Schneier on Security

Clive Robinson • September 13, 2022 3:32 AM

@ SpaceLifeForm, JonKnowsNothing, ALL,

Re : When is a spade not…

With regards,

There may have been a legal SCIF at MaL but there is no way it is legal today.

This has always been an issue of “certificates” and similar including identity documents like passports, and it’s all to often used for “rights stripping” by those in a position to control the certification[1].

So an example, I build a fire / earthquake shelter “to code” and it is then examined and “officialy” accepted as such. Then they change the code and over night the shelter is nolonger “officialy” accepted as such.

But is the shelter any less safe than it was?

Well it’s not changed so of course not[2]…

With regards the potential of a SKIF at the hole in the ocean, there will however be a “bureaucratic reason” for it no longer being certified… Like there was not “a certified guard” standing at the door scratching his butt 24×7, but then that is the nature of the beast…

Oh and don’t think that having it tested for compliance will work… There will be some clause somewhere that says some of the necessary tests are “secret” so… No independent testing possible, as they say “Not my first rodeo” on such nonsense.

[1] Look up what happened to Ed Snowden for example.

[2] You see the same but opposit problem with calibration certificates, one day a test instrument is allegadly tracable to national standards the day after it’s not… But in reality it was only tracable at the time it was tested not for the duration of the certificate. Because logically “entropy” tells you at some point it will fail in some way and it may not be immediately obvious, you do not know when, it could be five seconds after the calibration it might be five decades or more you just do not know (hence older wiser engineers do check tests before and after equipment use).

Clive Robinson • September 14, 2022 7:07 AM

@ SpaceLifeForm, ALL

Re : Things a’twitter.

“Media reports were already out early saying it was a foregone conclusion, yet the business day was not yet done. Stock price barely moved.”

Yes it was expected that the vote would be to get out with a profit.

Anyone who looks at a graph of Twitter share price csn see the Venture Capital “Pump and Dump” peaked out before the Musk offer. If you assume a straight line continuation the share price would be well well below where it currently is. Thus the real value of the Twitter share is probably below a third of what it currently is…

Thus the share holders all want to cash out like fleas hopping of a dying dogs carcus.

If the peding court case looks like it’s going to go Musk’s way expect the current share price to not just slump but tank like a lead kipper.

It might save time if Twitter changed it’s name to Toast…

The more information that comes out about how certain members of the Twitter board have shall we say been “fudging the figures” the lower the real share value will be. Beyond a certain point the “bull run effect” will happen and you would find the share certificates would probably not be worth hanging on a nail in an old style 1800’s out house…

Clive Robinson • September 16, 2022 4:04 PM

@ fib,

Re : Other sentience.

“… must [be] acquired through an evolutionary process, with the integral participation of sensors [image, acoustic, pressure, chemical] capturing reality in real time, and processing it without having to solve a specific problem.”

I’ve highlighted two parts because they are effectively contradictory.

All “evolutionary processes” are driven by the need “to solve” one or both of two very “sprcific problems),

1, Survival of the organism to breed
2, Survival of the species of the organism

Whilst they are often the same thing they can appear to be directly contradictory, such as when a female rodent eats her helples young, or a female insect eats her mate.

When looked at correctly they are “purposeful” under “evolution”.

The development of more than rudimentry skills is purposeful, thus it does fall within consoderation for evolution.

However with “sentience” a third eveloutionary pathway opens up,

3, Survival of environment

That is an organism can chose not to breed and in fact can chose to not keep their species alive if they will destroy the environment thus kill all life.

The question thus arises as to what the difference is between nonsentient and sentient life.

At the risk of being shot at, I would claim it is the maner of “free will” with regards “agency”.

Religion in particular is “anti-freewill” for the individual yet has to have organisational free will to exist and remain in existance in the face of changes in the environment.

It’s why I consider there are no deities just humans seeking control over other humans by use of a dictate / argument that is so one sided it leaves the power in just one or two hands, of people you realy do not want dictating your life (look at what we call “cults” to see why).

For those that doubt this look at the historical maturity of deities from worshiping that which is comprehended through to that which is effrctively philosophical musing on the notion of “good” but is in fact an evolutionary imperitive.

Thus I would argue that sentience in a spiecies can be found by looking first for societal grouping. Then by looking for signs of religion or moral evaluation as to “good or bad” not by participents in an act but by independent or third party observers. Further by a societal desire to “strive” toward a set of goals that are always on the edge of “reachable” thus as a species evolves and achives goals the goals either evolve or are replaced by other goals. So a societies generalised or sociatal goals evolve staying just ahead of where most in society would wish them to be (thus mores become norms and so on).

