K₁, K₂, ... , K_n. He has an event E (he exists) and he computes the probability L_i of E under each submodel: L_i=P(E|K_i), i = 1,2,...,n. There is no reason that these n L_i's should add up to any particular number: they are probabilities of an event under different models (ie, events in different conditional sample spaces). Indeed, I have repeatedly tried to explain to Jabba that, for his hypotheses H and ~H, P(E|H)=1 and P(E|~H)=1, because he can only observe the universe that he exists in.

jt,
- Please try explaining that again.
- If H is true, isn't my current existence a REALLY unlikely event? And if ~H is true, isn't my current existence actually likely (accepting for the moment my prior probabilities for my simple hypotheses)?

 

[Jabba]

What's more, one can write it out further so that we have equal priors and implicit evidence E' such that an application of Bayes' theorem to those equal priors with E' gives the priors we started with. Given that successive applications of Bayes' theorem are an iterated multiplication, and multiplication is commutative, this further means that we can just consider E along with the equal priors first, and have agreement on that first result. We don't have to start with the priors people disagree on, we can change the order of the evidences (implicit or not) however we want.

If people disagree on priors they could just start with equal priors and leave their individual, disagreed upon, information for afterwards.

Caveman,
- Sounds good to me.

 

[Jabba]

The partition itself is bizarre. I suspect it was invented as an attempt to satisfy the criticism that ~H is a composite hypothesis.

jt,
- Why is it bizarre?

 

[Jabba]

I don't really have a problem with made-up priors. Priors are just an expression of personal belief. There's no hard reason that any two people should agree on them; after all, we each bring different background information to any particular problem.

Furthermore, with a little algebra, we can write Bayes Theorem in terms of the odds of H, like this:

P(H|E) / P(~H|E) = [ P(E|H) / P(E|~H) ] × [ P(H) / P(~H) ] .

This shows that the posterior odds of H can be factored into the likelihood ratio and the prior odds, each factor contributing independently to the posterior odds. Since the data only affects the posterior odds through the likelihood ratio, even people who hold radically different subjective prior odds of a hypothesis should be able to agree on how much their odds should change given new data E. In this sense, Bayesian inference still "works" in the face of disagreements about the priors. But it requires H and ~H to make unambiguous predictions about the data. Thus, lack of consensus on the priors is tolerable, even expected. But lack of consensus on the likelihoods, which implies that the models (hypotheses) have not been adequately specified, is fatal.

jt,
- Have my models not been adequately specified for you?