The only way to bridge the gap between the Singularitarian vision of a radically different future and the shared reality we live in now is to bridge the gap between NOW and THEN.

A lot of AGI-related nontechnical thought revolves around imagining a sharp chain reaction where a superintelligent machine pops into existence and everything goes haywire. We have to find out what the intermediate steps are, and if we really understand general intelligence well enough to build it I would hope we can say something about those steps.

What business opportunities are available for people with machines that have a “little bit of general intelligence”? How will “somewhat generally intelligent” machines make incremental changes to everday life? What early steps can demonstrate that actual progress is being made?

If it is impossible, if AGI actually is an all-or-nothing thing, then there isn’t anything we can do because we really are just waiting for a sudden blinding techno-rapture, with no convincing arguments about when it is likely to occur beyond a 10 order-of-magnitude difference in intuition about what kinds of computer “should be enough”. If we can’t sketch intermediate steps with comprehensible human impacts and tangible economic benefits, are we sure we really understand what we’re doing well enough to get to the goal?

Beyond that, it’s interesting that we don’t really have agreed-upon answers to the oft-repeated obvious objections (everybody has their own favorite private answers and we don’t even accept each other’s version):

“AI has failed for 50 years so why will it succeed now?”

“You can’t explain consciousness, love, or even intelligence in a believable way to me, so your claim that you can build them into machines is silly.”

“Real AI researchers think this effort is premature at best. How can all these brilliant PhD-certified scientists all be clueless?”

Put simply, the SIAI statement at the top of this web page is: “In the coming decades, humanity will likely create a powerful AI.” But there is no convincing argument that this is true. It’s a tough case to make to the average guy (or even the smart but skeptical journalist).

Somehow people have to become convinced that God doesn’t mind what we’re up to. I don’t know how, but if AGI and the singularity only resonate with radical atheism that’s a problem if we want broad acceptance.

I am familiar with the standard examples but most of these are one-liners.

More such examples may help in understanding the phenomenon.

Oh, this is a common “theory” amongst AGI detractors. So what if every Singulitarian did yearn for an eternal life? That wouldn’t be an argument against the possibility of creating an AGI.

I’m not sure if putting a lot of effort to reaching out to mass media is a good thing at all, as 99,99% of people certainly won’t do anything helpful no matter how much we talk to them, but might do something harmful. Outreach should be mostly directed to audiences that have significant intelligence.

That resulted in a general ban on reporters at Hacker’s with a very few exceptions such as John Markoff.

I.e., if you don’t want stupid stories by people who really don’t understand the topic, don’t invite them.

Keith Henson

I think the outreach to a broader scientific and business community–those who are truly interested in learning about the technology–does make sense. People from many different disciplines will be needed to address these issues. But those informed folks will not be deterred by popular press articles making fun of Singularists.

Think about it: there are 300 million Americans, and approximately the same number again here in western Europe. If you can even persuade 1% of all of those people that transhumanist/singularity memes are worth listening to, then you’ll have a huge level of support.

Its a question of knowing which section of society you are trying to appeal to. So, forget the bible bashing types, and forget people who are totally scientifically illiterate.

I think that someone from SIAI should write a rebuttal to the above piece and send it in to the wall street journal. The paper would probably publish it because papers like controversy. Perhaps it could even be an open letter co-authored by a few of the more famous transhumanists, people like Kurzweil. This would encourage the paper to actually print it.

The point is that the section of society that we should be trying to appeal to will - by definition - listen to reason.

Outreach to the general public via mainstream media is going to be iffy. Most journalists write puff pieces most of the time. Summarizing a topic like AGI or the singularity for a lay audience is beyond what they are prepared to do, and probably beyond what they are capable of. It seems pointless to worry about such first impressions considering how little respect media gets from readers anyway. Most readers understand the flighty and shallow nature of media and are perfectly willing to take what they read with a grain of salt. The article gets us out there into the public consciousness in a small way. It’s a start.

In the end, real impact on mainstream perceptions comes from narratives, not from analysis. 1984 created a powerful set of symbols for us to use in considering technological topics which had been discussed exhaustively but which never quite impinged on the public imagination. Star Trek and The Matrix and Blade Runner promote widespread discussion of technology issues more effectively than ten thousand newspaper editorials.

Your loss.

But here’s what I tried to send ya:

You’ve got a multi-faceted problem to deal with, here. The best kind, imo. I like the challenges involved.

