Artificial Intelligence and Human Immortality

Chat Topics: Artificial Intelligence and Human Immortality

Ben Goertzel, CEO of Biomind LLC and Novamente LLC and and former founder of Webmind Inc., joins ImmInst to discuss his thoughts on the possibility of human physical immortality propelled by Artificial Intelligence.

Chat Time: Sunday Mar 7, 2004 @ 6 PM Eastern
Chat Room: http://www.imminst.org/chat
or, Server: irc.lucifer.com - Port: 6667 - #immortal



Ben Goertzel, Ph.D., is CEO of the start-up Biomind LLC, a firm focused on analyzing biodata with advanced AI; and the leader of the Novamente project, which aims to construct an artificial general intelligence. From 1998-2000 he was CTO/founder of Webmind, an AI software company. From 1989-97 he held professorships in mathematics, computer science and psychology. He has authored 6 books and many research papers.

Bio: The Edge

http://www.goertzel.org

----



AGIRI -- the Artificial General Intelligence Research Institute -- is a small team of individuals committed to bucking the trend toward AI conservatism, and explicitly working toward the grand goal of true artificial general intelligence.

http://www.agiri.org/

--

Recommended Reading:



My DNA, a fabulously long chain of amino acids, a copy of which is contained in every one of my cells, contains a large percentage of the information required to produce me, Ben Goertzel.

This is an amazing thing, really.

Extract my DNA from any one of my cells, and feed it into a “human producing machine,” and out comes a clone of Ben Goertzel, lacking my knowledge and experience, but possessing all my physical and mental characteristics. Of course, we don’t have a human producing machine of this nature just yet, but the potential is there: DNA seems to encode most of the information required to produce a human being.

This is the glory and the romance underlying the Human Genome project, a huge initiative launched in 1990, which aims to chart the whole human genome, to map every single amino acid in the DNA of some sample of human beings. No one could doubt the excitement of this quest: It has a simplicity and grandeur similar to that of putting a man on the moon.

Once you get past the excitement and mystique and into the details, however, the Human Genome Project slowly begins to seem a little less tremendous. One realizes that the actual mapping of the genome is only a very small part of the task of understanding how people are made, and that, in fact, the design of the “human-producing machine” is a much bigger and more interesting job than the complete mapping of examples of the code that goes into the machine. In other words, embryology is probably a lot subtler than genetics, and in the end, much like putting a man on the moon, the Human Genome Project is a task whose scientific value is not quite equal to its cultural and psychological appeal.

More: http://www.agiri.org/path/chap10.htm

Suggested Reading:

Universal Mind Simulation as a Possible Path
to Stably Benevolent Superhuman AI

Ben Goertzel
January 26, 2004
--------------------------------------------------------------------------------
Acknowledgements. These wild & wacky ideas have benefited significantly from discussions with a number of individuals, including Eliezer Yudkowsky, Philip Sutton (a long-time advocate of making Novamente simulate things), Kevin Cramer, Izabela Lyon Freire, Lucio de Souza Coelho, Zarathustra, Zebulon and Scheherazade Goertzel, and the Reverend Chuan Kung Shakya. Of course, none of these fine folks actually agrees with all these ideas; so to the extent that they’re foolish, the fault is definitely mostly mine….
--------------------------------------------------------------------------------

For those of us who take seriously the notion that humans may soon create AI’s with massively superhuman intelligence, ethical dilemmas loom large.

The crux of the problem is simple: If we build an AI and it suddenly becomes 1000 times smarter than we are, why should this AI want to keep us humans around? Perhaps it will prefer to utilize the mass-energy that we occupy in some fashion more suited to its tastes.

Of course, there’s no specific reason to believe such a being would do nasty things to us. Perhaps it would disappear into another universe altogether … perhaps it would reach a state of peacefulness in which no inclination to absorb further mass-energy arises. Attempting to foresee past the Singularity at which technology escapes the grasp of the human mind, is a futile game. But nonetheless, it seems worthwhile to think about how to increase the odds that superhuman AI’s we create will be benevolent toward us and other beings that we care about. If it’s too hard to think about post-Singularity scenarios, at least we can think about how to make slightly-superhuman AI’s treat us nicely in the nearer-term future (the time period after superhuman AI’s exist, but before they launch the Singularity!).

