Interview with Peter Voss, Full-Length Transcript
March 03 2008 / by memebox
Category: Technology Year: General Rating: 8
Transcript of Peter Voss Audio Interview
A: Hi this is Alvis Brigis with MemeBox. Today we’re interviewing AI pioneer Peter Voss. Hey Peter, thanks for taking the time to speak with us today.
P: Hi, Alvis!
A: Let’s start with a basic question: What do you do and how is that related to the future?
P: I run an AI startup. We’re developing basic artificial general intelligence technology and how it’s related to the future, well, if this pans out it’s going to have a significant impact on all areas of life and I guess we’ll be talking about that and the specifics a little further on.
(0:47) A: What separates Adaptive AI from other Artificial Intelligence research companies?
P: Well what we’re focused on is specifically on developing the core technology for what we called AGI, artificial general intelligence. And I’ll try and explain a little bit of how AGI differs from normal AI. In fact it’s sort of come full circle. Originally when people thought about AI it was really building machines that could think and learn like humans but that was like 30 years ago, but then over the decades it hasn’t really lived up to that promise – people were just too optimistic about what they could achieve in a short period of time. And so AI over the decades has really become very specialized problem-solving, like a chess-playing machine or medical diagnosis or traffic control systems or financial prediction systems. So the AI systems currently are primarily systems that are hand-crafted by humans to solve very specific problems so the knowledge and skills are specifically designed into the system.
AGI in contrast are systems that can learn, like humans, they can learn pretty much anything, and they can reason about it, they can improve their knowledge, they can experiment and so on. So it’s the ability of having a system, designing a system, building a system that can inherently can learn, acquire skills. Now, this is pretty darn difficult but once you have a system like that of course you can train the system to do any of the other tasks that current AIs can do.
(2:46) A: So how will we know that AGI has been achieved? Is it going to all of a sudden, you know just happen one day, or is more of a general curve into AGI that we’re going to experience?
P: It’s not really clear exactly how fast the take-off of this technology will be. In terms of, from my perspective, we’re certainly not talking decades, I don’t believe it’s going to be decades, or centuries that this is going to take but it could certainly take five to ten years between the first, good indication of AGI to having systems that are super intelligent. I don’t think it will take an enormously long time. A couple of reasons, but part of the reasons for that is that to have a useful agi application, a system that can learn stuff and is actually commercially viable, it already needs to be reasonably good, otherwise there’s no market for it, and then to improve the system to a point where it’s smart enough to learn a lot of different things I don’t believe will take all that long. But you know, it’s new technology and we’ll have to see how it pans out.
A: Can you give us an overview of some of the different approaches being used to achieve AGI?
P: Sure, well first of all I must say there aren’t a lot of people working on this at all. For largely historical reasons, and they actually did a whole long, there’s a whole long list of reasons: part of it’s because AI is over-promised and so on, but at the moment there aren’t really a lot of people working on AGI. Even within that the field is actually quite fragmented. There are people who come from a traditional AI sort of <> approach, there are other people who come from a neural network approach, then there’s a robotics approach, people taking essentially robots or cars or whatever and trying to get them to generally learn how to do stuff, then there’s brains or reverse engineering where people say well, we have the human brain that can do this, so if we can just figure out how to do that electronically then that’ll work, so they basically take the brain as a model and reverse engineer it. Then there are evolutionary approaches where people say well we don’t know how to do it but we can maybe guide it through experimentation to achieve what we want and then there are some various simulations. So as you can see the field itself is actually quite fragmented as small as it is.
A: I’ve heard some people say that Google may be the first to evolve AI due to the huge volume of information that they have at their disposal and all the processing power they also have. Can you speak to that a little bit?
P: Well I certainly wouldn’t write off Google with the amount of talent and money and infrastructure that they have and I don’t know what they’re doing behind closed doors but as far as the basic search technology turning into an AGI ,a system that can by itself learn specific things and follow instructions and so on, I don’t see that as a natural outcome of the Google design. So Google may be designing something specifically. There’s a big problem with bandwith and communication with distributed processing so having a huge amount of processing power that is distributed and has especially good problem with interprocessor communication, that is one computer talking to the other, it would be, well one can sort of imagine the neurons in your brain, were hundreds of miles apart or something like that, well it wouldn’t be as far as that, but if the neurons had to communicate over many meters instead of centimeters, it wouldn’t work very well. But as I say, I wouldn’t write off Google but I don’t think that their main thrust is toward AGI.