What I have to offer first is a work in progress. A a draft of a three part monograph called, “Social Singularity.” It addresses one facet of the problem you face wrt the problem as stated in the article referenced above.

Whether you know it or not, you don’t have a choice about dealing with the general public on this one. The good news is, most people are interested in the truth and in transparency and accountability. In the very near future, as based upon my research to date, I think that the credibility of all of us will be based primarily upon those three attributes.

I have other material to offer that may serve as an example of effective presentations to the general public. For example, several months ago, I assembled a list of material from peer-reviewed papers and articles to show a few friends that mankind presently has the technology to create a functional cyborg. It wouldn’t be easy or cheap, but we can do it by using ‘off the shelf’ technologies. It is no longer theoretical. Yes.The Six Million Dollar Man could be a reality…but he would probably cost a bit more. In the end, he would probably resemble the Borg more than he would Lee Majors. Presented in a more informal fashion, my science and technology challenged friends immediately made the connections for themselves. They immediately understood the reality.

We can also use ‘off the shelf’ technology to create a working “Star Trek” holodeck. The only lack would be a fully functional haptic interface, but even that limitation could be overcome if we weren’t concerened with certain ethical considerations. I presented the material in a similar fashion as in the previous example. Again, my tech-challenged friends immediately ‘personalized’ the information.

The approach I used worked very well to introduce cutting-edge technologies to a number of people who weren’t otherwise familar with leading-edge science and research.

I believe a similar approach could be used with presentations of AGI research and technology in outreach efforts to the general public and the media.

If you are interested,

This means that it will be important for SIAI to go out of our way to provide Singularity 101 to the journalists and bloggers at the next event so they digest our key talking-points for their pieces.

They just need a headline that they understand - next year it should be: Valley companies making great strides in AI, Nanotech etc.”

We’ll need to leave the metaphysical discussions to breakout sessions and keep the main stage focused on progress and key challenges.

- Jonas

—————————————————–

Editor,

I must have been at a different Singularity Summit from the one that Lee Gomes wrote about so derisively on September 19th. First of all, the technological singularity gets its name from the metaphor of a black hole, a collapsed star that has such intense gravity, light cannot escape. The technological singularity, refers to a fantastical period of time in the near future where things we value will change so profoundly that it is extremely difficult, from our current perspective, to shed light on what will happen beyond a certain event horizon. The timing of the singularity depends on whether there is a hard take or a soft take off in the chain of recursively, self-improving, thinking machines that will probably result after we build the first one. Much discussion in the artificial intelligence community revolves around ethical questions related to quarantining or slowing down the development of thinking machines so that we can minimize the risk of being swatted like flies before they evolve into godlike entities that might feel more kindly toward us.

Gomes wrote that the projected date for the emergence of a new super-intelligent life form was from nine to 40 years hence. Actually, Ben Goertzel, one of the speakers at the Summit, said that Artificial General Intelligence (AGI) could happen tomorrow if the right software could be written. Other speakers made guesses that ranged up into in the category of “who-knows-when” to “maybe never.” A Singularity Institute poll of members by Bruce Klein done just before the conference, showed that half thought it would happen by 2050. Seven percent thought it would never happen. Many others didn’t want to make a prediction, or had some other idea. The 800 attendees certainly could not be lumped into a class of quasi-religious zealots or UFOlogists longing “for an intergalactic messiah” as depicted by Gomes. Science is a search for the truth, not the search for god, even though it might ultimately–in some sense–turn out to be the same thing.

In any case, even if the odds were only 5% of it happening, the potential effects are so astounding that we need to take it seriously both as an existential risk and as something potentially wondrous. If a meteor was hurtling towards the Earth with even a 5% chance of destroying it, a huge global effort would be made to do something about it. Gomes, completely ignoring the startling pace of accelerating technological progress, rhetorically asks, “…don’t singularity people know that AI researchers have been trying to make such machines since the 1950’s, without much success?” Here’s another rhetorical question: “Haven’t we learned our lessons from history?” It was commonly predicted in 1900 that heaver-than-air craft would never fly based on efforts going back thousands of years. It was even portrayed in mythology as an affront to God. Sixty-seven years after Kitty Hawk, we were on the moon, and the rate of technological change has continued to grow exponentially since then. Gomes also wrote that long ago he spent an evening with Arthur C. Clark, the creator of the fictional HAL (the intelligent computer in the movie 2001) but came to a different conclusion, even though he “studied with the master.” He should go back, study his lessons, and read what Clark wrote in 1962 about the three “laws” of prediction:

1. When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.
2. The only way of discovering the limits of the possible is to venture a little way past them into the impossible.
3. Any sufficiently advanced technology is indistinguishable from magic.

