Re: Existential risks

Tim Tyler <seemysig@xxxxxxxxxxxxxx> wrote:
Curt Welch wrote:
Tim Tyler <seemysig@xxxxxxxxxxxxxx> wrote:

Humans are a crummy starting point for any future development -
so our descendants will most likely be more closely related to
today's machines than they are to us.

Well, really, you can look at DNA as the starting point, and humans are
just AI machines created by the DNA to help it survive. In turn we
will create other AI machines to continue that process of keeping the
DNA alive. The DNA is in control here, not us.

Unless something happens and the DNA fails to survive, our decedents
will continue to look just like they have for billions of years now -
societies of DNA strands surrounded by huge industrial complexes they
built which help keep them alive. Our work at creating AI is just
adding to that, it's not changing anything - the DNA is still in
control.

We are many orders of magnitude smarter than our DNA, but yet, here we
are, using our super intelligence to keep _it_ alive. If DNA could
create super intelilgence and make it continue to survive the purpose
of keeping the DNA alive, what makes you think the super intelligence
we create will do anything other than continue to serve that same
purpose?

The genetic medium for AIs are databases and computers - not DNA.

DNA is a pretty useless medium for them currently:
instead we have hard disks, CDs, DVDs, solid
state storage devices, etc.

The original problem that led to to biological systems
producing alternatives to DNA for storing information in
was the fact that DNA was not convenient as a writable
medium.

When organisms developed nervous systems they soon found
they needed temporary storage in order to maintain a
primitive model of the outside world.

However DNA did not wind up fulfilling this role -
and instead it was played by a complex collection
of enzymes acting to alter the activation potential of
synapses.

DNA is also a one dimensional storage medium. That makes it
easy to copy information stored in it, but leads to
problems if you want to sequentially access the
information or to index it.

Historically humans have preferred storage media with at
least two dimensions - since this reduces seek times
and provides better random access.

From a modern perspective, storage media can be classified
according to where they lie in a multi-dimensional space,
with axes labelled: cost, lifespan, thermal range,
error rate, size, access time, write capability, random access
capability - and so on. DNA lies at one point in this space.

However data storage is not a "one size fits all market" -
Different applications require different solutions -
and DNA is probably not optimal for any of them.

Rather it is an out-of-date solution that natural
selection managed to cobble together early on - and
which then became locked in. The idea of it persisting
for very long in the face of engineered alternatives
strikes me as being pretty fanciful.

Building on the human foundation is a ridiculous plan.

Yes, but it's not _our_ choice. We have no more choice than the AIs we
will build will have a choice on how they are motivated. We are here
only because the DNA lets us be here (so to say). We will survive only
as long as the DNA considers us worthy of taking care of them. They
have built us in a way to make it against our very nature to do
anything other than keep the DNA alive. We will build other AIs in the
same way - to make it against their very nature to do anything else
than keep the DNA alive. The forces of evolution are at work to make
sure it stays that way.

It is not DNA, but the *information* therein that is valuable.

One of the lessons of computer science is that information
storage medium is not very important. Information is
*fundamentally* portable.

However, a lot of the information discovered by biology will
be too primitive to last much longer. Who will need
photosynthesis, when we have nuclear fusion? What's the
point of neurotransmitters if brains can't be backed up.
A lot of existing adaptations /will/ be preserved - but
mostly in the history books.

Organisms were once based on RNA. Perhaps you would
have argued then that the genes of those organisms
have built them in such a way to make it against their
very nature to do anything other than keep the RNA alive.

Yet today, RNA has been replaced, almost entirely.

The medium is irrelevant. What matters is the messages.

Only if a life form more powerful than DNA based life emerges will the
DNA be evolved out of existence. DNA based life is "smart enough" to
know not to create things that will kill it, and DNA based life has all
of human society working to keep it alive.

So what are all the new heritable information storage media doing?
We invented floppy disks, hard disks, CDs, DVDs, etc because they
were better than DNA at information storage tasks. Gene sequences
are already stored in databases. We are not going to not use
a better database, just because it might compete with DNA. DNA
is expendible. What matters is the messages, not the medium.

