George Rebane
The Singularity must really be near when we have MIT physicist Max Tegmark climb aboard with the spate of science and industry luminaries who have recently discovered that intelligent machines will someday surpass humans, will systemically displace human labor, and (in a TBD form) will become the dominant lifeform on Earth. Dr Tegmark has just published his epiphany in Life 3.0, and joins Stephen Hawking, Bill Gates, Elon Musk, Nick Bostrom, … in spreading the news of the incipient Singularity to the general public. Of course, none of them dare call it Singularity; for if they did, then it would instantly identify them as Johnny-come-latelies.
For those of us in the field, as recorded herein, the possibility of Singularity has been known for at least thirty years (and more decades in science fiction). And science savvy entrepreneurs like Ray Kurzweil (now at Google) and Jeff Hawkins (founder of Numenta the developer of hierarchical temporal memory) have long staked out businesses that are pioneering the only branch of computer science that promises to deliver the Singularity – systems that are capable of learning from vast amounts of data present in their ‘environments’. Here I don’t want to revisit the ‘learn vs program’ debate on the path to super-intelligence other than to say that the ‘program school’ of computer science has woefully been the dragged anchor, diverting resources from the ‘learn school’ which now finally dominates the field of AI. All of the above people and technologies have been introduced and discussed here over the years (see RR’s Singularity Signposts section).
Deep learning is the label given to the latest generation of massive artificial neural networks (ANNs) that have learned to do amazing feats requiring intelligence and cognitive processing. Most of these feats already far surpass humans’ ability to do the same tasks. And what is happening today in the field of deep learning will literally blow your mind, in the sense of blowing it away as redundant, in the event that you decide to compete with it in the workplace – this brings to mind the story of John Henry, the steel drivin’ man (here).
Companies large – e.g. Alphabet/Google, Microsoft, Amazon, Facebook, Baidu, Alibaba, … – and small are racing to integrate deep learned AI into their operations and products at breakneck speed. The heads-up companies are launching in-house education and training programs to introduce deep learning to their technical and management staffs. As an example, I can modestly cite my computer scientist son-in-law Roland Fernandez at Microsoft Research who has co-developed and operates that company’s online course in deep learning (q.v.).
What the sclerotic (‘programming school’) side of computer science did not realize for some decades, with some still in their benighted darkness, is the message Dr Kurzweil has been telling (preaching?) to people for years – that technology is advancing exponentially. The programmers could not conceive that very soon there would emerge computers and databases that are large and fast enough to implement ANNs with thousands of layers that learn to ‘instantly’ manipulate millions of parameters in tasks like voice understanding, image recognition, medical diagnosis, concurrent large cohort control, and on and on. And that’s just today.
For more on Professor Tegmark’s new book, I bid you read the excellent review ‘When Machines Run Amok’ by Frank Rose.
rebanesruminations.com 
Leave a comment