Re: How Robots Will Steal Your Job

[Roedy Green]

Does that make us not
intelligent? Or are we intelligent because we did manage to do it at
all?

All I have asserted is that there is a speed component to
intelligence. If entity A can come up with a solution faster than
entity B, it is more intelligent.

In this sense, computers are more intelligent than humans in solving
problems for which they have been programmed.

You can't really speak of intelligence without specifying the problem
or problem domain to be solved.

 

[Constantinople]

You need both. I think I would recognise a solution to global
problems if I saw one. However, I can't for the life of me think it
up. I don't have the compute power to generate enough alternatives
and evaluate them.

In some problems, the evaluation is the easy part: e.g. does this
transmission line design violate any of the safety design criteria?
How expensive would it be to build.

The hard part is fishing around in the sea of all possibilities
looking at the most promising possibilities since you don't have time
to consider them all. One technique for doing this is called dynamic
programming. I used it in Optow. We also used it managing water
levels in interconnected dams.

I think that the density of good solutions in all possible guesses (or
mutations) is key to how we even exist at all. If it took too many
mutations for an organism to hit on an advantageous one, too many
evolutionary "guesses", then evolution could not work; organisms would
degenerate back into random noise. The space of possible mutations has
to be to some degree "dense" with advantageous mutations.

It must be therefore to some degree "not too hard" to randomly come up
with an advantageous mutation. But if that's the case, this should apply
also to learning.

Moreover, we would be sunk if solutions hit upon early on did not tend
to last, did not tend to keep being useful beyond changes in
environment. If every little change in environment invalidated all our
adaptations, then life would be a fragile thing, like a computer program
that utterly breaks down if a single bit is misplaced. Clearly our
adaptations need to be robust. But this means that they need to be
general. I.e., we evolve in one environment, but the features we gain,
or the things we learn, remain applicable and useful, maybe with minor
adjustments, in new environments. This robustness is the same as
generality. Our lessons are general. We learn, or evolve, in specific
surroundings, but the specifics of our classroom (or primordial
environment) are not the only environments in which our lessons apply.
In short, the lessons we learn tend to be general. But that is certainly
not due to any special virtue on our part; the adaptations successful in
one environment simply happen, often, to be successful generally.

This means that not only is the space of possible mutations dense with
advantageous mutations, but the space of advantageous mutations for a
given environment is dense with advantageous mutations for more general
environments.

Or in other words, a given environment is in a way a model, a microcosm,
of the wider world. If one can learn lessons applicable to the wider
world merely by focusing on one particular, limited environment, then
that limited environment clearly is acting as a fairly useful model of
the wider world. This means that the world is filled with self-
similarity. It means that modelling tends to be successful, or tends to
be more successful than we might have any right to expect a priori. It
means that we can expect that by toying with models of the world we are
surprisingly likely to hit upon truths that apply to what we were
modelling. It means that modelling is curiously effective. Mathematics
is a kind of modelling. So it means that mathematics is curiously
effective.

Einstein asked,

"At this point an enigma present itself which in all
ages has agitated inquiring minds. How can it be that
mathematics, being after all a product of human thought
which is independent of experience, is so admirably
appropriate to the objects of reality? Is human reason,
then, without experience, merely by taking thought, able
to fathom the properties of real things?"

It's not just mathematics that admirably appropriate in helping us to
fathom the world outside of mathematics; but environments are,
generally, admirably appropriate in preparing organisms for the world
outside of those limited environments.

To be sure, an animal adjusted to a particular environment does not
necessarily do well outside of it. But when we consider the similarity
of animals that have adapted to all corners of the world, then it
suggests that while animals are not prepared for the wider world EXACTLY
as they are, nevertheless they're at a point where minor adjustments are
sufficient to launch them into a new habitat.

 

[Corey Murtagh]

Can I try it in C?

void magicTimeTravelUserInput()
{
do {
data_entry_form();
} while (!test_error());
}

Don't know enough Forth but I'd guess it goes something like

: magicTimeTravelUserInput while data_entry test_error do ;

Nah, it's something like:

for each entry field
push state
get input
if user requested undo
go to prior entry field
pop state

Fairly simple, depending on how you define 'state'. Assuming it's just
an entry form, 'state' is probably just a struct containing the output
data. Shouldn't take more than a few minutes to implement the core
logic for it anyway.