The important thing to understand is that at the base the physical environment is “bound” thus effrctively a “zero sum game” all that changes is how we increase any given resources utility by the application of energy as “work”.

There is unfortunately much nonsence talked about “increasing productivity by improving efficiency” well it takes no great brain to work out that with a fixed quantity of physical resources and that the efficiency of any process is ultimately limited, there is only so far you can go. And infact with a little further thought realise that beyond a certain point (around 67%) you start in on the “law of diminishing returns” as the processess cease to be robust to change and become increasingly fragile. It’s why we are now seeing the damage of over doing “efficiency” by the increase in supply chain issues.

As sentient beings we should be able to see when and where the 67% is worth passing and why, and when we should just jump to some alternative. But for some reason on mass humanity appears incapable of being sentient in this way…

So the, “Physician heal thyself” question arises about humanity… Which might account for our current inability to come up with a “theory of mind” or more esoteric, a “theory of sense” or “theory of inteligence” and importantly how they came about.

Whilst Roger Penrose claimed a “quantum theory” and since we have found biological quantum processes, we can not answer the “How?” question.

Clive Robinson • September 17, 2022 7:06 AM

@ JonKnowsNothing, fib, weather, All,

Re : Species of AI.

“Early AI endeavors attempted to replicate the method of “thinking” of a human.”

That still exists, and at one point was distinquished as “Hard AI” with the implication that other areas were “Soft AI” which at the time covered “Expert Systems” and later drew in “Fuzzy logic”.

Hard AI appears to get further away with every constructive step we take in that direction. Because what is realy being sort after is non human “creativity” that is ingenuity and invention capable of moving not just knowledge but the world forward in new and original ways.

Soft AI is realy about “codifing existing knowledge” and has no “creativity” even though it might appear to. Originally you would get an “Expert” to make the rules and these would be tested and further refined, hence “Expert Systems”.

The result was in effect no more than a flow chart decision maker with a fancy computer front end. But just as static as if the rules had been chisseled in stone.

Which brings us to,

“Then the definition changed to what we have now: AI == Pattern Recognition with a finite list of responses at the end-exit node.”

It was a bit more drawn out and nuanced than that.

The next stage was to make the human expert more “efficient” in rule making. In essence statistics on data sets instead of being used to test the rules, were used to find anomalies that needed addressing. As part of this non obvious patterns started to become seen.

Which appeared to some to be like magic, thus require abilities beyond theirs so therefor must have some form of intelligence… Which says rather more about their capabilities than those of the software. This process became known as “Knowledge Mining” a term still used today but has since had additions.

As I’ve mentioned before, back in the 1980’s I was involved with what was the first easily available and usable commercial system. The Company was called “Brainstorm Computer Solutions” and the product called “HULK” for “Help Uncover Latent Knowledge”.

I learn’t a lot from that, primarily why keeping out of Hard AI was advisable if you wanted a career[1] 😉

My input was to “Move the drinks cabinate”[2] by adding the use of “fuzzy logic” with randomized tree spanning to help match goals in more optimal ways. As it turned out even though it worked and worked very well it was not realy necessary.

Others investigated via “weighted drunkards walks” what were to become “Annealing Algorithms” to optimize desired outcomes.

These annealing algorithms showed quite significant returns, but not in AI but in manufacturing. They were ideal to minimise wastage on things like “glass cutting” in the making of double glazing and the like where standard sizes were not the norm and “cutting” had to be done in compleate straight lines such as “cleaving”.

The reality is AI, is still not “delivering” but some have discovered a “New Magic” for it. Most here will have heard,

“There are lies, damn lies, and then statistics”

Well the “New Magic” is to replace the word “statistics” with “Artificial Inteligence” or what ever the snake oil salesmen are calling it today.

I won’t go into the details of how you transfer a desired “bias” into such systems but as current research is showing it’s not exactly difficult due to “First look sensitivity”.

But one correction,

“Even animals know things fall DOWN and they don’t take courses in physics”

And they are wrong… As Sir Issac proved they move towards the respective center of mass in a vectored way. The reason “down” has a direction is we are already falling in that direction on a parabolic path.

As for,

“Horses can calculate speed, velocity, angle of attack and intersecting arcs and they never take courses in mathematics.”