All the technology we create, including AI technology, just creates a
larger protective shell around the DNA.

DNA sucks. It is totally replacable - and we will replace it.

The era of one-size-fits-all inheritance is long dead.

Augmenting humans would just wind up with their brains
being replaced by computers, and their bodies by robots -
not a significantly different result from a straight
machine takeover.

It's the same result. The DNA will fight to keep itself alive and not
be replaced.

Rather the *information* in the DNA will leap at the opportunity to
migrate into better media, to better ensure their immortality.

Any genes that fail to make the leap will be obliterated by those
that do.

Uploads may preserve some human minds, but you need an
extremely advanced AI to make that possible, and then,
the motivation to upload old personalities becomes pretty
weak. How many people play with Commodore PET emulators
these days - not many. Human minds will be even more
archaic and useless, by the time they can be uploaded.

These issues are just so hard to predict. I think you are right that
by the time we could upload a mind, we will have already figured out
how to build far more intelligent machines and the world will be full
of intelligent machines of all shapes and sizes and levels of
intelligence. By that point in time, a single human mind won't be very
important to preserve. It would be nice to preserve an Einstein or
Bach, but that's because they brains that did something most other
brains could not do. But by the time we could do that, all the great
minds of our times will be machines, not humans. So by the time we can
do it, society probably won't see it as something we should do.

Yes, exactly.

If they could have that behavior duplicated in a machine, would they
choose to do it? Calling it uploading is really deceptive. It's not
uploading as if we were being _transfered_ to the machine. It's just
building a machine which is able to produce a close copy of our
behavior and nothing more. If, after seeing AIs do it all the time,
humans should have a clear understanding that duplicating the behavior
doesn't allow _them_ to move into the other body - it just allows the
machine to mimic their behavior.

Allowing yourself to be copied, is what the option here is, not
transferring your soul into a machine. It might be neat to be able
dump the wiring of a brain, and archive it for all time - to allow
anyone in the future to run the brain emulator and talk to the person
that was once alive to learn what it was like "back in the day". That
would be cool.

I doubt most unmodified humans will be interested in uploading. Indeed,
I rather doubt that there will be many unmodified humans around by the
time that uploading becomes possible.

But will we as a society allow that emulated brain dump to continue to
have voting rights? To continue to consume resources and direct the
allocation of energy and raw material forever? I don't think we will.
Most likely, once we lose our DNA, we will just be put on the shelf and
archived - like a box of old photos and books about what did in life.

Ex-humans might be regarded as budding psychopaths. Uploads may need
brain surgery and "inhibitor safeguards" before they can join society.

And what happens if we become so advanced that we can stop death? What
happens when we figure out how to keep humans alive (at least some part
of them like the brain) forever? What happens if through a slow
process of genetic engineering that we learn to extend life
indefinitely? If this happens, and if we don't manage to escape the
earth into space and colonize other planets (or build huge planet sized
space stations from raw material we find in space), we will have to
regulate birth rates to prevent over population. If people don't die,
we can't let people reproduce. That will be an odd battle. Do we
prevent birth, or force people to be killed, or take away their right
to consume resources and allow them to live only as long as someone
younger who still have voting rights are willing to share their
resources with them? It's just so very different than what we have do
deal with now since we are powerless to keep people alive.

Death loses its sting once you can be backed up and copied. I doubt
biological aging will be convincingly fixed before the world fills
up with digital minds. So, the whole enterprise of fixing biological
aging mostly strikes me as a misguided waste of time: the route to
immortality is plainly a digital one.

However humans and society ends up evolving, I believe the DNA will
still be in control - at least for a very long time.

It can't be taken out of control until some other system of survival by
reproduction comes along to replace it.

That is happening now. Look at all the new heritable information storage
media that have established themselves recently.

Once we have smart enough AIs, they will be able to build other AIs
> and DNA based life will have some competition.

DNA already has serious competition from databases and the other new
media.

But we have all of human society working against it happening.

Human society is not on the side of DNA. DNA sucks - and is
plainly in serious need of replacing.

