Email to Elon Musk, re Safe Trips to Mars. What Would a Free Will Experiment Look Like ? Boolean Logic for Applications Beyond Digital Electronics ?

Some weekend thoughts.

(i)??????????????Wrote an email to Elon Musk with the previous blog article contents, i.e., safe trips to Mars, all probability and statistics empirical, with only local-to-Earth error bars. What value empirical input without error bars, etc. ?

Elon Musk does not have an available email address. In the end, I wrote to [email protected] and ElonMuskOffice@Tesla pointing out that I could not find a direct address. I also posted to a couple of Elon Musk company webpages, but drop-down titles were not appropriate, so no knowing where those web-mails will end up. Maybe one of those emails or messages will end up on an appropriate desk.

I'll see if I get a reply. Often when I email experts, I'll get a considered response back, to an interested question. Sometimes containing an insight that would be hard to extract from the literature or spare time reading. In this case, a reply from a Space-X engineer indicating how they tackle the safety critical systems problem would be very interesting. Reasonable assumptions always. But how 'reasonable' to be, in the most safety critical applications ?

(ii)?????????????Fatigue lifetime modelling employs the Weibull distribution, which is based on card from deck selection. And the point at which that frequentist argument is meant to contact the physical world looks totally invalid to me. Thus, another guessed ideal distribution, good to empirical accuracy. No explicit modelling of the science or the measurement process directly. And how important fatigue modelling for a trip to Mars ? Similarly, interpolation and extrapolation from limited data points, on Earth. So what to do ? Design always with large tolerances on every empirical input ? E.g., normal distribution (dangerous non-zero, fat tails); Weibull distribution; all probability and statistics, e.g., p ~ ? for the symmetric fair coin flip, not p = ?, etc. which is not justified even in ideal models. Otherwise, reasonable assumptions and is that good enough for safety critical systems ? At very least, make good estimates for all the numbers. How unconsidered would it be not even to do that ? So is it done as a matter of routine ? Don’t know. Would hope so.

And return to the free will question. The author thinks about that on and off.

Free will question, as per previous blog. Plus how ever to demonstrate lack of free will experimentally ? Could not assert lack of free will everywhere. That is unscientific. Would have to test every point of the Universe to attempt show that free will did exist, then fail in all cases. But what about demonstration of a single case, e.g., claimed mind programming, etc., conspiracy-theory-like, as a thought experiment.

Any experimental test system with a time delay, would be open to fraud. Any experimental test set up with a spatial distance, would be open to fraud. So how might an experiment be set up in demonstrably sound fashion ?

One definition of lack of free will might be fully predictable behaviour ? But that is a non sequitur.

Given determinism, and definition of free will as ‘I have free will if I think I have free will’ then determinism is irrelevant.

How about forward predictability as a definition ? Again, a non sequitur. This would say something about ability to take 4D spacetime photographs at a different locations in spacetime. Thus, not the same as determinism, more like precognition. So interesting if doable, but not a test of the free will argument.

Consider a crazy psych’ performing a free will test. If based only on external observations, then this is 100% subjective. Even if monitoring is ‘local to mind’, e.g., picking up mental verbalisation by near future engineering hi-tech, say, then this is still fully external to test subject’s self and mind, thus again 100% subjective.

Thus, the crazy psych’ might think she could demonstrate lack of free will in a test subject, but that would be entirely a matter of personal perspective.

And then the ‘if you think you are at the controls of my machine, how do you know I am not at the controls of your machine’ argument ??Thus, the crazy psych’ would have to demonstrate her own free will, as part of the experiment, rather than merely assuming it. Maybe somebody else could be at the controls of both machines.

And as soon as questions as big as free will are considered, then all the reasonable assumption, subjective reality variants must be tested, thus in extreme cases, movie plot type scenarios, such as: The Matrix, Total Recall, The Trumann Show, Star Trek Holodeck, life as computer simulation, etc.

Given reasonable assumptions, no time delay, and ignoring spatial separation, consider demonstration of mind programming, in a single test subject, puppetry, (the Silmarillion), by displaying side-by-side images of the monitored test subject in real time, versus a predictive model also running in real time. How else to verify the assertion of lack of free will in the test subject ?

Fraud would just fabricate or swap the two live streams.?So, assume the live streams were verified as good. What might this prove ? At most, precognition perhaps ?