(7:43) A: What impact would a fully-operational AGI system have on the world economy? Any specific examples? I assume it would be pretty disruptive.
P: Yes, it certainly would, will be disruptive. The simplest way one can see that is there would be of course many tasks that humans do right now AGI could do as well or in many cases better, so the crux of many, especially desk-bound applications, are going to drop dramatically, which of course, as before with automation will have the end effect of vastly improving the standard of living and the wealth across the world but it will be disruptive for the people doing those tasks they will end up doing something else that computers aren’t as good at.
(8:50) A: What’s the likelihood that AGI will be achieved by a private company as per se a government?
P: Well, I guess the story I tell my investors, those investing in my company, is that I believe our company has an excellent chance at achieving that breakthrough. Standing back if I have to give an objective assessment of ways it’s likely to happen, I still believe it is at the moment looking around likely to happen from a private company rather than government. I don’t government see having any concerted effort to actually put that together, their thinking doesn’t seem to be along those lines. But again, it’s hard to know what’s happening behind closed doors, it really just takes one or two people with the right vision in a government position to put together some effort like that. They certainly have the money they can tap into and skills and university affiliations and so on, so it could happen but right now I actually see it being driven by industry.
(10:13) A: Many people have a favorable view of AI and the possible benefits of AGI but some people don’t. Where does opposition usually come from? What sorts of arguments do people who are against this sort of technology usually post to you in conversation?
P: Well, it comes from a very wide spectrum. In sort of circles of futurists, often people just like having a good spirited debate [A: Sure] and be controversial so a lot of those discussions people are often pushing the arguments to just see what comes out. But of course there is the ground swell of opposition to any new technology, to any change, which just comes from people’s fear of change in AGI more fear of the changing landscape of job markets and so on. But the technology, AGI technology, will be very disruptive in many ways there will be major change and I think that’s fundamentally what people are scared of. Then of course there’s also the huge sort of anti-technology movement in the world now, that technology is inherently bad, that we’re messing up nature in a way. Now obviously I don’t share that view at all but that fundamentally we should be slaves to nature. But the XX inherently see opposition from those areas as well. However most people I come across really seem to be excited about the prospects and the real benefits that new technology and AGI in particular can provide.
(12:27) A: Do you view technology and AGI and AI as an extension of that as natural?
P: Yes of course, yes. … Let me qualify that one gets in sort of philosophical debates and you have to say well in what context are you using the term. And I’m saying yea I regard it as natural in the sense that humans clearly are part of nature and our ability, our nature is in fact to be able to invent things and improve things and improve our lives and in that sense it is perfectly natural.
(13:14) A: Would you also see it as a natural extension of the processing that is already going on in nature in biological systems, in complex adaptive systems?
P: I personally have some difficult in wrapping my brain around that level of abstraction of equating natural evolution, Darwinian evolution, with technological progress because technological process is directed specifically by humans, even if we don’t necessarily know the outcome we go into an endeavor specifically trying to achieve a goal. I find it difficult to equate the two and say that technological evolution is the same as natural Darwinian evolution, I think they are fundamentally different so in that sense I do not see them as the same or as one being an extension of the other.
(14:20) A: Okay, to switch gears a little bit in one of your answers you mentioned that you’ve spoken to a bunch of futurists, particularly about AI. How would you describe a futurist? What does it mean to be a futurist?
P: I should imagine that the generic term futurist is simply somebody who is knowledgeable about discussions/issues relating to the future, and I guess supportive of it, but I would regard the generic expression as someone who is knowledgeable on the subject. So that’s pretty broad. I describe myself as an extropian, which is a subclass of futurist, and there are many different flavors of futurist and transhumanism are those people who see humanity transcending our biological basis and that we are on the way of becoming something post human, so transhuman is basically someone on the way to it, and extropians are again the sub genre of that and we believe in technological progress to achieve that and personal responsibility and dynamic proactive optimism that we have as opposed to a transhumanist or futurist who may not share those views.
