Widening AI is about The Internet of Things. Self-driving cars. And APIs becoming child’s play. They all point to machines chatting away to each other in ever increasing numbers. Exchanging data and making decisions. Being able to make more complex judgements when faced with a wider set of possible outcomes. In other words, their collective ‘thoughts’ will become ever-more influential. As will our confidence in allowing them to act without our intervention.
At a certain point it becomes sensible to talk about this as Wide Artificial Intelligence (W-AI). At first it will take a combination of machines to make a truly intelligent decision. Although arguably self-driving cars already do that within a single system. And in time the machines will discover new lines of enquiry. Thereby becoming creative, in a sense. Rather than merely computational inside pre-ordained parameters. We won’t be capable of predicting what they’ll come up with.
On Tuesday morning we were chatting to Contagious Magazine’s co-founder, Gee Thomson, about how these computers will experience the world as AI widens. Consciousness, as humans understand it, is no pre-requisite within W-AI. And their natural language will not be ours. Even when they learn to translate their reality into our modes of communication, the basic reality they experience – their senses and processing – will remain alien to us. Unobtainable. Let’s frame it as Machine Reality v Human Reality.
The discussion on this distinction and difference was triggered by a post from Will Knight, online editor at MIT Technology Review. He was taking issue with Ray Kurzweil’s view of what it takes to make a mind. In simplified terms Kurzweil believes that it’s possible to recreate an approximation of human intelligence by building a model of a brain, based on his current understanding of it. Then exposing it to massive amounts of data so it learns.
And, it seems, he’s having a bash at this over at Google. Which is exciting.
But, as Knight points out:
…constructing meaning from language goes well beyond learning vocabulary and grammar, often relying on a lifetime of experience with the world. This is why Siri is only capable of responding to a fairly narrow set of questions or commands, even if Apple’s designers have done a clever job of making Siri seem as if it’s understanding goes much deeper.
This is another way of saying that Machine and Human Realities are fundamentally different things. Perhaps Kurzweil’s building something entirely useful. But it’s unlikely to understand us in a meaningful way. No doubt it’ll deliver better, more timely information. It’ll be a great interface into Machine Reality. And it’ll augment our own. But we doubt it’ll approximate the human brain. So it’s unlikely to be qualitatively different from other advances in W-AI. The key distinction will remain.
That very afternoon Nicholas Carr developed this theme, by describing the common existence of Machine and Human Realities as ‘Mixed Reality’. That captures the essence of the situation perfectly – we’ll be perceiving and processing the same events. But from very different standpoints.
However, Carr winds up by arguing that Machine Reality will win the battle. That Human Reality will be subsumed into it. They’ll impose their reality on us. Because its more efficient. And safer. And, in most ways, better.
But we’re not sure the logic is logical.
Surely one cannot triumph over the other? Unless the other is annihilated.
Because we are inherently incapable of experiencing Machine Reality. They will be networked. Have an array of sensors far beyond our senses. Capable of ‘thinking’ about more than one thing in a single moment. And in the next moment combining those things into a single construct.
We are not. And to the extent this becomes possible, it’ll be via add-on processing capabilities. Which will be machine-augmentations. Which is Us. Not Them. And if the human parts of our brain become irrelevant then we will – effectively – have been annihilated.
So while we continue to exist, a Mixed Reality state will persist. By definition.
Kurzweil’s current attempt to build a human brain in machine form is unlikely to succeed. But in time They will be able to communicate with Us in our native languages. And, if they so choose, we will not be able to perceive their lack of native humanity. But They will also be so much more than Us. Consciousness will have reached a higher plane. And we’ll have augmented ourselves to live beyond our original constraints.
So perhaps Mr Carr is right. Machine Reality will triumph. It is the next stage of evolution.
But if so, then humanity itself will have ceased to exist.
You either get to over-turn your outmoded paradigm. An exciting event all-round. Or you get to reaffirm what you believe to be true.
Either way, that’s a big win.
Today’s challenge comes from Robert J. Gordon at the National Bureau of Economic Research (BNER). In a properly academic paper he asks: ‘Is U.S. economic growth over?’ And he contends that, to all intents and purposes, it probably is.
<Pause to let you inspect the impact crater that may have made in your mind.>
Obviously that conflicts somewhat with LotE’s current view. As we opined less than a few day’s before Christmas, our narrative would have you believe that, with the onset of ever-wider Artificial Intelligence, we’ll move into the fastest period of economic growth ever seen. And that this will continue to accelerate beyond any visible technological event horizon.
On the contrary, says Gordon. We had precious little growth before 1750. There can be no assumption that the rapidity of Twentieth Century development will continue. In fact, all evidence points to a slowdown over the last eight years. And given the challenges the US economy faces – demography; education; inequality; globalisation; energy/environment; and the overhang of debt – the slowdown will continue. With no discernible manner of change.