And what meaning an assessment of likelihood for any such explanation, or for other putative explanations should they appear necessary, given the primary experience of free will, e.g., effects related to telepathy, or time travel say, as just two illustrations, without hard probability numbers ? All probability numbers are empirical, often with no meaningful error bars. The Universe simply is as it is, whatever current best science appears to imply. If it's not old science, it's new science, and it's always mixed science, which can be hard to disentangle. So stick close to the data and don't over-extrapolate. Thus related to the free will question, if I think I have free will, I have free will. And don't deny the obvious without good reason. Science not scientism, as discussed in my previous blogs. Scientism being denial or acceptance of anything upfront, without careful examination of evidence and due analysis of all possibilities. At minimum, pre-science in the form of reasoned thinking based on evidence, internal world and external world. And science requires building of testable input->output models, falsified systematically against experiment, to experimental accuracy, and statistical reproducibility, with reasonable assumptions. Until then, scientifically unexamined and / or undetermined.

So how would the relative likelihood of any putative explanation for any odd stuff, including any apparent results or implications of this free will thought experiment, be assessed meaningfully statistically ? We do not have access to the parent population of all conceivable weird and wonderful, odd stuff. Any assumed parent population would be a guess, perhaps verified usefully, to some limited degree of approximation, after the fact.?The usual statistical approach. And what applicable statistics for a single case study ?

Mind and self would appear to be always non accessible by any broadly system-response, external observation methodology (including internal to brain, e.g., based on brain I-V / EM or other measurable characteristics). Mind and self accessible only by interviewing the test subject, a broadly cognitive psychology type approach, and believing his answers. Any test subject declaring himself to be lacking in free will, likely to be considered mentally ill. Though I have no idea if this is at all a prevalent world view, and it would be one of many interesting perspectives to explore by systematic cognitive psychology interview to ascertain the broad spread of introspective mindscapes, something beyond MBTI character types. I do not know if such surveys are made routinely. I have not come across any such results in my brief recreational reading.

As argued above, any assertion of Universe-wide total lack of free will is unscientific. And on the above arguments, it is hard to envisage how demonstration of lack of free will might be achieved even for an individual case, if performed and analysed in demonstrably methodologically sound scientific fashion.

And the author does logic --?

What is the usefulness of Boolean logic ? Certainly, it is good for digital electronics. But what about deductive inference in real world applications ?

Boolean logic evaluates T or F. Not every statement needs to contain an equality, but consider.

T = T,

Bill is alive = T

Bill is human = T

Bill is alive = Bill is human not valid as inference either way. Thus, how to prove meaningful shuffling of meaningless symbols, as opposed to meaningless shuffling of meaningless symbols ? What reason to expect that local isomorphism between knowledge systems will be generated or preserved, by inferential deductive logic, in any domain of application, without defining terms up front ?

The above illustrates misuse of the term equality, in p / e / maw application.

p / e / maw = physical external mutually accessible world

Also, all statements, such as the above, will be relative and subjective. So, what value Boolean logic with T / F evaluation ? Something closer to fuzzy logic would look more appropriate to real world knowledge systems.

And even if using Boolean logic in its generic deductive T / F form, any statement substituted for a meaningless symbol, would be ambiguous.

Any term can only be defined up to the last implicit assumption. It is not possible to define terms absolutely unambiguously even for self. Thus T / F inference on always ambiguous statements in any real-world domain of application. Thus, good to reasonable assumptions.

The above argument being based on the notion of possible coupling to, or embedding within, a larger system. If that is always a possibility, then how to know that any knowledge system is complete, correct or fully accurate, in its domain of application ?

?

And the author exchanged some interesting Quora Q&A's with a helpful correspondent. My conclusion:

Truth functional and logically functional, look the wrong way round as definitions.

Truth tables are actually logic tables. Arguably, truth requires meaning. Thus, first order logic (?), with substitution of meanings for symbols, at which point the possibility of truth arises, but the technical definition is logically functional rather than truth functional.

And a final thought for the day, prompted by the various considerations of the free will argument, above.

From the security perspective, writing stuff down, or not writing stuff down, could be irrelevant if mentally verbalised thoughts were accessible to near future hi-tech. Thus, a key algorithmic security question might become broadly: what can be achieved assuming full question, full solution development and reasoning, and full algorithmic solution are all visible, with public domain key ? How to secure or verify communications ? Signatures, for a start. Also, time delay, order of message delivery, etc. Still thinking about that one. But more doable than zero, I think.

The author is not a very paranoid person. He posts his thoughts on LinkedIn. :-)