More: http://www.goertzel....AllSeeingAI.htm

<Ben> buenas tardes
<Ben> Anyone else on here?
<jmgj1> hi ben
*** Joins: gustavo (~gustavo@pool-138-88-112-214.res.east.verizon.net)
<John_McC> Sorry Ben, I've been waiting for BJ to start the festivities
<Ben> No worries...
<John_McC> We have fewer people tonight since 6 pm is a little early for a immst chat.
<Ben> Yah -- he had to reschedule since I forgot the time I was previously scheduled for ;-)
<John_McC> Plus, there seems to be some sort of SIAI volunteer meeting on the SL4 channel at 7 pm.. dunno what that is about.
*** Joins: ChrisRovner (~cr@200.59.172.50)
*** Joins: Guest (~Guest@pcp08124930pcs.nrockv01.md.comcast.net)
<Ben> Yeah, SIAI just hired a new director... which will be great for them I guess...
<Ben> Anyway i don't mind if this chat is short ;>
<Ben> my kids are eager for me to get off the computer, I promised I'd watch Terminator 3 with them
<Ben> My 6 year old daughter is convinced it's the most likely future for humanity...
<Ben> hopefully she's wrong...
*** Joins: Eliezer (~sentience@c-24-98-32-87.atl.client2.attbi.com)
<Ben> Anyway I guess we can start chatting about pertinent topics now eh?
<Ben> I don't know what the format is for these things...
<John_McC> I'd love that... can you talk about AGIRI or the Novamente project?
<serenade> O dpm
<Ben> and I generally despise chat due to my fondness for sentences and paragraphs...
<serenade> oops, ignore that
<Ben> Sure ... there are 2 things I could talk about:
<Ben> 1) Novamente, 2) Biomind
<Ben> They're related of course but also distinct to an extent
*** Joins: outlawpoet (~Justin@24.130.29.164)
<Ben> Novamente is an AI system aimed at general intelligence
<Ben> but it's also a software system usable as an "AI toolkit" for building
<Ben> narrow AI applications --
<Ben> for instance bioinformatics applications
<Ben> Biomind is a company we founded to apply Noavmente to
<John_McC> As a project, how far along in a project life cycle is it? (percentagewise)
<Ben> bioinformatics ... mostly analyzing gene expression data
<Ben> About 50-60% of the Novamente design is implemented and tested
<Ben> However, this does not include GoalNodes and does not include
<Ben> procedure learning -- things that are totally critical for
<Ben> autonomous general intelligence, self-modification, etc.
<Ben> but are not necessary for our current apps in bioinformatics
<Ben> and text analysis etc.
<Ben> The Biomind Analyzer product uses only a few parts of Novamente
<John_McC> So you have functional subsystems or infrastructure done now?
<Ben> and has now been released and delivered to some customers
<Ben> Yah, we have functional subsystems, such as first-order
<Ben> probablistic inference, evolutionary learning of data patterns, etc.
<Ben> These are the stuff used in Biomind and our other apps
<Ben> Regarding human immortality, our most direct attack on that
<Ben> problem is via applying Biomind software to understand
<Ben> DNA repair and apoptosis pathways and so on...
<Ben> But of course achieving a true AGI via Novamente would indirectly
<Ben> contribute to human immortality --
<gustavo> sorry, what is apoptosis
<gustavo> ?
<Ben> altho I should stress that human immortality is by no means my sole goal!!
<Ben> apoptosis is pre-programmed cell death
<Ben> your cells are programmed to kill themselves
<Ben> If a cell is too screwed up, it's best for your body for it to kill itself
<gustavo> ah, ok
<gustavo> thnks
<Ben> only cells in multicellular organisms have apoptosis, it seems
<Ben> It's an interesting conclusion that it may take an AGI system
<Ben> to fully understand human biology...
<Ben> the breadth of bio data out there now seems to boggle human biologists' minds...
<Ben> I'm sure there are major discoveries implicit in the data that's online right now
<Ben> in bio research papers and databases... but humans can't integrate it all..
<Ben> it's not what our minds are made for...
<Ben> So... wanna hear about Novamente or about Biomind...
<Ben> or neither ;-)
<John_McC> Both, really... It's been pretty obvious for a while that biotech is going to need serious AI to get anywhere iteresting.
<Ben> Hmmm... ok i'll start w/biomind then
<gustavo> about Biomind, what part is still played by the human mind in the discovery process? (in addition to designing the AI system)
*** Joins: hkhenson (~hkhenson@CPE002078ca5697-CM.cpe.net.cable.rogers.com)
<Ben> Right now the role of biomind products is to aid scientists in making discoveries
<hkhenson> wassail
<Ben> we do not have anywhere near an autonomous AI bioscientist
<Ben> the AI can make hypotheses based on analysing quantitative
<Ben> and relational and textual data ... then biologists
