Sunday, September 21, 2014

Peering into the Future: AI and Robot brains

In Singularity or Transhumanism: What Word Should We Use to Discuss the Future? on Slate, Zoltan Istvan writes:

Singularity-word-cloud"The singularity people (many at Singularity University) don't like the term transhumanism. Transhumanists don't like posthumanism. Posthumanists don’t like cyborgism. And cyborgism advocates don't like the life extension tag. If you arrange the groups in any order, the same enmity occurs." 

See what the proponents of these words mean by them...
...and why the old talmudic rabbis and jesuits are probably laughing their socks off.
==Progress toward AI?== 
Baby X, a 3D-simulated human child is getting smarter day by day. Researchers at the Auckland Bioengineering Institute Laboratory for Animate Technologies in New Zealand interact with the simulated toddler, reading, teaching, smiling, playing games, even singing into the computer's microphone and webcam. The blonde youngster mimics facial expressions, laughs, reads words, even cries when he is left alone.
1400832509352"An experiment in machine learning, Baby X is a program that imitates the biological processes of learning, including association, conditioning and reinforcement learning. By algorithmically simulating the chemical reactions of the human brain— think dopamine release or increased oxytocin levels— and connecting them with sensory digital input, when Baby X learns to imitate a facial expression, for instance, software developers write protocols for the variable time intervals between action and response. Effectively "teaching" the child through code, while engineering such a program is no cakewalk, the result is an adorably giggling digital baby with an uncanny ability to learn through interaction," writes Becket Mufson, in the Creators Project.
This is precisely the sixth approach to developing AI that is least discussed by “experts” in the field… and that I have long believed to be essential, in several ways. Above all, by raising them as our children – even fostering them to homes in small robot bodies – we will gain many crucial advantages – that I lay out (somewhat) in Existence.
Meanwhile, Cornell's Robo Brain is currently learning from the internet -- downloading and processing about 1 billion images, 120,000 YouTube videos, and 100 million how-to documents and appliance manuals, all being translated and stored in a robot-friendly format, accessible to 'helper' robots who will function in our factories, homes, and offices. “If a robot encounters a situation it hasn’t seen before it can query Robo Brain in the cloud,” said one researcher. Follow its progress on the Robobrain website.

It's coming. "You can call it a Virtual Personal Assistant, an Intelligent Agent, an Intelligent Interface or whatever you wish. We call it inevitable," writes Dag Kitlaus in A Cambrian Explosion in AI is Coming in TechCrunch. An interesting assessment of the era of the digital assistant and the likely tsunami of new variations, building upon variations.
Meet Jibo, advertised as "the world's first family robot." Kinda creepy but attractive too…
Asimov-three-laws-roboticsEver hear of “neuromorphic architecture?” Silicon chip design that uses transistors — (5 billion of them in the latest IBM chip) - to create analogues of the nonlinear response patterns of biological neurons. The latest version, from IBM, is called “True North” and it is simply spectacular. Its prodigious pattern recognition capabilities are only matched by its stunning (by four orders of magnitude(!)) power efficiency. This is where Moore’s Law, augmented by new neuronal and parallelism software, may truly start delivering.
Now… How to keep what we produce sane? And where on the chip - pray tell - do the Three Laws reside?
Ah, well… I have explored the implications (yin and yang) of the Asimovian laws in my sequel which tied up all the loose ends in Isaac’s universe – Foundation's Triumph. Meanwhile, serious minds are grappling with the problem of “how to keep them loyal." For example…
==Creating Superintelligence==
"Risks that are especially difficult to control have three characteristics: autonomy, self-replication and self-modification. Infectious diseases have these characteristics, and have killed more people than any other class of events, including war. Some computer malware has these characteristics, and can do a lot of damage...
"But microbes and malware cannot intelligently self-modify, so countermeasures can catch up. A superintelligent system [as outlined by Bostrom would be much harder to control if it were able to intelligently self-modify." writes Bostrom.
Nick Bostrom makes a persuasive case that the future impact of AI is perhaps the most important issue the human race has ever faced. Instead of passively drifting, we need to steer a course. Still, his litany of “be careful what you wish for” parables is taken straight from the pages of a century of science fictional “what-if” scenarios. Geeky sci fi archivists need to be present, during the programming, to point out: “you may want to rephrase that… cause way back in 1947 Leigh Brackett showed that it could be misconstrued as...”
== and more on "intelligence..." ==