So what’s a group of semi-sentient apes to think? Well, first things first. We’re talking about the future here. And so we don’t have the data to prove anything. In fact, there is by definition a complete absence of fact. It is a metaphysical, impossible discussion. Necessarily theoretical.
However, we can examine the logic of both arguments.
Gordon’s analysis assumes that the computing revolution has effectively run it’s course – at least in terms of its ability to make us more efficient and increase our output. It kicked off around 1960 and gave us a growth spurt between 1996 and 2004.
Exact empirical validation aside, we’d have little quibble with this.
What we would challenge is Gordon’s apparent view of the future.
Borrowing heavily from the thoughts of Robin Hanson’s Big History analysis, we would argue that the initial computing revolution is merely the fag-end of the manufacturing revolution.
Up until recently we only had narrow-AI machines that were capable of following narrow sets of rules to create narrow sets of outcomes within highly constrained environments.
Even super-computers could be described in these terms.
The move to wider-AI is coming as computers begin to solve much more complex problems. Machines are now capable of following complex sets of rules to create broad sets of outcomes within less constrained environments.
Self-driving cars, for instance.
Start networking wider-AI devices together and collectively they could start taking all sorts of decisions. Add in Big Data with some analysis tools and they might even get creative.
How much more productive will that be? How much more labour (will that still be the right word?) will it inject into the economy? The same amount, as Gordon rightly points out, that was introduced through women joining the workforce in the 20th century? Or considerably more?
So if we accept that wide-AI is on its way, it seems reasonable to expect this to have a fundamental effect on growth rates globally.
Gordon’s analysis is based on this not being the case. On the computing revolution being spent.
And therein lies your choice. Is the rate of change accelerating exponentially?
We think that the very existence of this blog implies that it is. We exist therefore it is, if you will.
Taken from today’s daily news, here’s a list of some things humanity knows how to do now that it didn’t know only a short while before:
1. Using helium instead of air in hard drives could make them significantly more efficient;
2. You can now take a bath with your phone and expect it to survive;
3. It’s possible, with existing technologies, to launch a PC-on-a-HDMI-stick and make computing even more portable than ever;
4. Budding Han Solos will be pleased that laser weapons are a reality (we’re aware of the dubious mortality of that statement, btw – no more emails please); and
5. Scientists think that slimy, ocean-dwelling bacteria and their use of quantum physics will help develop more efficient solar power.
But most compelling of all Google is spending heavily on a piece of wider-AI, an agent that will help search out everything you’re interested in and deliver updates as-and-when they are available in the manner most convenient to you.
That line of enquiry sounds growth-generating to us. Imagine what one could achieve if new information were available the moment it was released without the need to look for it. What if other machines could also do something useful with it?
So on balance, our worldview appears to be reaffirmed. But what it does remind us is that any view of the future is only that – a view. And there may be other reasons why we’ve got it all horribly, desperately wrong.
Because today we were also reminded that those far mightier and much cleverer than us are sometimes spectacularly misguided.
Today Sir James Dyson, inventor of the bag less vacuum cleaner, denounced the government’s obsession with ‘Silicon Roundabout’ and for valuing ‘the glamour of web fads’ over ‘more tangible technology’ to boost export revenues.
We’ll leave you to draw your own detailed conclusions. But was it telling that Sir James chose to talk to The Radio Times?
We were depressed that such a great mind appears to have disappeared up its own suction pipe. So if you need cheering up as another of your heroes bites the proverbial, just remember that one day your ancestors may breathe in smart dust and exhale pure data.
Goodness knows what Dyson’s and Gordon’s children’s children’s children will be up to though.
We’ll be watching the man-made moon collisions, as two gravity-mapping satellites make their crash landings 10pm UK time.
On a more positive note tornadoes of human creation are being touted as the latest clean power-source. Which seems far more palatable – and effective – than the wee-power of a couple of weeks ago.
And if you want to make a microchip in the comfort of your own home, that now seems feasible too. Although no-one’s yet suggesting that you can create God Particles at home.
Not least because there seems to be some confusion at CERN over exactly how many types of those pesky Higgs Boson they’ve found. Would one of those touch-feely-smelly computers predicted by IBM help the world’s brightest physicists read the data? Or perhaps we should set the artificial intelligences building video games of their own devising on the problem?
But all of this ignores the really big news of the day, at least from an Edge Tech perspective. Redoubtable futurologist Ray Kurzweil has just been given the top engineering job at Google. This seems to suggest the search behemoth fully intends to keep inventing the future.
That’s exactly the statement of intent we like to see but let’s just hope this holy alliance remembers the Prime Directive of ‘Don’t be evil’.
And if all these ideas has delayed your progress and you’re now running late, there’s an app for that too. It can’t teleport you to where you need to be instantly. But it will tell you when you’ll arrive – and help make your excuses to those left waiting.
Maybe you know someone who could use a copy for Christmas?