<Ben> look at the hypotheses and make their own assessments,
<Ben> and design wet lab experiments to confirm/deny...
<Ben> In principle though the process could be automated fairly fully--
<Ben> there are robo bio labs out there, and there's no technological reason
<Ben> why the cycle can't be closed, and the software used to generate
<Ben> hypotheses then run experiments to test them
<Ben> As of now the software isn't savvy enough to design radical new kinds
<Ben> of experiments but it could certainly order up and control routine
<Ben> robo lab work to verify its hypotheses...
<Ben> That would be more a matter of putting lab assistants out of work,
<Ben> not putting creative scientists out of work...
<Ben> For now, I'd say, Biomind AI Engine (built on Novamente)
<Ben> has a different kind of intuition than humans -- complementary...
<Ben> It's not really general intelligence, it's very clever "narrow AI"
<Ben> built out of components of a system designed for general AI
<Ben> But when you look at the complexity of pathways involved in aging
<Ben> it seems clear that human intuition isn't gonna be enough...
<Ben> The data is not quite there yet for us to fully map out these pathways
<Ben> but I guess it will come within the next 5 years
<Ben> due to the very fast pace of innovation in experimental biology
<Ben> ???
<Ben> Hmmmmmm....
<Ben> Hey, perhaps we should try this again some other time...
<Ben> I'm starting to feel like I'm babbling to myself --
<hkhenson> that's ok
<Ben> except I usually find more original topics to babble to myself about ;-)
<hkhenson> I presume you were answering somebody's question
<Ben> Indeed I was answering someone's question, which was a general question
<Ben> "tell me about Biomind"
<Ben> But hey -- you wanted to ask me about Novamente... so this is a decent forum for that I guess
<Ben> ask away ;-)
<hkhenson> hmm.
<John_McC> Is there a plan to add the Goal nodes and procedural learning?
<hkhenson> what I fail to see is something that really gets a handle on the process of intelligence
<Ben> John: yeah of course our plan is to complete implementing the entire design,
<Ben> it's just a matter of proceeding step by step through the whole thing,
<hkhenson> I take it from what I read is that you have built something that is basicly a bunch of links
<Ben> and the parts of more use in our short-term narrow-AI projects have gotten prioritized
<Ben> Keith: our data structure consists of a weighted typed hypergraph
<Ben> But that is a data structure not a process
<hkhenson> been there, done that, xanadu
<gustavo> another basic question: what do you call textual evidence?
<Ben> Keith, a weighted typed hypergraph -- like an array of numbers --
<Ben> is a very general math structure that can be used in very many different ways
<Ben> Novamente has almost no relation to Xanadu...
<Ben> any more than two different software systems both using arrays
<Ben> are necessarily related...
<Ben> The key is in the dynamics of the hypergraph as it evolves --
<Ben> in the mechanisms for reflection and learning
<Ben> The "basic process of intelligence" as I see it is very simple:
<Ben> You have a system that recognizes patterns in its world and itself,
<Ben> and then determines actions based on these patterns
<Ben> (again using pattern recognition in observation of past actions)
<Ben> where actions are chosen to maximize system goals
<Ben> To do this given infinite computational resources is trivial,
<Ben> as Marcus Hutter showed nicely in his recent work, building on solomonoff's classic work
<Ben> To do this using limited computational resources is hard
<Ben> and requires making a lot of clever compromises
<Ben> the brain makes these compromises one way, Novamente does so another way
<hkhenson> as I put it, getting intelligent behavior out of a cockroach
<hkhenson> I think the brain does it with massive parallel computation.
<Ben> Yes, the brain clearly uses massively parallel computation
<Ben> but that is an implementation fact, not a feature of the
<Ben> "general nature of mind"
<Ben> Not all minds need use massively parallel hardware, just cuz the brain does
<hkhenson> well, in theory, you could impiment a mind on an apple II
<hkhenson> it would just run mighty slow
<Ben> And you'd need a big pile of floppy disks...
<hkhenson> right.
<hkhenson> heh.
<Ben> Yes, there is no rigorous math telling us how much hardware we need
<Ben> to make a Novamente as smart as a dog, chimp, human, whatever...
<hkhenson> might be worse than a year's stack of DVDs for the LAPD badge camera output
<Ben> On this point we are going based on our intuition based on playing with the system as it is now,
<hkhenson> 1600 foot high stack.
<Ben> and based on what we know of cog sci
