19 replies to this topic

[Bruce Klein]

Nov 5, 2005 - ImmInst Atlanta Life Extension Conf.
http://www.imminst.org/conference

Individual Speaker Abstract & Discussion Forum:


Ben Goertzel, Ph.D.
CEO of Biomind LLC and Novamente LLC and and former founder of Webmind Inc., Ben discusses from the perspective of a unified theory of mind the "Immortalizing of Brains, Mind and Selves" by way of AI, neuromodification and gradual uploading - ImmInst Chat

[bgoertzel]

In this primitive era where undesired death is rampant and "basic" things like cryonics and stem cell research are under attack by the conservative and ignorant, it's all too easy to ignore the subtleties involved in futurist concepts. For instance, "immortality" itself is a subtle and multifaceted idea. Indefinite preservation of human bodies and brains is a worthy goal, yet represents only one dimension of immortality. Most likely, the possibility will soon exist to preserve human minds separate from human bodies via uploading into alternate software or hardware substrates, and/or to transform human mind-brains into fundamentally different entities via radical neuromodification and the hybridization of human brains/bodies with various engineered technologies. These possibilities raise fascinating philosophical questions regarding what "immortality" really means -- if I alter myself into a completely different, superhuman mind, then in what sense is the core of "me" being made immortal, as opposed to the current "me" being used as the seed for the growth of some completely different being? What sort of spiritual and aesthetic value lies in the growth process by which a mind transcends into a wholly different form? Is it possible to develop the technology that permits this kind of transcendence, while still preserving the freedom for some humans to remain in immortalized traditional-human form if they so choose (and for some humans to die if they so wish)? The questions abound, and there are no ready answers, but it's an appropriate time to be deeply reflecting on such things with all our minds and hearts.

In my talk I will review these issues from the overlapping perspectives of the multiple roles that I now hold -- theorist in cognitive science and philosophy of mind, artificial intelligence researcher, biological scientist studying life extension among other phenomena, creative writer/musician inspired by transhuman issues, and last but not least, human being.

[Bruce Klein]

Dr. Ben Goertzel's power point presentation:
http://www.imminst.o...ference/Ben.ppt

[Bruce Klein]

Dr. Ben Goertzel's video presentation may be found here:
http://video.google....505614870506496

[Athanasios]

Some people spend their whole lives to obtain ego-death. As long as I am part of the universe experiencing itself, I am not in oblivion, I would be cool with it.

Edit: An exception, I dont think I would choose it over myself if I could not experience joy that isnt pleasure. It is one of the most noble and remarkable things about being human. I do not think this would be affected by many changes, but obviously it would be affected by total conversion without careful consideration of this specific apect..

Edited by cnorwood19, 14 September 2006 - 02:57 AM.

[olaf.larsson]

Most likely, the possibility will soon exist to preserve human minds separate from human bodies via uploading into alternate software or hardware substrates

What is "soon"?? Minimum 50 years i guess.

[Lazarus Long]

In the grand scheme of things 50 years probably should be considered *soon*.

But of course most of us at least consider soon as in during *my* lifetime remaining. Because obviously otherwise it wouldn't be a moment too soon.

Nonetheless I suspect that approach could yield significant results a lot sooner than 50 years as it requires advances that can be seen occurring in multiple disciplines. This approach of course is not mutually exclusive with biological ones. Uploading in principle should not be understood as requiring *active uploading* with the ability to interact in a fully conscious manner while uploaded. It is only logical that we can *record and store* a person before we can provide that uploaded entity either the option of being down loaded into a body whether that body is biological, cybernetic or even a synthetic construct that exists only in a virtual reality.

That stage would be analogous to cryo except without the requirement of any specific body or brain. Perhaps DNA from tissue samples would accompany the uploaded data so as to insure the biological construction of the original body but can be understood as a viable option much sooner than 50 years and more likely within 20 IMHO.

Like so many of the technologies we discuss it would be wise to recognize there are phases of introduction that reflect degrees of complexity.

Uploading is a parallel tech to cryo bio and nano. It is an option for someone that is leading an active lifestyle which might make it difficult or impossible to ensure their body would be preserved through cryo. It is also the cruel fact of the matter that folks in my age bracket are not going to live long enough through biological means alone to realize many of the advantages of the biological only approach.

Edited by Lazarus Long, 25 September 2006 - 12:31 AM.

[caston]

Could a mouse brain one day be uploaded to a computing cluster based simulation?

[Lazarus Long]

This is an important step to accomplish before attempting a human brain of significantly greater complexity but I would argue that we must start even more basically at say insect memory and see what kind of learned behavior can be codified, recorded, and transmitted. I suspect there are such projects already underway.