(16:14) A: So as a transhumanist and more specifically an extropian how long do you think you’re going to live?
P: I’d like to describe it as an indefinite lifespan. I hope I’m within the envelope, I just turned 54 so the medical technology would have to advance over the next two decades to start reverse aging for me to have an indefinite life span and I believe that is quite possible with the advent of AGI helping to promote medical research. So yes I’d like to have an indefinite lifespan and I think that that may happen, and that possibility will exist for people living today.
(17:15) A: Can you give us a general overview of what you think the next 20 years will be like? What kind of super disruptive changes will occur that will effect the common person?
P: I publicly made the prediction, or I won’t call it the prediction but the estimate, of AGI, that personally I would be surprised if it’s more than ten years before we have human level, or effective AGI, and I think it could be quite a bit less than that, as little as five years. So when that happens of course there will very dramatic changes but this will enable other technology as well. AGI will allow us to accelerate nanotech development, medical research, that will allow us to deal much better with all sorts of problems, of course disease and aging, but also just reduce the cost of production of all sorts of goods and foods very dramatically and also helped with environmental issues so there will be a snowballing effect started by Agi development. In terms of what will happen and in what year and what chain of events, I have no way of really putting any more numbers on that. Once we have machines that are as smart as humans and we can employ them to help us develop other technologies I think things will happen quite quickly. On the other hand things like aging and some of the diseases are really really hard problems and they have certain constant time delays in them. If you want to try a new drug in humans, it takes a certain amount of time to really concern that it works. You can do a lot of simulations but ultimately they have to be tested in real humans and that takes time. So it’s very difficult to predict the interaction between those various dynamics – the superpower of the computer and the simulations and the rapid research one can do on the one hand versus the actual constants in the real world – delays of FDA approval and people getting funding together and trying things out and doing experiments.
(20:09) A: When it comes to the near-term, what do you see happening in 2008 that would really peak your interest? Any really cool stuff on the immediate horizon?
P: I don’t really have any predictions in the near term. There’s cool stuff coming out all the time, I don’t personally see anything on the horizon that would blow us away in this year or next year but it’s the nature of the beast, that these things suddenly pop up but I don’t have any predictions.
(20:50) A: How about for 5 years out, say, through 2013?
P: I believe that there’s a very good chance that AGI will start being employed and extensively employed within five years. That obviously is a huge change as we will see some of the early AIs general learning ability being employed and the change that people will start to think about what the implications are and then obviously the ongoing improvement of the XX.
(21:40) A: How about by 2018? Ten years ahead. Do you see a trillion dollar AGI company existing at that point?
P: Yes.
A: Could you be a little more specific? What do you think the specific chances are that that’s going to be Adaptive AI?
P: I think we have a reasonable chance, but there are lots of things, all sorts of funding and finding the right people. It’s not just the technologies, developing the technologies is one thing but of course you need to turn that into the kind of business and have the right group running the business and dealing with regulatory and all sort of business issues, competitive issues, the staff and growth and funding. One needs to get most of those right to build a successful company. Yes, at this point it looks like we have a chance at being a dominant player in the AGI field, perhaps the dominant player in the AGI field.
(20:51) A: So what are the next steps ahead for you and for Adaptive AI? What can we can expect to see or hear from you in the near-term future?
P: A the moment we’re still in a stealth mode in our technology. Over the next two years we actually expect to start commercializing some of our, parts of our technologies. I can’t talk about what that will be but we do expect to be up there with some new products, with obviously limited intelligence, general intelligence, or using general intelligence technologies, and that should be up there within the next two years.
A: Great! Good luck with that, Peter, and thanks for taking the time to talk with me and with the MemeBox audience.
Comment Thread (1 Response)
-
Thanks for bringing this interview to us! I was wondering if the audio version of this is available somewhere.