The singularity may seem impossible to the uninitiated, but it is a reasonable hypothesis about what is coming based on an overwhelming range of evidence. It deserves serious attention from a serious newspaper.

Michael E. Arth

Certainly it is not helpful to trot out the old “they said man would never fly either!” — since it can (and has) been used to support ANY technological forecast, it is always dismissed with a chuckle and the entire argument along with it by association.

1. AGI is obviously impossible
2. This community is centered around AGI and its consequences
3. Something is “wrong” with people that believe in something obviously impossible.
4. Suggested stereotypes.

Once 3 is established, it’s just too easy to throw in the nerd, socially inept, or religious fanatic stereotype. As someone here has already pointed out, it really is irrelevant whether the community is full of nutcases or not for the purposes of determining whether AGI is possible. But the article is not concerned with whether AGI is possible or not, it is assumed axiomatically that it’s not, and the community is described starting from there.

Now, following lamis’ request, here’s my take as to what people must associate with SIAI, hope you get the idea from these vague points:

More science, less science fiction
More AI less Singularity
More math, architecture, code, less text and vague speculation.

I wont elaborate but consider:

The amount of “hard” AI research produced by SIAI was until very recently, ZERO. And without concrete and tangible research, it’s too easy to classify SIAI as vaporware, nutcases, etc as did wsj. The steps taken with Ben Goertzel as director are definitely the way to go, and im happy to see concrete research proposals in the pdf.

1. The SIAI should spend in the region of $50,000 to $100,000 per year sponsoring selected researchers doing new and innovative work on AGI. I’m assuming that the SIAI can afford that? These researchers should be chosen by a panel of people drawn from today’s “elite” of private AGI research - Eliezer Yudkowsky, Ben Goertzel, Bruce Klein, etc, and a few people who hold academic positions at good computer science departments; the criteria for funding being heavily skewed towards proposals which try new approaches to the problem of AGI, and mechanisms for making AI which lend themselves well to friendliness optimization. There should also be a focus on mathematical theories of mind which allow us to make notions of friendliness more precise, because at the moment we are stuck discussing all of these issues in natural language with all of its ambiguities and inconsistencies.

2. The SIAI should sponsor an individual or small group of researchers to perform a thorough literature review of AGI research; that is to prepare a long report which details all of the approaches to AGI which have already been tried, and what their strengths and weaknesses were; this should be published on the SIAI website, and publicized as widely as possible. Perhaps such a review already exists? If so, find it and put it on the site.

3. The SIAI website should have a section which acts as a forum for exchange of ideas amongst AGI researchers, and as a point to publicize the importance of AGI research to computer science students. Perhaps it would be a good idea to host AGIRI on the SIAI website?

http://www.agiri.org/wiki/Main_Page#About

What I’m trying to get at here is that SIAI needs to spend a lot more time developing contacts in the academic world, and it needs to be seen as a hub for serious AGI research. This will serve two purposes: firstly it is good for the credibility of SIAI - as discussed above by bascule - and secondly it will re-inject some life into the field - which is rather unpopular at the moment.

If an intelligent critic could be convinced to engage in an extended dialog about their criticisms, to discover the root points where actual disagreement lies, that would be good also.

Now that two singularity summits have gone by, it might be time to get down to nuts and bolts for the next one — focus on business opportunities, measurable targets for the future, and recaps of recent successes.

Somebody (perhaps Ben Goertzel?) once suggested a style of online discourse where two opposite sides of a concisely-stated position iteratively edit their position statements until both sides think they have done the best they can. If SIAI could find a web developer to set something like that up and convince a few people to participate, it could be a good way to debate some of these issues.

The important thing is to engage the Singularitarians in an open and ruthlessly rational effort to focus on good answers to objections and maybe discover exactly what we do not know or agree on at this point.

This might be a weak response, but it’s the best I can think of. It might be that nothing productive can really be done for broad outreach until there is some concrete and impressive progress to report, in which case (as others have suggested) ignoring the detractors and focusing on the business people and thinkers who intuitively “get it” is probably more fruitful than trying to go mainstream prematurely.