When did homo sapiens become a more sophisticated species? Not until our skulls underwent “feminization." Interesting article! In fact the mystery of the First Great Renaissance... the burst of human creativity around 45,000 years ago... is discussed in EXISTENCE!
But -- if I may mention it -- the real correlation with this notion… that sexual selection resulted in gentler, more “feminized” males, was presaged by this paper of mine… Neoteny and Two-Way Sexual Selection in Human Evolution.
==Developing Brains==
EMPATHYResearcher Talma Hendler has found evidence for two types of empathy, each tied to a different network of brain regions. One type she calls mental empathy, which requires you to mentally step outside yourself and think about what another person is thinking or experiencing. Parts of the frontal, temporal, and parietal cortex that make up this network. The other type she calls embodied empathy; this is the more visceral in-the-moment empathy you might feel when you see someone get punched in the guts. Very cogent and thought provoking.
This interesting article in Wired explores how movies exploit both of these networks to make you identify with the characters. Only the manipulation is now going scientific!
And veering a bit... When did modern humans arrive in Europe, and by how much did they overlap with our fading cousins, the Neandertals? New studies suggest it all happened earlier than most had assumed, perhaps around ...45,000 years ago.
Now throw in.... Children and adolescents with autism have a surplus of synapses in the brain, and this excess is due to a slowdown in a normal brain “pruning” process during development.
==and organs==
Scientists have for the first time grown a complex, fully functional organ from scratch in a living animal by transplanting cells that were originally created in a laboratory to form a replacement thymus, a vital organ of the immune system.
By deciphering the detailed gene expressions by which a lizard regrows its tail, scientists hope to re-ignite regrowth processes in mammals like us, that have been dormant for 200 million years.

 Both of these stories are straight from my story “Chrysalis” in this month’s ANALOG! Have a look and see where all this may lead!
Scientists report using laser light in ultrafast pulses to control the quantum state of electrons contained inside nanoscale defects located in a diamond, and also observe changes in that electron over a period of time. The findings could be an important milestone on the road to quantum computing.
SCIENCE-TECHNOLOGYAnother team has devised a way to make microscopes magnify 20 times more than usual. This magnification allows scientists to see and identify substances and matter as minuscule as or even smaller than a virus.
Direct synthesis of ammonia from air and water? At low temperatures and pressures? If this membrane method can bypass the usual harsh processes, the news can be significant for liberating poor farmers everywhere to make their own fertilizer.
Looks plausible… if amazing! A transparent luminescent solar concentrator developed in Michigan can be used to cover anything that has a flat, clear surface. Visible light passes through. But organic molecules absorb invisible wavelengths of sunlight such as ultraviolet and near infrared, guiding those packets to the edge of the solar panel, where thin strips of photovoltaic solar cells pick it up and convert it into energy. Fascinating… another potential game changer.
what-if-munroeRecommended: what if? Serious Scientific Answers to Absurd Hypothetical Questions by Randall Munroe (of the brilliant xkcd).
Researchers from UC San Diego's structural engineering department are using drones to capture unique views of the earthquake damage to Napa's historic landmarks. Our own Falko Kuester explains how this new tech is helping.

How to tell if a Chelyabinsk style meteorite came from an asteroid? Here's the basic rule of thumb. “The speed of whatever collides with Earth’s atmosphere depends on its orbit, which in turn depends on its source. The impactor’s entry at 19 km/s means that it came from the asteroid belt between Mars and Jupiter, not from a ballistically launched missile, whose speed is less than 11.2 km/s; a short-period comet, with an average speed of 35 km/s; or a long-period comet with an average speed of 55 km/s. As investigators began retracing the path of the meteor that blazed across the sky, their reconstructed orbit bore out that provenance.”  
Oh, anything much faster than 60 kps either fall naturally from outside the solar system... or was accelerated by someone with boojum powers and maybe ill intent!
And finally:
Don’t bogart that puffer, my friend. Dolphins pass around a puffer fish — apparently to get high off its toxins. After a few chomps, you no longer give a fugu.


Jonathan S. said...

I think I'd rather keep my autism than have a drug that randomly prunes neurons. Sounds more like a non-invasive lobotomy than a useful therapy to me.

Paul451 said...

New paper on Gamma Ray Bursts as a solution to Fermi's Paradox.

Puts some numbers on the frequency of GRBs vs location. Basically, the bulk of the galaxy is too crowded, leaving only the outer, thinner regions. And even there, the early galaxy was too small/dense for the first 2/3rds of its existence. Only after a bunch of mergers and expansion would there have been outer regions far enough away from the action for life to develop without too frequent GRBs.

Via Crowlspace: Adam speculates that Earth may have had, and lost, macroscopic life in the Archean, setting evolution back a billion years.

["isaympag appears": And all the worlds trembled.]

David Brin said...

The Gamma Ray burst hypothesis is certainly a compelling one.

ZarPaulus said...

How many people here actually use Siri or Cortana (isn't that the AI from Halo?) as anything but a dictaphone?

locumranch said...

True North sounds like a significant advance when compared to the 0101 binary microprocessor.

Baby X does not. It cannot be said to 'learn' because it uses human coders to simulate learning and, since it appears incapable of either self-programming or initiating action, it does not meet AI criteria.

On a more biological note, Neoteny is a well-established evolutionary concept but 'feminization' is not; and, as it is quite unclear how one can differentiate between the feminine and the merely juvenile in the absence of gender, it sounds eerily reminiscent of cultural projection bias.


Alex Tolley said...