Posted by: gremlinn March 05, 2008
Vote for this comment - Recommend
Trackbacks (62 Responses)
-
fwdfjexs
Posted by: fwdfjexs September 18, 2008
-
tfivxuyl
Posted by: tfivxuyl October 02, 2008
-
alprazolam
Posted by: alprazolam October 04, 2008
-
order tramadol
Posted by: order tramadol October 04, 2008
-
order phentermine
Posted by: order phentermine October 04, 2008
-
prilosec
Posted by: prilosec October 04, 2008
-
cheap carisoprodol
Posted by: cheap carisoprodol October 04, 2008
-
cheap tramadol
Posted by: cheap tramadol October 04, 2008
-
xenical
Posted by: xenical October 04, 2008
-
purchase soma
Posted by: purchase soma October 04, 2008
-
buy adipex
Posted by: buy adipex October 04, 2008
-
buy soma online
Posted by: buy soma online October 04, 2008
-
order cialis online
Posted by: order cialis online October 04, 2008
-
generic zoloft
Posted by: generic zoloft October 04, 2008
-
generic paxil
Posted by: generic paxil October 04, 2008
-
buy propecia
Posted by: buy propecia October 04, 2008
-
keflex
Posted by: keflex October 05, 2008
-
cheap levitra
Posted by: cheap levitra October 05, 2008
-
adipex online
Posted by: adipex online October 05, 2008
-
buy valium
Posted by: buy valium October 05, 2008
-
hydrocodone
Posted by: hydrocodone October 05, 2008
-
buy vicodin
Posted by: buy vicodin October 06, 2008
-
kenalog
Posted by: kenalog October 06, 2008
-
cheap valium
Posted by: cheap valium October 06, 2008
-
buy ambien
Posted by: buy ambien October 06, 2008
-
cheap adipex
Posted by: cheap adipex October 06, 2008
-
fosamax
Posted by: fosamax October 06, 2008
-
generic norvasc
Posted by: generic norvasc October 06, 2008
-
lorazepam
Posted by: lorazepam October 06, 2008
-
purchase xanax
Posted by: purchase xanax October 07, 2008
-
zyban
Posted by: zyban October 07, 2008
-
adipex online
Posted by: adipex online October 07, 2008
-
lorazepam
Posted by: lorazepam October 07, 2008
-
buy xanax
Posted by: buy xanax October 07, 2008
-
buy hydrocodone
Posted by: buy hydrocodone October 07, 2008
-
phentermine
Posted by: phentermine October 07, 2008
-
buy levitra
Posted by: buy levitra October 07, 2008
-
generic levitra
Posted by: generic levitra October 07, 2008
-
premarin
Posted by: premarin October 07, 2008
-
order viagra
Posted by: order viagra October 07, 2008
-
fosamax
Posted by: fosamax October 07, 2008
-
order carisoprodol
Posted by: order carisoprodol October 07, 2008
-
furosemide
Posted by: furosemide October 07, 2008
-
generic wellbutrin
Posted by: generic wellbutrin October 07, 2008
-
cheap carisoprodol
Posted by: cheap carisoprodol October 07, 2008
-
xanax online
Posted by: xanax online October 07, 2008
-
buy fioricet online
Posted by: buy fioricet online October 07, 2008
-
purchase tramadol
Posted by: purchase tramadol October 07, 2008
-
soma
Posted by: soma October 07, 2008
-
generic cialis
Posted by: generic cialis October 07, 2008
-
triamcinolone
Posted by: triamcinolone October 07, 2008
-
carisoprodol online
Posted by: carisoprodol online October 08, 2008
-
amoxil
Posted by: amoxil October 08, 2008
-
singulair
Posted by: singulair October 08, 2008
-
hydrocodone
Posted by: hydrocodone October 08, 2008
-
buy xanax
Posted by: buy xanax October 08, 2008
-
ambien
Posted by: ambien October 08, 2008
-
cheap vicodin
Posted by: cheap vicodin October 08, 2008
-
zopiclone
Posted by: zopiclone October 08, 2008
-
tadalafil
Posted by: tadalafil October 08, 2008
-
levofloxacin
Posted by: levofloxacin October 08, 2008
-
nasacort
Posted by: nasacort October 08, 2008