What would bug me is if there was an error in the 2nd of 20 items, and
all the rest were the same. From what I gather Abundance would, on
detecting the error, unwind everything you've done until it reached the
saved state where you were entering data in field 2, then force you to
re-enter everything again. That means I've just spent all this time
entering data, only to have the data entry system forget it all because
of some little mistake I made.

Sounds pretty stupid to me. I'd much rather it just tell me there's a
problem, set input focus to the problem field, and let me make just that
one change. Whether this happens immediately after I enter the data, or
when the system is validating all the data in the form, is irrelevant.

 

[R. Steve Walz]

Yet humans will submit to short term pain for long term gain, whereas
animals will not.

-----------------
But that's still not as potent a motivator as ignorant instinct.
We didn't win the planet because we suffered more, we won because
THEY did. We ATE them.

Example:
My cat hates wearing a collar. Yet the only way she is let outside is
with a collar. This is a long term relationship: collar = outside.

If she wants to go out to play, she sits by the door and waits until
someone notices. She does not get the collar, she does not sit by the
collar. In fact, she will sort of resist the collar being put on. Once
it IS on, she goes straight to the door and waits for it to be open
(proves that she knows about the relationship).

Humans undergo surgery to augment some part of their bodies. They KNOW
that post surgery pain will be present, yet they submit to it.

------------------
Not without anaesthesia.

The Dune series even had an "animal test" where a human will make
themselves tolerate pain, knowing that it IS a test and they must pass
it, whereas an animal will immediately remove their hand from the
pain.

 

[Corey Murtagh]

Isn't this simply a built-in "socialize with other cats" type of
behaviour?

I'm not certain she's aware that she's a cat. Her repertoire of
responses to other cats seems to be limited to spit, fight or run away.
Of course this is likely to be a territorial imperative, but she quite
definitely disdains social contact with other felines. Humans she
likes, especially if there's an available lap... mine for preference,
but any lap will do in a pinch.

Is she aiming for a warm lap or for being patted? If the latter is
involved, then it seems like normal (group) animal grooming behaviour
to try and "pat" you so that you will reciprocate and pat her.

The patting is one of her ways of gaining my attention. She's not
grooming me, she's trying to make sure I know she's waiting for
something. Similarly, playing with my hair involves a variety of
non-grooming activities... swatting at it, chewing the ends, etc. In
other words it's fairly basic attention-seeking behavior. The repeated
attempts to get over my shoulder, or under my arm, or attacking the
keyboard cable, etc. convince me that the goal is to get into my lap.

 

[R. Steve Walz]

I watched a dolphin practice for days until he could balance a ball
under his chin. I would say he had a purpose to his actions.

-----------------------
And he did this himself? Or was he being directed/rewarded?
if so, document this.

Your religion has given you

-----------
I have no such religion.

an advanced case of species vanity.

 

[R. Steve Walz]

The main thing that convinces us that humans are special is speech.
For example, nobody came up with a way a non-speaking animal could
convincingly demonstrate it had a notion of past and future.

----------------
They're supposed to come up with it themselves, WE did!

I worked in a research lab on a dolphin communication project.

Dolphins can communicate complex information to each other. You can
prove this by putting two dolphins in two different pools and giving
some information to one dolphin and seeing if the other dolphin knows
it. You connect the two by high bandwidth hydrophones. As you reduce
the bandwidth, the ability to communicate disappears.

What gets really baffling is listening in. Sometimes the total
conversation will be extremely short, and sounds the same for
different information transmitted.

If you study delphinic communication with an FFT, you can see it is
very rich. Lilly estimated they communicate ten times as many bits
per second as we do.

 

[R. Steve Walz]

Steve,

we're now well past the ability to use the off switch. The
computer systems that are running our economy today do have off
switches, but people are effectively prevented from using them.

If you go around and switch off computers now, then one of two
things will happen.

1. You will end up in jail.

2. You will end up in a psychiatrical institution.

 

[R. Steve Walz]

Steve,

three counter-arguments.

1. I would respect an animal even if it is not quite as far on
the path to intelligence as we humans are.

------------------
I would respect it and use steak sauce.

I would respect an animal if it has a concept of self.