In this view, a response to this article would simply be “History grinds up the shortsighted for fuel. Always has, always will.” The Singularity Summit should have talks about Jupiter Brains, the evolution of Second Life, and designs for nanotech light sabers. Technohippies should read poetry and there should be a fashion show for wearable computers emphasizing the merging of technology and humanity.

By not taking ourselves so seriously, we could lay some groundwork for the time when results make the ideas less easy for unimaginative critics to simply shrug off.

As a charity, SIAI should do research that nobody else will. I don’t think that SIAI-funded near-term nuts and bolts AI research is appropriate. There is one thing that is appropriate, though, and it’s your main stated research goal: Developing Friendly AI theory toward buildability.

I am personally a bit skeptical that it’s even possible, and am further skeptical that Friendly AI will somehow be easier to build than “normal” AI built with good intentions. However, I contribute to SIAI precisely because I might be wrong and Yudkowski et al might be brilliant enough to pull off a theoretical triumph. Alas, from the publication list, there isn’t any indication that any progress has been made for years and years. How many years are left to crack this nut?

Even new but still preliminary steps could provide some guidance, food for thought, or specific structure to reduce risks.

But don’t singularity people know that AI researchers have been trying to make such machines since the 1950s, without much success?

I compare the processing power available from the 1950s through the 1970s with today. I reference some of Ray Kurzweil and Hans Moravec charts.
http://advancednano.blogspot.com/2007/09/ai-1950s-and-now-or-did-you-make-that.html

some have indicated that we now have enough hardware if we are clever enough with programming to achieve AGI. I think having a lot of extra processing power will make the programming a lot easier. Microsoft makes better versions of excel with manageable programming by programming in a wasteful way.

===
It is simple to communicate to the every day person how things have changed and why what was not possible or failed in the past is now becoming possible and something to be concerned about.

I am concerned that a ten year VCR could be programmed into an AI ? No.

how about a supercomputer in ten years with ten thousand petaflops with a million qubit quantum computer coprocessor and another billion neuron coprocessor rendered in hardware ?

How about in twenty years with a MNT enabled machine with a trillion petaflops, a trillion qubit quantum computer and a thousand trilion neurons device ?

It’s not about an unpredictable future. It almost the opposite: we want the world to be predictably better than we left it, to a reasonable standard of prediction. Extreme change is not a necessary consequence of a Friendly AI intelligence explosion. Extreme positive change is a different matter, but this sentence doesn’t imply that.

“The timing of the singularity depends on whether there is a hard take or a soft take off in the chain of recursively, self-improving, thinking machines that will probably result after we build the first one. ”

This jargon is not very clear to the first comer.

“Much discussion in the artificial intelligence community revolves around ethical questions related to quarantining or slowing down the development of thinking machines so that we can minimize the risk of being swatted like flies before they evolve into godlike entities that might feel more kindly toward us.”

The important question is about how to design a thinking machine which doesn’t want to swat us. Thinking machines don’t evolve, they deliberately self-modify. “Godlike entities” — true in the sense of power, but not “personality”. “…that might feel more kindly to us.” — it’s not about guessing things which “might” help for some unspecified reason, but designing things that do help for technically describable reasons.

Why this talk of creating a God? SIAI has never intended to create one. How familiar are you with SIAI’s existing literature?

The last two paragraphs seem good, though

As a more personal opinion, ill say that research into “surrounding ethical and social issues” is a potential time sink if done prematurely. Prematurely means before AGI and relevant concepts have been defined precisely. For example, the notion of friendly AI is way too general a concept to attempt to pin down in the abscence of a specific AGI proposal. If you try to define friendly AI without any AI context, i get the feeling you’ll just get into philosophical matters that can go around in circles endlessly. But as i said, this is a personal opinion, I could easily be wrong.

“desire to actively help even better”

Definitely would be willing to contribute, if possible.

For 2008, I very much am interested in building a sideshow around our core content that will be compelling to a broader audience. Promoting startups / investment opportunities that are somewhere along the Singularity Enablement vector (even if it is search engines or games or fashion) is the kind of step that we need to resonate with the popular press. The truth is: They will cover us regardless of if we embrace them or ignore them. We need to do a better job presenting a compelling story so we can attract a broader class of benefactors to move the mission forward.

AI’s failure is not a failure to build a big enough computer, it’s 50 years of failure to understand the mind well enough to write a program. Let’s pretend you have the fantastic computer you write about. Where’s the program you’ll run on it?

AI has failed for 50 years to find that program, or even get close to it, independent of whether a computer is available to run it in real-time. I see no exponential curve in the philosophy of mind or psychology… so the question remains: Why is a reversal of this failure imminent now?

