Recently in AI Category

http://www.bbc.co.uk/news/science-environment-11041449?goback=%2Egde_37945_news_178002176

Nice article about the need to spend some SETI time looking for AI signals and hang-outs, rather than focussing solely on the activities and locations of biological life-forms. Personally I still think that "first contact" is likely to be with a dormant/dead AI controlled Von Neuman type probe in orbit around a gas giant or a NEAR type orbit.

Loebner 2009 Video

| | TrackBacks (0)

Great video report from Erwin Van Lun of the 2009 Loebner prize. Having read (and analysed) the transcripts it's hardly surprising that none came close to winning. Looking closely at entering in 2010.

http://blog.wolfram.com/2009/03/05/wolframalpha-is-coming/

This could be interesting - from the guy who created Mathematica - a plain language factual question answering system.

"It's going to be a website: www.wolframalpha.com. With one simple input field that gives access to a huge system, with trillions of pieces of curated data and millions of lines of algorithms.

We're all working very hard right now to get Wolfram|Alpha ready to go live.

I think it's going to be pretty exciting. A new paradigm for using computers and the web.

That almost gets us to what people thought computers would be able to do 50 years ago!"

http://blog.thebrain.com/megabrain/

I have Kurzweil's TheBrain on my laptop and have used it a bit, but not this extent!

Humaniti 2100

| | TrackBacks (0)

I must try and get back into the habit of a Friday afternoon blog post. So here's something I've been putting together for a while on my Palm - it is unashamably direct and does not particularly hedge its bets - where's the fun in doing that?.

I still reckon that Greg Egan (in Diaspora) has probably got the most accurate view of how humanity may 'evolve' over the next few centuries. Inspired by him here's my take on where we could be by 2100.

NATURALS

Naturals are those who have refused any sort of body mod or digital existence. In 2100 there will be a frighteningly large number of people living (even in poverty) for whom this may be the only option. Increasingly though this is a moral choice. However general health advances mean that life expectancy is 100+, with good quality of life to 90+ for those in the developed world.

AUGMENTED

The Augmented are those who have taken advantage of the transformative technologies of genetic and nano engineering and digital/cyber mods, but who see their 'self' as purely their organic mind. If they use 'scapes and virtual worlds through avatars it is for specific, non-persistent, purposes. Augmentation itself may range fm slight (e.g. just regular use of an avatar or life-logging systems) to extreme (bio-electronic cyber-systems).

MULTIPLES

Multiples are those of organic descent who have created digital copies of themselves which exist persistently (and probably in multiple instances) in 'scapes across the globe, planets and (by 2200) the stars. I still think that 'uploading' a mind from the brain could be in the near-impossible category. lnstead the first 'copies' will come from explicit teaching/programming, quickly followed by automated learning from email & social media, and ultimately by eavesdropping on everything we do from birth - our lifelog becomes us.

The real challenge for multiples is the re-integration of learnt experiences. Copies can easily just copy data, but what about the organic prime? Again I think that uploading memories is probably a no-no, but in-silico memory accessed through some sort of personal agent or augmented reality (or even brain-jack) would seem achievable.

Freed of a corporeal existence the copies can explore the stars in starship borne 'scapes, and even be beamed from star to star at the speed of light (and bear in mind that since the Multiples sense of time is dependent only on processor clock speed the journey to Alpha Centauri could pass in seconds or millennia - again see Egan, this time in Permutation City).

But perhaps the most telling feature of Multiples is that they can be immortal - so whilst your organic atom based self may die your digital Multiples can live forever - perhaps we might even call them ghosts.

And if we can create simple Multiples now (and I think we can) then it means that we can create simple (multiple) immortality right now - and just think of the moral and ethical issues that raises.