<gustavo> is that a proved fact or a theoretical assumption? that there are no limitations for building a mind intrinsic to the hardware you are using
<hkhenson> partly afk for a bit
<Ben> Well, if you buy the Church-Turing thesis then as keith said it is possible to build a mind on any hardware
<gustavo> in other ways, that the mind is a digital computer, and that the dichotomy hardware / software applies to the mind
<Ben> I don't believe that "mind is a digital computer"
<Ben> I believe a mind is a system of interrelated patterns
<Ben> and that digital computers are capable of generating systems of interrelated patterns of arbitrary complexity
<jmgj1> a mind is a analog system.
<Ben> In my view minds are neither digital nor analog --
<Ben> minds are systems of patterns ...
<Ben> But these issues of philosophy are quite subtle...
<Ben> My view isn't quite the same as standard computationalism...
<Ben> I don't agree with Dennett and Eric Baum that experience doesn't exist, for one thing...
<jmgj1> a mind is the final comupution of all computationalism.
<Ben> ??
<gustavo> what about consciousness, intentionality, emotions, etc?
<gustavo> self awareness
<Ben> I wrote an essay on this topic recently, check out
<gustavo> great, do you have an URL?
<Ben> www.goertzel.org/dynapsyc/, it's called "patterns of awareness"
<gustavo> thanks
<Ben> I do believe selfawareness exists and isn't fully computational in nature
<Ben> but it's not biological in nature either
<Ben> and it may associate with computer programs as well as with brains or neutrino plasms or whatever
<Ben> (By selfawareness in the above, I really mean "qualia", "raw awareness")
<Ben> As for emotion, I think it's a particular kind of experience
<Ben> that comes out of having our rational minds largely yanked around
<Ben> by our reptilian and mammalian brains...
<Ben> AGI's may not experience emotions like we do, due to having
<Ben> greater control over all parts of their software & hardware...
<Ben> Just as mystics and gurus, as they gain more & more control over their minds,
<Ben> experience less of what we call "emotion"
<gustavo> we can delete the program then?
<Ben> eh?
<Ben> I'm not making a value judgment about emotions...
<BJKlein> yikes.. sorry i'm late.. thanks for starting Ben.
<Ben> just pointing out that they're not necessarily an intrinsic part of mind
<Ben> and not necessarily desirable
<gustavo> right, that 's what I meant by my question. thanks
<Ben> Hey BJ ... I gotta go at 7 due to a date with my daughter..
<BJKlein> k.. no problem
<Ben> Personally, gustavo, I enjoy my emotions a lot
<BJKlein> just a quick question for ya.
<Ben> But I might sacrifice them to gain superior mentality in other ways
<BJKlein> do you subscribe to the idea that death = oblivion?
<Ben> BJ, I think it's very very likely that my individual memories, emotions
<Ben> and cognitive biases will not survive the death of my body
<Ben> Of course, I don't have certain knowledge about this
<BJKlein> thus, the continuance via brain uploading is the only option for you?
<Ben> I think that the core of my awareness -- the "I am " at my center --
<Ben> lives outside of time and space anyway, and "survives" my body's
<Ben> death in a sense...
<Ben> but that's either a weak or strong form of immortality depending upon your attitude and mood!!
* BJKlein tilts head
<Ben> Anyway I definitely have an interest in causing my memories and cognitive biases and feelings
<Ben> continue to exist indefinitely....
<BJKlein> thus you have metaphysical belief in the "I am"
<Ben> Yep, I guess I do... though "Belief" seems a wrong word, I guess it'll do...
<BJKlein> right.. that's cool..
<BJKlein> very interesting actually..
<Ben> My ex-wife was a Zen Buddhist...
<BJKlein> do you subscribe to any religion..
<Ben> I appreciated a lot about that tradition,
<BJKlein> ahh.. ok.. yeah
<jmgj1> the Ben can you copy all the atoms of my brain.
<Ben> but I'm pretty rabidly anti-religious when it comes down to it
<BJKlein> right.. agree
<Ben> can I copy all the atoms of your brain? Not at the moment, no!
<Ben> Sorry if someone told you I had super powers..
<Ben> it's a common misconception ;-D
<John_McC> So Ben, do you see any path for creation of an AI that can imitate an existing human brain?
<BJKlein> heh
<Ben> Imitate a human brain: sure it'll be possible eventually
<Ben> My own Novamente work is oriented differently,
<John_McC> What about collecting the data, and building it later?
<Ben> toward making a fundamentally nonhuman AGI system
<Ben> but my Biomind work may well help out with molecular neuroscience
<Ben> and thus help work toward this goal
<Ben> As for collecting the data in the human brain,
<Ben> there's no good way to do that now