That is because they use a varient of what was once called “fuzzy logic”.

“Fuzzy logic is based on the observation that people make decisions based on imprecise and non-numerical information.”

( From https://en.m.wikipedia.org/wiki/Fuzzy_logic )

Which drives a simple “split the difference and re-evaluate” huristic for two or more moving objects, which is kind of essential for “running eith the herd”. It is the same thing that gave USAF Colonel John Boyd the idea that gave rise to his “Observe Orient Decide Act”(OODA) loop huristic approach.

https://en.m.wikipedia.org/wiki/OODA_loop

[1] It was not long after that I got right out of the AI Game, and went back into product design of “Intrinsic safety” and “Safety Critical” systems where I unfortunately became “an expert” which I hated, so then went back into my “first love” telecommunications, oddly in FMCE product design. However the horror that most of the world thinks of as AI, that which underpins Soft AI still keeps coming back in my life in one way or another. For instance in the likes of “Interface Design” where the Interface learns from the users behaviours. At another level be used to optomise database organisation. Anything where adaptability is desirable will probably end up using the tools that Soft AI does, but without any kind of Inteligence, because it’s not needed when “a drunkards walk” gets you more that half way there, and adding even a little annealing gets you atleast eight tenths of the way there and often well over nine tenths…

[2] See “Black Adder” where the expression was coined as a sarcastic euphemism for an expensive but ultimately futile act.

Clive Robinson • September 17, 2022 9:38 AM

@ Winter, fib,

Re : AI and sensors

“eg, add emotions, sensors, and actuators. But no one knows how to do it.”

Not quite true as written. We know how to add sensors and actuators, and even via the likes of adaptive huristics we can get a machine to link some sensors and some actuators (preasure sensor and grip actuators) in a feedback loop such it grips but does not crush.

What we can not do, nore it appears have any clue how to do is get a machine to “want to play” just by it’s self without prompting / forcing / directing. Which when you think about it is how humans calibrate their huristics, and then go on and learn about things like hot, cold, etc.

As far as we can tell play is about “fun” which is a form of “pleasure” and we humans get rewarded for it by a release of chemicals that our brains “like”.

But those words fun, like, and pleasure, are just tags devoid of actual meaning they are just “names” for things we just do not understand.

Just like “random” and “determanistic” we can and do ascribe useful meaning that we can process to determanistic, but the only meaning we can apply to random is “appears non determanistic” and it’s not usefull.

To see why, come up with a “usefull meaning” for “joy” and why it is quantatively and measurably different to “happiness” and “pleasure” thus can be correctly recognised.

The only thing we do know is that “pleasure” in it’s broadest sense is a driver for evolutionary processes.

But we do not know which is cause and which is effect. That is does seeking pleasure cause evolutionary behaviour or does evolutionary behaviour cause pleasure. So far we say “Chicken or Egg”.

What we do know is that without regulation thinks become either obsesive or addictive depending on your point of view as an observer.

Which brings us around to the notion of “judgment” and “good and bad” we have absolutly no idea how we decide good from bad and we use “arm waving argument” to “talk it into the long grass”. Part of which is to talk about “goal setting” and “achiving”.

Which brings us back to a subset of “pleasure seeking” and how much pleasure is good and how much is bad, without any way of measuring it empirically.

Thus arguably “emotion” is based on a “feeling” that can be either good or bad but does have intensity. The purpose of which is apparently to keep an organism one a usefull evolutionary path…

But it’s not very usefull because the first question where it all falls appart is the obvious,

“How is the evolutionary path defined?”

Followed by,

“How was the emotion control mechanism arived at?”

We don’t know but we do know that it can be seen in people that are “driven” and they often have high IQ’s, and it can show at an early stage.

For instance I have a very fond memory of my son’s first Christmas. We had got him one of those “driving consoles” for 18-36 month olds, figuring he would learn in his own time. It was the “Fisher-Price” sort that has a steering wheel and accelerator lever, both of which change the sound it makes. Even though like most under one year olds he could not realy “sit up” properly after being shown how to play with it, he in effect glued himself to it and kept playing untill he slumped over it fast asleep… It was the first indication of just how focused he can get when something engages his attention.

The question arising is “Why?” when other children of the same age mostly don’t, their attention span can be measured in seconds or just a few minutes at most. The next question is “Does it have an evolutionary advantage?” and if so “Is it for the individual or the species?”