-----------------
That's only apes and MAYBE cetaceans.

(The mirror test comes to
mind. Can the animal grasp that the picture in the mirror is
itself?)

-----------------
Questionable, but I'll allow it.

2. We cannot always be sure whether the animal speaks to us or
could speak to us, because we have not sufficiently studied its
language.

----------------
That's not up to us, we developed speech BECAUSE we were intelligent!
They must do so also.

I would not wait for proof, but instead give the
animal the benefit of the doubt.

 

[R. Steve Walz]

Roedy,

let me propose a definition. If a being has a model of the
world, at least of its surroundings, and if in this model the
being itself is included, then we call this conscience (or
self-awareness, which, to me, is about the same thing).

Hans-Georg

---------------
Models are possible without a self. They are needed for perception.
I'm more interested with whether they have a Self to observe and
use this model and believe they exist, or whether the model is
merely a structure that runs on its own, like a squirrel that knows
where the trees are, where the ground is, and buries nuts and forgets
them. Without a Self like ours to use that memory, it isn't a Alive.

-Steve

 

[R. Steve Walz]

If consciousness is a quantum phenomenon,

 

[R. Steve Walz]

Well, the account in Genesis suggests we were created with an additional
purpose over animals of having a relationship with God.

--------------------
It's not an "account", it's a Babylonian fairy tale told to children.

Hmm. Interesting assertion, but you'll need to do more than just state
an opinion as if it were fact. Can you communicate with animals?

------------------
We're in charge and kill them. If they were intelligence they would:
1) resent that, 2) try to convince us not to eat them. They don't.

If
not, how can you know anything about what they think? ----------------
Observation.


How do you know
they don't know they are here?

---------------
If *I* knew I was here I'd do everything in my power to prove it to
the chief species and ensure they wouldn't eat me1

What does it mean to know one is in a
particular place anyway - do you recognise your surroundings and from
that determine your location, which is possible, given that when we
don't recognise our surroundings we consider ourselves lost. Or do you
mean we know we exist?

----------------
That knowledge must be in the context of a total knowledge of our own
existence such as we have.

Assuming you could communicate with an animal, and that it fully
understood the question, exactly how would it answer "Do you exist?"

------------------
It need not know the word yes, but:
It would have to be obviously frantic to inform us, knowing what we
might do since we are in control, and then make every effort to
symbolically communicate. If it can't, then it doesn't exist. Symbols
are part and parcel of what we mean by awareness. In other words it
better take up a stick and start scratching in the dirt and making
sense.

If it understands the question, then it can only answer "Yes."

What if you were to ask a Frenchman if he existed? If he doesn't speak
English, the chances are he'd say something like "Quoi? Je ne comprend
pas. Où est le thé?" So are French people who don't speak English
self-aware or not?

-----------------------
Disingenuous, they are using symbolic methods.

If they are, then non-communication with animals doesn't automatically
imply a lack of self-awareness.

What if animals really are not self-aware? How would we determine this?

-------------------
As I said.

If we can't draw conclusions from a lack of ability to communicate in
a manner we understand, then what?

-------------------
We can. What awareness *IS* is symbolically communicative and desperate
to prove it's value so that it not be killed.

Another H2G2 quote seems appropriate here. "Humans think they are more
intelligent than dolphins because they have created the wheel, wars, New
York etc, while all the dolphins have done is muck about in the water
having a good time. Curiously, dolphins think they are more intelligent
than humans for exactly the same reasons."

-----------------------
I have no such prejudice about productivity.

Self-awareness of computers is another interesting question. Are they?
If so, are we committing ratiocinicide when we turn them off? How
will we know they are self-aware until they can tell us?

--------------------
No entity is a being and has a right to life until and unless they
are bright enough to demand those rights in a manner that any other
such being can understand. That means symbolically.

They AREN'T till they make the effort! Nothing IS alive or DESERVES
the assumption that it IS alive, until it can demand that it be
respected! Until they can do that, they are FOOD!!

 

[Corey Murtagh]

And can you /currently/ demonstrate that to the satisfaction of any
non-human? If not, your objection isn't really relevant, is it.

 

[Roedy Green]

Fairly simple, depending on how you define 'state'. Assuming it's just
an entry form, 'state' is probably just a struct containing the output
data.

