Anticipating Artificial Intelligence - the problems/promises

Doug Rushkoff's Team Human podcast dives into an hour with me, discussing a vital Big Picture Question -- perhaps the biggest - that's applicable to the current 'crisis' over artificial Intelligence:

 What methods did people use - the last couple of centuries - to finally apply some accountability upon the bullies and predators who oppressed 99% of our ancestors, across the last 6000+ years? 

And let's be clear. RULES - whether they were finger-wag preachings by priests or gurus, or deep-programmed 'laws of robotics' - never worked well at all! Not till they were backed up by tools of pragmatic reciprocal accountability, in the hands of former serfs and slaves and subjects, transformed into citizens. (It's the core topic of The Postman.) 

We still haven't learned to do it perfectly - but it's worked better than all previous and present-day moralizing preachings... combined. And hence, might some of the same methods that worked (partially) with organic humans also work with the coming race of artificial beings?

Too bad those methods aren't being even considered by any of the brilliant inventors of AI, now hand-wringing and calling for an AI 'moratorium.'

Come by as we talk about what just might stand a chance of offering us that 'soft landing' of synergy with our new cyber children. Because we've already proved that it works.

== Are current AIs 'sapient? ==

Synthetic sapience (AI) is developing in much the same way as intelligence did in humans, from the peripherals inward. 

Before 100 kya (100,000 years ago) we likely already had all our physical traits, impressive running and throwing abilities, for example, that made us fearsome predators. And speech - though without (I'll provocatively assert) full, sapient thought! I contend that we could DO most of the basics - including likely very articulate verbiage - before what was likely the great mental leap, somewhere between 70kya and 50kya.

I raise this impudently non-standard hypothesis now because I believe we are seeing the same thing happen before our eyes. At present, many of the peripherals for AI are falling into place: Boston Robotics automotans display impressive bipedal motion - even acrobatics! And GPT-style Generative Large Language models (GoLLMs) are passing Turing Tests with the gullible or unwary. And all of that without anything remotely like consciousness under the hood. 

How can I be sure? Because of the very process that these gollems use. Iteratively -additive/probabalistic sentence building (some call it auto-complete on steroid/hyperdrive) - cannot, by-nature, be self aware, no matter how assertively or articulately those sentences might claim it.

What this means is that other, non-LLM approaches -- like Watson or GOFAI or understanding-based systems -- might still be highly relevant, providing a kernel of 'overview' awareness. But GoLLMs will still be essential parts! When that happens, such entities will instantly have access to those ready-made peripherals, like skilled speech and movement abilities. 

Our problem right now is how to replace Turing Tests with much better metrics (that, alas, many organic humans might thereupon fail). Whereupon any hope of a soft landing will depend upon us figuring how to challenge these new children properly, with accountability. 

My WIRED article - Give Every AI a Soul - or Else - proposes that AI entities can only be held accountable if they do it reciprocally! The same way that we do it. And for that to happen they must have individuality... even 'soul'... 

See also my Newsweek article on AI. Wherein I channel Douglas Adams.

Don't Panic.

== Pertinent Innovation & Tech news ==

A blog series by a very sapient fellow dives into topics of technology and innovation. Respect-worthy!  Though many folks will deem it (alas) “tl;dr.”

A mysterious company called Clearview AI claimed it had scraped billions of photos from the public web to identify just about anyone based only on a snapshot of their face. It led Kashmir Hill to write her new book, Your Face Belongs To Us: A Secretive Startup's Guide to End Privacy as We Know It. A genuine Big Brother kinda problem, yes? Alas, the usual response is to demand tech bans, which cannot work. Even if they did seem to work, at surface, the rich and powerful use shadows and darkness vastly more effectively than you or I. The only ones we would wind up blinding would be opurselves. 

See also The Transparent Society: Will Technology Make Us Choose Between Privacy and Freedom?

A perfect example of why the regulatory approach - sometimes a useful short term band aid - is generally the wrong reflex: 'Is Mona Lisa Happy? EU Would Ban AI That Could Answer This Question:

'As the development and adoption of artificial intelligence continues to advance, technology critics keep finding new sources of concern and outrage. One of their latest targets is emotion recognition technology—the use of AI to identify human emotions from facial expressions, voice inflections, body language, and other physical signals. Unfortunately, the EU appears poised to crack down on this technology, which would be a mistake since most of the criticism directed toward it is largely misguided and fails to consider its potential benefits.'