<Ben> but I'm sure the tech will come
<Ben> Eugen leitl has written some good stuff on this,
<hkhenson> Ben?
<Ben> but I don't recall the URL
<Ben> keith?
<hkhenson> are you up on william calvin?
<BJKlein> Ben, not trying to rehash to much.. but your feeling about the "I am".. does this preclude any major focus on protecting your existance (cryonics, etc)
<Ben> Yeah, I know the guy and have read his books
<Ben> BJ: I'm an Alcor member
<BJKlein> wonderful..
<BJKlein> same here
<hkhenson> ben, out of curiosity, what or who influenced you to become an alcor member?
<BJKlein> any luck with family members in this route (cryonics)?
<Ben> I guess Peter Voss pushed me over the edge into signing up
<Ben> but I had been meaning to cryonically preserve myself since early youth
<hkhenson> with me it was eric drexler
<Ben> when I found out it was possible
<Ben> About family: my kids badly want to be cryo-preserved but
<Ben> my ex-wife won't allow me to sign them up. I will when they turn
<Ben> 18.
<hkhenson> chance are good they will make it to 18
<Ben> My new partner is in the process of signing up though...
<BJKlein> Ben, we should reschedule another chat sometime soon.... many good questions for you...but don't want to keep you from your daughter
<Ben> See, my ex-wife believed in reincarnation! not compatible with Alcor...
<Ben> I can chat a few more minutes BJ... she seems to be distracted playing Harvest Moon
<MichaelA> Can you tell us a little bit about your new partner? bwuahaha
<Ben> Sorry that's classified information
<MichaelA> Awww
<BJKlein> thanks for adding me to your Orkut friends list, Ben.
* BJKlein is an Orkut addict now
<Ben> She's a Novamente programmer and a mean bluegrass bagpipe player though...
<hkhenson> and a redhead?
<gustavo> you still running for president ben?
<Ben> Well not in this upcoming election
<Ben> I need to get Novamente to the stage of being a powerful AGI first
<Ben> The only way an obvious freak like me could get elected President
<BJKlein> what's the financial hurdle for pres candidate (if any)?
<Ben> would be to become as famous as Einstein first ;-)
<Ben> It's not the financial hurdles that worry me -- it's the conservatism
<Ben> of the US public, that would preclude someone as odd as I am (from their perspective)
<Ben> from getting any votes
<BJKlein> heh.. yeah.. no kidding. i suspect creating an AI president would be easier
<jmgj1> this is for ben
<jmgj1> Stem Cells: From Blank to Brain
<jmgj1> By Kristen Philipkoski | Also by this reporter Page 1 of 1
<jmgj1> 10:35 AM Nov. 30, 2001 PT
<jmgj1> Two teams of researchers have turned human embryonic stem cells into brain
<jmgj1> cells, a significant scientific step that could lead to the treatment of nervous
*** Joins: jmgj1 (jmgj1@dialup-171.75.72.93.Dial1.Weehawken.Level3.net)
<Ben> yeah, that's very cool!
<Ben> stem cell research is definitely important
<Ben> Biomind is not working on that sort of data now but could well do in future
<gustavo> Ben, a basic question left from before
<BJKlein> Ben, you hang out with John Smart much?
<gustavo> what do you call "textual evidence"?
<Ben> BJ, I know John Smart pretty well and we trade emails back and forth regularly
<Ben> I don't hang out with him in person since I live in DC and he lives in california...
<BJKlein> ah k
<Ben> gustavo: "textual evidence' in what context?? what are you talking about?
<gustavo> you were talking at the beginning that biomind used 3 types of evidence
<Ben> ah, I mean information contained in bio research papers, eg. papers
<BJKlein> Ben, as the first ImmInst Book submission is past.. I hope you consider submitting something for us for the next book...
<Ben> referenced o nmedline etc.
<Ben> BJ, maybe I'll submit something for the next book, sure
<Ben> Lately my "writing time" has been devoted to completing the
<Ben> series of 3 books I'm writing about Novamente, plus some
<Ben> bioinformatics research papers...
<BJKlein> coolness
<Ben> Anyway i wasn't sure I had anything original to say about
<gustavo> about textual evidence: can the AI system be fed on journal papers? how do you feed textual evidence into an AI system?
<Ben> immortality...
<BJKlein> right.. understand
<Ben> Because we have a natural langauge processing capability
<Ben> Which works imperfectly right now...
<gustavo> ok
<Ben> Two modes: information extraction (fast, but misses a lot)
<Ben> Interactive -- requires a human to correct its mistakes... goes slowly but doens't make errors
<gustavo> Hello, computer, let me help you reading this
<Ben> anyway.. guess I gotta go...
<BJKlein> thanks Ben! have fun.
<gustavo> thanks! it was great
<Ben> ok, i guess we'll do this again sometime.. have a good night everyone...
<jmgj1> bye ben
<Ben> bye