(and if you doubt this whole section take a look at this recent DoD requirement)

DIGITALS

Digitals (who also normally exist as Multiples) are personality constructs that are not derived from an organic, living source. At their most basic, and in current tech terms, these could be virtual receptionists or game NPCs, but very shortly we'll be able to create autonomous, self-motivated avatars - the fore-runners of true digitals. We might also create Digitals from historic personalities, and we could even use software DNA to allow Digitals to breed and evolve (for what a baby Digital might experience read the opening of Egan's Diaspora). The key point is that within the virtual space Digitals are not differentiable from Multiples or the avatars of the Augmented.

GLEISNERS/ANDROIDS

But Multiples and Digitals need not be confined to virtual spaces. Once we have a digital self controlling an avatar body there is no reason why we can't have the same self controlling a robot body. Indeed in building our own AI engine we created a Sensor and Action mark-up language to isolate the AI 'brain' from the embodiment technology. So the same AI that controls a Second Life avatar could also control, and live through, an AIBO or ASIMO. Fast forward 50-100 years and our 'human-level' Multiples and Digitals can walk the atom-based physical worlds in human (or non-human) bodies (what Egan calls Gleisner robots). And whislt embodied Digitals may sound weird just think what it would be like, as the human root of a Multiple, to shake your own hand.

When I first wrote this I entitled it Humaniti 2200 - but I've now put it to 2100. For either date though it's a matter of degree - particularly for the 'organic' element. There is also though the potential of a 'singularity' effect; once we have one fully fledged Digital we could rapidly clone or evolve it (or it could clone or evolve itself) so that the Digital population could go from single figures to thousands to millions in years or decades.

So how might percentages or numbers go? This is probably not even a guess, but its always useful to have some strawman figures.

Type / Date200920152020205021002150 2200
Naturals99%95%90%80%50%30% 10%
Enhanced1%5%10%20%50%70%90%
Multiples0100s10k+100k+10m+1bn2bn
Digitals010s50s+100s+10k+1m++100m++
Gleisners01s10s+10s+1k+100k++1m++
Figures for naturals and Enhanced are % biological human population. + means +/- 1 order of magnitude, ++ means +/- 2 orders of magnitude

And if you doubt the 2015 figures I fully expect Daden to be running at least 1 Multiple (probably me!), 1 Digital (an enhanced Halo) and even a Gleisner (Halo in some 2015 equivalent of a Femsapien).


But the real message is that we can already create Multiples and Digitals, and even Gleisners - they just aren't very good yet! But it means there has been a shift. This is no longer a question of when or if or can, but just one of how good, and how fast will we improve.


Virtual Dialogue Application for Families of Deployed Service Members OSD09-H03

Interesting DoD RFP for a chatbot for absent serviceman's families. Too late for us to respond directly (and the project overall is probably a bit too big for us and US based). but would sure be something interesting to contibute to and I'm sure our AAML/ASML, extended AIML and SL technology would play in nicely.

Question remains though - do you have to pass the Turing Test as part of the project?

I've just posted a new video to the Daden YouTube channel showing our Halo automated avatar being put through her paces. This is a recording of the demo we gave at the BCS AI Special Interest Group conference in December, and includes a demo of the emotion work we have been doing with the University of Wolverhampton.

http://spy.appspot.com/

Neat app, monitoring Twitter et al for "hot topics". Does it do this automatically or when people ask to search? Another neat feed for Halo to monitor.


http://bloggingheads.tv/diavlogs/15555

Neat concept - one on one interview by webcam. This one is Jaron Lanier
Microsoft Corp., University of California-Berkeley, Jaronlanier.com and Eliezer Yudkowsky
Singularity Institute, Overcoming Bias.

Thanks @mendicott for the heads-up

Wiki Sections

Flickr

www.flickr.com
This is a Flickr badge showing public photos from halo4256. Make your own badge here.

Categories

Powered by Movable Type 4.1

About this Archive

This page is a archive of recent entries in the AI category.

Avatars is the next category.

Find recent content on the main index or look in the archives to find all content.