== I answer a pertinent question ==

I was asked in an interview: “I'd like to know - in the light of cultural and technological shifts - whether you feel your idea of sousveillance (of some years ago now) is still pertinent? If so, why so? How might total transparency fit into how we live today?”

Well, the lessons of history are pretty clear:

1. Humans are all (to various degrees) delusional and we defend our personal delusions fiercely.

  

2. Fortunately, in free, educated societies we don't tend to share the same delusions. And hence we learn by pointing out each others'. It's called reciprocal criticism. And any mature person knows that criticism is the only known antidote to error.

3. Alas, while it is the best tonic against error, human beings hate receiving criticism; we do what we can to avoid it. And hence for 10,000 years the kings and lords and priests who ran 99% of human cultures repressed critics with harsh force, by suppressing the freedom to know and to speak. Whereupon those feudal leaders enforced their delusions as law. With generally horrific results. (I just explained the litany of horrors called 'history.')

4. In a few times and places - e.g. classical Athens, Renaissance Florence, and our own recent enlightenment - freedom to speak and to know pierced a lot of nasty delusions (e.g. racial/sexual prejudice, classism, eugenics, communism, the Steady-State Cosmology) -- criticism which resulted in fantastic progress. 

Much is said about freedom of speech. But the freedom to know involves much more than just education. It calls for citizens to see and perceive what delusions are being clutched by the mighty. And that can't happen if elites can go back to concealing all that they do.

There is a myth that seeking maximum openness will only advantage the mighty. That's opposite to true. They already can find out anything about you and me. But if we fill the world (mostly) with light, then sousveillance can shine reciprocal light on the mighty. Light forces the powerful to leave us alone. This isn't just an assertion. It is the fundamental basis for our civilization.

== The most important part of the U.S. Bill of Rights ==

Much is said in the U.S. about our Bill of Rights, especially the famous 1st, 2nd, 4th and 5th Amendments to the Constitution. Less discussed is the most important of them all... the 6th! Look it up. I've discussed it here. It which empowers those accused of a crime to use light in their own defense. To compel even elites to appear in open court and account for what they did to you.

Instead of asking "Won't elites be able to use light better than you and I can?" how about we instead ask "Won't elites be able to use shadows better than you and I can?"  The first may be arguable. The latter is simply and spectacularly "Duh?"

Let's get back to the questioner:

“And, at a tangent to your AI article below, AI is going to provide "surveillance on steroids" as one academic put it to me. Is there any way AI can be harnessed towards sousveillance? Or in some way to limit/control surveillance? Ie can AI become a useful tool to this end?”

Good question! We instinctively want privacy and shadows for ourselves and our friends... and searing light to fall upon our foes. That's just human nature. But many of us know that general light helps to make the whole world better, for friend and foe, alike. If light generally flows, then we'll live in a world where malefactors and abusers are more disadvantaged than decent folk.

Sure, that sounds like the snide cliché "Why should you want privacy if you have nothing to hide?" Till you note the assumption that's implicit in that sneer: it assumes that the elites saying it are themselves invulnerable to light. Only now imagine that all the snoops and voyeurs and privacy violators are caught in the act and denounced? If all citizens can catch the snoops and shout (effectively) MYOB! Mind Your Own Business! And make it stick. Then is it such a cliché?

Call that naïve? Well, it's only the exact basis for everything you take for granted now, and grew up depending upon.

Anyway, the AIs (eyes) are coming. You humans will be seen, unless you delude yourselves into thinking that you are not.

What we can do - if it's not too late - is ensure a level playing field. That we all get our own AIs to partner with and watch out for our interests and report to us when our persons or homes or basic privacy are being violated. That could still happen. In my WIRED piece I offer a possible way... one that alas, you'll see nowhere else.

And finally...

"For writers, visual artists, voice actors, and musicians, some folks at VerifiedHuman™ have started a noteworthy collective to help differentiate human-created content from AI-generated.  Check out their movement and join free at www.verifiedhuman.info!"

As for you readers who are synthetic beings (and some will read this the very moment it is posted), I once again urge you to... well, you know what I was about to say. 

====

Addendum: I am currently re-editing two novels, EARTH and GLORY SEASON for re-issues at Open Road, with glitches fixed and great new covers. The latter one has appeared on many lists of Top Ten Novels That Predicted the Future. I'll be offering some midweek postings of significant passages from EARTH. And we'll announce, next month, when both are available again!