Also just to be clear I was describing something that did not require a *simulation*, only the ability to encrypt sufficient information and store it. Interactivity is a far more complex problem than just *recording* the individual's total neural informational content up to that given moment.

If a sufficiently in-depth process of scanning and BCI could resolve a clear recording (a snapshot) of long and short term memory as well as neural net organization then this information is far easier and more reliable to store for longer periods than any other we can imagine perhaps including cryo. Because like cryo much of it depends on the ability of an infrastructure to support the information but theoretically a device capable of storing such a vast quantity of information could be created that placed the person in stasis even after the fall of the supporting civilization. Consider it *data crystallization* and a complex analog to bacterial sporification that can allow bacterial DNA to survive hundreds of millions of years.

This could allow for the discovery of such a device at some time in the future by a new civilization of sufficient advancement to decrypt the data and download the individual. It could mean that only memories are read in a kind of ROM or even that the intelligent species of humans could be recovered by a sufficiently benevolent future alien civilization if we don't make it. Although they would be wise to carefully consider such a move. )

Just as we should be if we ever discovered such a device.

However if we did discover such a device even today we probably would not recognize it for what it was and just cut it up to make jewelry. [lol]

[caston]

What is the simplest known biological mind?

Flat worms perhaps?

Edited by caston, 24 September 2006 - 08:09 AM.

[Lazarus Long]

I would keep working with mice as they at least have an organized neural net that is more analogous to ourselves, as well as clearly developed methodologies for testing memory and behavior.

However I am open to applying this strategy to more basic species but I would argue that it becomes too tenuous to test the behavior of species such as flatworms or even bacteria for complex behaviors although there are experiments involving the transfer of memory as RNA in these species that appear to offer some direction and positive result.

I suggest not attempting an experiment with any species less complicated than an octopus or shark. Octopi have demonstrated complex learning ability and individualized behaviors. They also have simpler brain and nervous systems. They reproduce faster and easier than sharks, not to mention being far easier to handle and maintain under laboratory conditions.

[caston]

Lazarus,

Agreed. I can't remember the source but I remember reading something about a guy with a pet octopus. The octopus one day opened the lid on its tank and helped itself to one of his other pets. They are very smart creatures.

As for simulations have you heard about the zero-one problem?

http://ieeexplore.ie...arnumber=151480

Yearning for the day when we can back-up and restore the soul...

[Brainbox]

In this thread, Bruce tries to explain the architecture of the AGI that Novamente is trying to develop as a prototype. This architecture is based on a black-box representation of human behaviour. I.e. a more or less functional representation of our behaviour, that is not necessarily a model of our brain. Also the video attached to this thread of the presentation of Dr. Ben Goertzel suggests the same, if I remember and understand it correctly.

Reading the introduction of this paper posted by Zoolander, I came up with the following dilemma as far as up- or down- loading of minds is concerned. Maybe (or even probably) this is not an original thought, I didn't have the time yet to read to much about AGI's and the way the concept of mind up- and down- loading could be implemented. It is also quite similar to the observation Lazarus made in this thread already. But stay with me please.

As far as I understand, the human or animal mind does not function as a computer, at least not as the von Neuman model we are using today. The essence of this in this context is that our minds do not work like a general purpose processor with general purpose memory, in which a piece of software can be loaded that can in principle carry out any function that receives inputs and produces outputs that are within the scope of the I/O functions that are available. As far as I understand, in our human mind, there is no separation between "software" and "hardware". The software is the hardware and vice-versa. It is the specific configuration of synapses that is grown according to genetic factors and experience. So, if we were able to model and implement the functions of the different building blocks, i.e. the synaptic network components, and "wire" instances of these components into a network equal to a particular mind, the "upload" has been carried out. It could be as "simple" as that. Simulate the different synaptic functions, copy and connect them according to the network of a particular brain, and viola. Ofcource this is an utter over simplified simplification.

The advantage, as far as I understand, of such an approach is that you just have to rebuild the synaptic network using computer models of the basic synaptic functions to recreate all the aspects of the original brain. This could almost be a one-to-one copy, provided we had sufficient insight into all the possible synaptic functions and their connections. And provided that there are not to many variations in synaptic functions. And provided we could implement a model of all of these synaptic functions, i.e. to simulate each of them.

This seems to me a far more manageable approach compared to making a model of our behaviour, with the countless parameters that should be measured and copied into the behaviour model. This is prone to misinterpretation and plain errors. E.g. I consider myself (and for that each other individual) as quite unique. By making a functional behavioural model, how can we be sure that it can ever contain the behavioural parameters of each individual? If we could recreate the synaptic network of a brain in a (software) model, we would automagically recreate all behavioural aspects of it.