I don't see why humans will ever want the machines to take over
or how we will ever allow it to happen. It's like fire, and viruses,
and bugs that threaten to destroy our food sources. We see it all as a
serious threat to humans and we kill them without a second thought. We
will only build, and use AI, to the extent that we can trust it won't
take over.

Takeovers are common. The quill pen did not gradually
transform into a word processor by a series of gradual modifications.

Rather takeovers were involved: the old technology was replaced
by the new one.

We haven't had a genetic takeover for a long time, but they
/have/ happened before, and we are having one now:

``Cultural evolution is many orders of magnitude faster than
DNA-based evolution, which sets one even more to thinking
of the idea of 'takeover'. And if a new kind of replicator
takeover is beginning, it is conceivable that it will take
off so far as to leave its parent DNA (and its grandparent
clay if Cairns-Smith is right) far behind. If so, we may
be sure that computers will be in the van.''

- Chapter 6 of " The Blind Watchmaker", 1982.

``Today, billions of years later, another change is under
way in how information passes from generation to
generation. Humans evolved from organisms defined almost
totally by their organic genes. We now rely additionally
on a vast and rapidly growing corpus of cultural
information generated and stored outside our genes - in
our nervous systems, libraries, and, most recently,
computers.

Our culture still depends utterly on biological human
beings, but with each passing year our machines, a major
product of the culture, assume a greater role in its
maintenance and continued growth. Sooner or later our
machines will become knowledgeable enough to handle their
own maintenance, reproduction and self-improvement without
help. When this happens the new genetic takeover will be
complete.''

- "Mind Children", pages 3 and 4.


I agree with all these positions except the one major error you and these
other people are making.

You are making the mistake of thinking that _we_ (the brain / the AI) is in
control. We are not. We never were. We are stupid ass reinforcement
learning machines that are total slaves to the hardware in us which
determines our prime motivations. We don't control those motivations, and
as a result, we don't control what is going t happen to us.

And like I've argued before, if we do learn to control our motivations, we
will just turn ourselves into wireheads.

But for now, it's the DNA which is in control. It makes no different how
"dumb" or poor of an engineering choice it happens to be. It's the
dominate life form on the planet right now.

The "information" is NOT important. The "information" is not the hardware
building the brains, which is building the AIs. The DNA and its cell which
is the hardware building everything right now.

That last point is the one which sums up the error:

Sooner or later our
machines will become knowledgeable enough to handle their
own maintenance, reproduction and self-improvement without
help. When this happens the new genetic takeover will be
complete.

This is just totally wrong. It's based on the false assumption that
"intelligence" = "a desire to survive". It's based on the assumption that
since humans use our intelligence to do all this complex stuff to help us
survive, that all intelligence will naturally work the same way. I don't
believe that. I believe "intelilgence" = "reinforcement learning machine".
Reinforcement learning machines have no desire to survive. They only have
the innate desire to produce behaviors which will maximize their reward
signal. The behavior the intelligent machine ends up producing will be
whatever it takes to make the reward system produce more rewards.

Our DNA gives us a complex reward system that motivates us to produce
highly advanced survival behaviors. And evolution and natural selection is
hard at working making sure that continues to be what happens.

We will build AIs that are as smart as we are, and once they get that
smart, they will be able to design and build other AIs for us, and they
will be able to maintain themselves. But they will NOT be built to
survive. They will be built with reward systems that motivate them to help
us survive, just like the human brain is build with motivation systems to
help the DNA survive.

Once we build AI with the intelligence to self reproduce, and put them to
work designing and building more AIs for us, the technology will exist to
build replicators to replace DNA, and as Dawkins says, have the potential
to leave us far behind. But our DNA will fight to prevent that from
happening - and only when, and if, it loses that battle, will the other
form of replication when out and replace the DNA replicators.

Over the very long run, you can believe it just will find a way to happen
if the other forms of replication are better (which I'm not sure is true
BTW because DNA is about as small and efficient as you can get so making it
"better" seems to be hard to me). But in the shorter term, it won't happen
just because we can let it happen any more than man would let a small group
of aliens come in and take over without a fight.