Mainstream AI programs are solving the automated car driving problem.

Mainstream AI programs handle robotic walking.

there are speech recognition systems with 99% efficiency.

there are vision categorization systems.

There are various expert and learning systems with varying degrees of success.

AI controlled financial trading.
AI medical systems.
AI search engines.
AI character recognition
AI robotic control.

Various programs are available that perform interesting and significant portions or tasks.

If I have 100 to 1000 times the hardware than I can run all of the programs simultaneously even using current methods. Then using XML and web services a loosely coupled integration layer can be generated.

Put of all the information gathering and interpretation systems with sensors that feed all of the specialty sub-systems, make them self-contained.

As I said, wasteful use of a lot of spare resources let us advance the integrated state of the art. Without further inspiration we end up with an AI Rainman collective. Each one can do something better (sometimes a lot better) than a normal person. How good can we get the task parser, context analyzer and coordinator who identifies which thing should be handled by which AI Rainman ?

Is it correct to assume that we do not make any further breakthroughs in programmnig ? It seems clear the AI Rainman collective is achievable. It seems we should be able to do better than that.

My advice to everyone. Forget it now. The article and the attitude it represents is meaningless. Any irritation you feel about it is simply a psychological reaction. Public relations to bring in aggressively ignorant “skeptics” is meaningless. We will get to the singularity step by step and each step will be taken as a goal in itself. I don’t think that the singularity will have to be “created” anymore than capitalism has to be created by a central authority. It will be an organic process which will grow on its own.

Let’s concentrate on each step. And concentrate on the people who “get it” to make each step. People who don’t get it are simply irrelevant. Personally, my feeling is that the less attention we draw to the ultimate goal the better because we would simply illicit negative responses, whereas the individual steps will not. But I acknowledge that this itself is probably just a feeling.

“Much discussion in the artificial intelligence community revolves around ethical questions related to quarantining or slowing down the development of thinking machines so that we can minimize the risk of being swatted like flies before they evolve into godlike entities that might feel more kindly toward us.”

To which you responded: “The important question is about how to design a thinking machine which doesn’t want to swat us. Thinking machines don’t evolve, they deliberately self-modify. “Godlike entities” — true in the sense of power, but not “personality”. “…that might feel more kindly to us.” — it’s not about guessing things which “might” help for some unspecified reason, but designing things that do help for technically describable reasons.
Why this talk of creating a God? SIAI has never intended to create one. How familiar are you with SIAI’s existing literature?

1. I was using the term “evolve” in the everyday sense of the word, such as in “Google is co-evolving with the net”, not necessarily in terms of natural selection. However, AGI may very well evolve in the more technical use of the word. Ben Goertzel, for one, writes about “The Evolving Mind” and “Self-Organizing Evolution.” Goertzel and others are working with self-evolving algorithms.

2. When I wrote “Science is a search for the truth, not the search for god, even though it might ultimately–in some sense–turn out to be the same thing” I was writing about how AGI will almost certainly seem godlike to us. The key phrase here is “in some sense.” It’s certainly not about Jehovah.

I write weekly about macroeconomics. What is most interesting to me in the present context is that the WSJ - the repository of stuff about the economy - should miss the implications of an event which may change the economy to an extent not seen since the Industrial Revolution. Now that is funny.

Ok, that is a common meaning of the word evolve.

You can find some sense in which a powerful AGI is like a god, but these are not the obvious ones. That is, comparing it to a god will lead to a bunch of mistaken inferences e.g. that the AGI would be worthy of or want worship, that it would be morally superior to a human, or that it would be a powerful being with its own agenda that may or may not involve being kind to humans.

For example, a Coherent Extrapolated Volition AGI would simply compute for a while, investigate the world a bit, then output a program. This program will do something humanity can be said to coherently want or nothing if anything goes wrong.

That’s not much like a god, except that its actions could shape the solar system. But then so could a supernovae, and one wouldn’t call that a god. “Godlike power”, perhaps, but there is little fear about making the wrong kinds of inferences about an exploding star since you know it doesn’t have a mind. There are more similarities between AGI and gods than this, but I don’t think they improve the comparison.

Mind you, I suspect sending the letter was a good idea. I’m just pointing out things that seem misleading or incorrect as I understand your words

p.s. Yes, men play a role in baby-making, (I didn’t miss that health class) but woman give BIRTH.