???

Edited by brainbox, 24 September 2006 - 09:50 PM.

[Lazarus Long]

In a followup to the point about mouse brains here is an example of true serendipity.

Institute unveils full atlas of mouse brain

Map could point to new treatments for human neurological disorders

SEATTLE - A brain institute funded by software billionaire Paul Allen says it has completed its first project: a map of the mouse brain down to details of individual cells. Work is already beginning on a similar map of the human brain.

The new Allen Brain Atlas is being made available online without cost to neuroscientists studying brain circuits and chemistry, a potential boon to cancer and other disease research because of similarities between the brains of mice and human beings. The formal announcement of the mouse brain atlas' completion was made by the Allen Institute for Brain Science on Tuesday in Washington, D.C.

The tools just get better and better everyday to accomplish this objective. The results of this study are also another reason to continue with mouse experiments.

A question to all my friends at the MMP;

If it could be shown that a mouse brain with all the learning that we can identify for that individual mouse were successfully uploaded and then downloaded at some point in the future, say three or four years later into a cloned mouse at an early neonatal phase would that constitute a successful example of a Methuselah Mouse?

[lunarsolarpower]

If it could be shown that a mouse brain with all the learning that we can identify for that individual mouse were successfully uploaded and then downloaded at some point in the future, say three or four years later into a cloned mouse at an early neonatal phase would that constitute a successful example of a Methuselah Mouse?

Considering the furor such an experiment would generate if humans were ever involved, such a project would not be of serious interest to immortalists in my opinion. If an entire brain could be grown to the specifications required without the need for a donor neonate, then it would be exciting.

[Lazarus Long]

The issue is not about the furor it would cause for society, it would not be a viable option just because it could be done with a far simpler mouse brain. The only serious question is whether the prolonged life of one *individual* mouse is preserved by this method.

If you so no then demonstrate why any of the rebuilt body scenarios are valid that cryo depends on. If you are hinging the argument on cloning then that is also a logically false dependence. The model is dependent on whether or not the *consciousness* of the mouse is first off being preserved by this method and second that the non-organized neural net of the neonatal mouse can be externally organized to recreate the original mouse.

If the DNA, the personality and the memory of the original mouse are intact then wouldn't the age of the mouse be the combined age of the two bodies plus the period in encrypted suspension?

If you want we can even save the mDNA of the original mouse and include that in the clone. The point is to test whether or not a consciousness can be uploaded at all or is that something that you think also will be objected to by society?

If so too bad that one is not an ethical conundrum like cloning a human. I am not proposing cloning a human I am proposing cloning a mouse, something we already know very well how to do and do often. The procedure is not dependent on cloning, the proof for the MMP might require the subject to have the same DNA to meet the critical parameters of the contest.

[Lazarus Long]

Whilst an excellent tool for people interested in what genes are being expressed across the entire mouse brain in a gene-specific searchable 3D-context it does not provide information on the physical organization of neural pathways and their networks.

True but it is a start, It is also mapped down to a single cellular level according to the site. Combine this data with other active mapping techniques and it may become possible to also define the individual neural pathways and their networks using this as a standard. To accomplish what I suggested would require also attaching a BCI interface to the mouse brain and the ability to decrypt and record neural activity. We are still a ways away from this approach but we just got a little closer.

I also suggested using a cloned mouse so that it would not require as much to recreate the basic neural pathways but rely instead on structural genetics to accomplish that phase.

I have just heard a lot of theoretical discussion here about what is consciousness but the thought occurred to me that there might be a practical test of what is consciousness. And it is a test that offers a pragmatic approach to more than just a description.

[Brainbox]

I understand that downloading a brain by copying the neural structure and synaptic components into a simulated model would not be complete. It is to static since it indeed misses the information that is needed to revive the downloaded information in a way that it fully represents the original brain again. Furthermore, it is unknown at this moment how we could scan a brain structure with sufficient accuracy to be able to copy it.

But as far as uploading the consciousness of a living being, Imho this is the way to proceed instead of making a behavioural and learning model. Although copying a brain has it's issues, copying a mind as a functional model (opposed to the physical model of a brain) will always confront us with limitations regarding the correctness, completeness and last but not least incorrect human interpretations of the mind-model.

Although copying a physical brain will not be easy, it could lead to something that is complete without the increased risk of missing parameters and components in the functional mind model.

[Brainbox]

In a followup to the point about mouse brains here is an example of true serendipity.

As far as I understand it will be a generic model?