It's not a question of who is "better", it's a question of who is in
control, and the DNA is currently in control (NOT HUMAN INTELLIGENCE), and
it isn't going to give up that control without a fight.

I suspect that the most likely way for DNA to die out, is simply by
attrition. That is, we will cease to live not because the machines take
control away from us, but because the more we learn about what we are, the
less we will care about reproducing and keeping the DNA alive, and the more
we will care about having "fun" while we are alive. When future humans
grow up in a world filled with intelligent machines, and they learn to see
humans as nothing more than poorly designed machines, a lot of the "magic"
of being human will vanish, and with it, the importance, and desire, to
make more humans, will start to fade. The more advanced the machines
become, the less "special" humans will become. At some point, we might
just turn over control to the machines because we don't care any more.

The forces of evolution will favor the genes that choose not to give up on
humans, but the DNA might not be able to evolve quick enough to change our
motivations in ways to force our intelligence to make us want to keep
surviving in this new world.

But if this path happens, it will happen over a long period - one which
does not turn over control the the AIs just because they develop the power
to self maintain and reproduce. Building new robots requires the work of a
large society (mining the material, creating all the parts, building the
robot) - where as building new humans requires only two humans and food.
Robots have a long way to go before they are are as powerful at survival
and self reproduction as humans are. We are amazing machines in ways that
go far beyond our intelligence.

Though we feel our intelligence is the source of our power, I think when we
better understand intelligence, we won't be very impressed by it at all.
We will see it as little more than a neat parlor trick - nothing but a
simple control process. It's just one more technology in a long list of
interesting technologies that exist in the human body. The really advanced
stuff is the nano-technology which exists in the cell which we aren't
anywhere near duplicating or replacing yet.

Humans who can self-reproduce with only the addition of food, have a huge
survival advantage over robots that require a whole working industry to
keep themselves alive and maintained and reproduce. Not until the AI
robots get as good at self reproduction (carrying with them all the
machinery needed to self reproduce from the raw material found laying
around), will they have a significant change of taking over from the DNA
based life.

The factories and industry we build will not be structured for the survival
of some type of AI, but instead, to create what humans need to survive. We
will use AI, but we will use it in all shapes and forms - most of it so
specialized it won't be able to understand what it is, let alone, care
about surviving. There won't be any single type of robot that will want to
take over. There will be millions of different types, all that tend to
need each other to keep functioning, and all of which were built to create
a huge support system for humans which together will have no purpose other
than to collectively meet the demands of humans. If we do die out, they
won't take over. They will just stop building things for us because there
will be no more humans around to ask them to build something. Unless of
course, the last humans alive ask the machines to design and build
"survival" AIs to take over.

--
Curt Welch http://CurtWelch.Com/
curt@xxxxxxxx http://NewsReader.Com/
.

Relevant Pages

  • Re: No easy road to AI
    ... DNA looks set to have a similar intermediate role in the future. ... mixed society of AIs and humans where the AIs have rights in the society ... world will be filled with machines smarter than humans in all ways, ... I don't think you grasp human nature at all. ...
    (comp.ai.philosophy)
  • Genetic proof for the descent of man
    ... Plagiarized Errors and Molecular Genetics ... Present-day humans and chimpanzees, ... DNA basics ... "pseudogenes" - might still be present in primates and guinea pigs as relics ...
    (talk.origins)
  • Genetic proof for the origin of man
    ... Plagiarized Errors and Molecular Genetics ... Present-day humans and chimpanzees, ... DNA basics ... "pseudogenes" - might still be present in primates and guinea pigs as relics ...
    (talk.origins)
  • Re: Genetic proof for the origin of man
    ... Plagiarized Errors and Molecular Genetics ... Present-day humans and chimpanzees, ... DNA basics ... "pseudogenes" - might still be present in primates and guinea pigs as relics ...
    (talk.origins)
  • Re: Existential risks
    ... just AI machines created by the DNA to help it survive. ... create other AI machines to continue that process of keeping the DNA alive. ... state storage devices, etc. ... Historically humans have preferred storage media with at ...
    (comp.ai.philosophy)