The implementation is remarkably simple. State is the contents of the
two stacks, plus some system variables.

However, it lets you write assertions and it forces the USER to
navigate around them if they fail by changing the inputs.

So for example you can insist that the area code and state be
consistent. The assertion jaunts back enabling the user to change
either to make them consistent. The place in Jaunts back to need not
be in the same routine.

You as programmer just concern yourself with writing assertions and
warnings (optional assertions). You don't have to write any code to
arrange navigation backward in time, not backward lexically as you
would in a loop.

This sort of thing would be extremely difficult to simulate in Java.

 

[Roedy Green]

And he did this himself? Or was he being directed/rewarded?
if so, document this.

No he was not being rewarded. He just did it on his own. The
dolphin's name was Joe. He was an adolescent male dolphin. I watched
this with my own two eyes.

I will be accused of anthropomophising, but he appeared delighted when
he finally mastered it.

I also watched his first encounter with an underwater TV monitor, and
also what happened when we put a live feed of an underwater camera on
the video.

I was also present when dolphins and man exchanged their first
"communications". I talk about it in an essay at
http://mindprod.com/intel.html

 

[Roedy Green]

The space of possible mutations has
to be to some degree "dense" with advantageous mutations.

Another thought is life in the primordial oceans would have been a lot
easier without the predators. Life is a competitive game. You just
have to stay equal with the competition to stay in the game.

The competition has gradually heated up.

You can see the same thing happening in sports as we evolve better and
better training methods. You only have to compete against your
contemporaries.


In other words, in the game of evolution, you don't have to find the
best solution right off the bat, just one good enough to let you stay
in the game.

 

[R. Steve Walz]

And can you /currently/ demonstrate that to the satisfaction of any
non-human?

 

[R. Steve Walz]

What are you saying:

1. All things that are alive have a Self (a soul, a thing inside that
feels and experiences)

 

[Corey Murtagh]

The implementation is remarkably simple. State is the contents of the
two stacks, plus some system variables.

Yes, simple.

However, it lets you write assertions and it forces the USER to
navigate around them if they fail by changing the inputs.

....where most modern systems are written to work around user error in
one way or another.

So for example you can insist that the area code and state be
consistent. The assertion jaunts back enabling the user to change
either to make them consistent. The place in Jaunts back to need not
be in the same routine.

My objection being: if you unwind the system state to an earlier point,
before the user had entered some piece of data, then the user is forced
to re-enter that data, whether it was correct or not. Using your
example, if the user entered the wrong State Name (I presume that's what
you meant by 'state' in the quote above) but the correct area code, then
wanted to change the State Name, then he would have to re-enter the area
code, even though it was correct the first time 'round. Whether the
assertions caused the unwind, or it was manual, is not relevant.

You as programmer just concern yourself with writing assertions and
warnings (optional assertions). You don't have to write any code to
arrange navigation backward in time, not backward lexically as you
would in a loop.

This sort of thing would be extremely difficult to simulate in Java.

It's really not as difficult as you seem to think. I could implement a
form description language and intpreter for same in C++ that does what
you have implemented in Forth. It would probably be a more difficult
task, since I wouldn't have Forth's self-redefining ability to rely on,
but once the interpreter is written there's nothing to stop me from
keeping a list of snapshots of the interpreter state and returning to
one of those on demand. Doing it in Java should be about equivalent to
doing it in C++... possibly more difficult, probably just different in
detail.

Forth has a certain versatility, but you pay for that in various ways -
like lack of type safety, etc. As an interpreted language it's quite
fast. But I wouldn't consider it to be more capable than either C++ or
Java.

 

[Corey Murtagh]

I didn't say aliens, nor are they required to satisfy my condition.

Your assertion appears to be: Since no other lifeform has independently
demonstrated /to us/ that it has the ability to conceptualize
past/present, therefore no other lifeform is 'conscious.'

My counter-argument is: Since we have thus far failed to prove exactly
the same thing to any other lifeform, despite our vaunted intelligence,
your assertion is not relevant.

I don't see that this follows. If your assertion is irrelevant, how
does that equate to extinction? All I see is that /you/ failed to
either a) express your point in a way that makes sense /to me/, or b)
make a /relevant/ point.