@ZarPaulus How many people here actually use Siri

As a Brit living in the US, I find Siri has a lot of trouble with my accent. It seems to be getting better, but at this point it is more annoying than useful. In another 5 years however...

@locum Neoteny is a well-established evolutionary concept but 'feminization' is not; and, as it is quite unclear how one can differentiate between the feminine and the merely juvenile in the absence of gender

I agree with this difficulty. Not sure about cultural projection bias, perhaps more wishful thinking? Estrogenic emulators in the environment are feminizing males, both animal and human with measurable characteristics, none of which are skull measurements AFAIK.

Alex Tolley said...

@DB Scientists have for the first time grown a complex, fully functional organ from scratch

Assuming we can do this as a standard medical procedure in the future, and that other life extension approaches work -e.g. mitochondrial replacements - how does this square with your general sense that human lifespans are already nearly maxxed out? Or do you consider these artificial approaches separate from natural lifespans?

A Cambrian Explosion In AI Is Coming

Not a new idea, but the brief article does raise the issue of monetization. Interesting. What alternatives might there be?

Tony Fisk said...

On a less spectacular but very useful note: surgeons are experimenting with replacing worn knee cartilage with nasal connective tissue

Tim H. said...

Thanks for that link, Tony. My knees give me the excuse to quote Marvin, "I ache, therefore, I am.".

Tony Fisk said...

A terrible pain in all the diodes down your left side?

Tim H. said...

More like a nagging ache in all the heavily used connective tissue, but that's the picture.

Anonymous said...

I'm always amused when digital logic is used to simulate non-linearity. Especially as the transistor is a non-linear device to begin with. I think I'd have used analog computing more, but as a discipline, it got eaten by digital.

Jumper said...

Anthropologists have used the terms "gracile" and "robust" to indicate "feminine" or "masculine" (such terms recognized as yes, cultural bias) forms. It ranges through many mammals in various epochs, having as much to do with the ability to be fleet and agile as opposed to stand-and-fight as anything else.

SteveO said...

Instead of Baby X and human coders, I would have set it up with an FPGA to reprogram itself in response to stimuli.

FPGA are chips that can reprogram themselves on the fly. If I was an ace programmer and had a couple, I would totally set that up to run for a few years...FPGA have already been used to create novel algorithms using genetic selection criteria. I think FPGAs or something like will be part of any AI we end up with. Otherwise, you have a static "brain."

Alfred Differ said...

The more I think on it, the more I think the AI folks are going to have to crack two problems to reproduce anything remotely like us. There is the obvious problem of simulating the brain and what happens on it. However, we aren't exactly the same kind of human in isolation, so they are going to have to consider the memescape and what languages really are.

Ultimately, I think the real distinction between us and our cousin species will come down to the richness of the memescape. Simulate only the brain and you might reproduce homo erectus. We are much too social not to address that space of interactions.

Tony Fisk said...

Having spent the first part of 'Zendegi' to describe the communications arms race between an entrenched oligarchy and a popular uprising, Greg Egan uses the second part to depict attempts to 'upload' human awareness. The final conclusion, from the AIs created by this process, is not to do so until it can be done *properly*. I suppose it's the same moral as the tale about the painted fish that demanded a soul.

locumranch said...

Horace Mann, the founder of the US educational system, believed
that women were morally superior to men. This superiority,
according to Mann, would permit females to instill their morals
in the common school students they taught, allowing them to "remove the evil” that resided within their masculine students and replace it with a feminine virtue, conduct and character more appropriate to republicanism and good citizenship, the converse being likewise true, meaning that 'homo sapiens', by virtue of being 'more evolved' than their brutish and animalistic predecessors, must also be more feminine and/or feminized, insomuch as terms like 'civil', 'moral' and 'proper' have become modern euphemisms for the condition of being feminine.


sociotard said...

recomended: this rube goldberg device based almost entirely on optics.

No fiber optics though, which is a shame.

Duncan Cairncross said...

Who are you?
And what have you done with Locumranch?

David Brin said...

Duncan be nice. Lately Locum has been very cogent, at least as often as when he's not. And I am starting to get a feel for what makes the difference. A fascinating fellow, actually.

Duncan Cairncross said...

Hi David
I agree - but I though that last posting was really good!

Another BIG step

TheMadLibrarian said...

I think the worst that might be said about locum is that he appears easily bored. Periodically he feels the need to 'stir the pot' and see what happens.


Paul451 said...

Fashions seem to shift back and forth whether women are proclaimed to be genteel and demure, lustful and demanding, fragile and emotional... the saner, stoic sex or the (literally) hysterical one.

(Of course, we also project onto those proclamations our shifting historical stereotypes about the guys making them.)

Paul451 said...

(I'm not sure why I italicised "the" in that last line.)

("Seaudy psaltery": Saucy psandwich.)

Alfred Differ said...

Any time someone suggests women are morally superior to men I get suspicious that their real intend is to exclude them from the business of men. I get there are differences, but I think there is some turf protecting going on when people focus too much on the supposed differences in our minds.

David Brin said...