The Futurist

ATOM Award of the Month, October 2018

For this month's ATOM AotM, we visit the medical industry, and examine a technology that seems quite intuitive, but on account of patents and other obstacles, has seen rapid improvement greatly delayed until now. 

Surgery seems as though robotics would be ideally suited for it, since it combines complexity and precision with a great deal of repetition of well-established steps.  The value of smaller incisions, fewer instances of bones being sawed, etc. is indisputable, from qualitative measures such as healing pain, to tangible economic metrics such as hospital stay duration post-surgery.  

Intuitive Surgical released its Da Vinci robot to the market in 2001, but on account of Intuitive's patents, they sustained a monopoly and did not improve the product much over the subsequent 17 years.  Under ATOM principles, this is a highly objectionable practice, even if technically they can still earn a high profit margin without any product redesigns.  As a result, only 4000 such robots are currently in use, mostly in the US.  Intuitive has achieved a market capitalization of over $60 Billion, so it has succeeded as a business, but this may soon change.  Now that Intuitive's patents are finally close to expiry, a number of competitors are ready to introduce ATOM-consistent exponential improvements into the competitive landscape.  

The Economist has a detailed article about the new entrants into this market, and the innovations they have created.  In addition to mere cost-reduction due to smaller electronics, one obvious extension of the robotic surgery model is for each robot to be connected to the cloud, where the record of each surgery trains an Artificial Intelligence to ensure ever-improving automation for several steps of the surgery.  With AI, greater usage makes it improve, and when thousands of surgeries around the world are all recorded, that makes each machine simultaneously better.  As costs lower and unit volume increases, the volume of data generated rises.  As the accumulation of data rises, the valuation of companies capturing this data also rises, as we have seen in most other areas of technology.  

This level of data combined with greater circuitry within the robot itself can also increase the speed of surgery.  When more of it is automated, and the surgeon is doing less of the direct manipulation, then what is to prevent surgeries from being done at twice or thrice the speed?  This enables a much shorter duration of anesthesia, and hence fewer complications from it.  

 

 

ATOM Award of the Month, September 2017

For September 2017, the ATOM AotM takes a very visual turn.  With some aspects of the ATOM, seeing is believing.    

Before photography, the only image capture was through sketches and paintings.  This was time-consuming, and well under 1% were prosperous enough to have even a single hand-painted portrait of themselves.  For most people, after they died, their families had only memories via which to imagine faces.  If portraits were this scarce, other images were even scarcer.  When image capture was this scarce, people certainly had no chance of seeing places, things, or creatures from far away.  It was impossible to know much about the broader world.    

The very first photograph was taken as far back as 1826, and black&white was the dominant form of the medium for over 135 years.  That it took so long for b&w to transition to color may seem quite surprising, but the virtually non-existent ATOM during this period is consistent with this glacial rate of progress.  The high cost of cameras meant that the number of photographs taken in the first 100 years of photography (1826-1926) was still an extremely small.  Eventually, the progression to color film seemed to be a 'completion' of the technological progression in the minds of most people.  What more could happen after that?  

But the ATOM was just getting started, and it caught up with photography around the turn of the century with relatively little fanfare, even though it was notable that film-based photography and the hassles associated with it were removed from the consumer experience.  The cost of film was suddenly zero, as was the transit time and cost from the development center.  Now, everyone could have thousands of photos, and send those over email endlessly.  Yet, standalone cameras still cost $200 as of 2003, and were too large to be carried around everywhere at all times.  

As the ATOM progressed, digital cameras got smaller and cheaper, even as resolution continued to rise.  It was discovered that the human eye does in fact adapt to higher resolution, and finds previously acceptable lower resolution unacceptable after adapting to higher resolution.  Technology hence forces higher visual acuity and the associated growth of the brain's visual cortex.  

With the rise of the cellular phone, the ATOM enabled more and more formerly discrete devices to be assimilated into the phone, and the camera was one of the earliest and most obvious candidates.  The diffusion of this was very rapid, as we can see from the image that contrasts the 2005 vs. 2013 Papal inaugurations in Vatican City.  Before long, the cost of an integrated camera trended towards zero, to the extent that there is no mobile device that does not have one.  As a result, 2 billion people have digital cameras with them at all times, and stand ready to photograph just about anything they think is important.  Suddenly, there are countless cameras at every scene.  

But lest you think the ubiquity of digital cameras is the end of the story, you are making the same mistake as those who thought color photography on film in 1968 was the end of the road.  Remember that the ATOM is never truly done, even after the cost of a technology approaches zero.  Digital imaging itself is just the preview, for now we have it generating an ever-expanding pile of an even more valuable raw material : data.  

Images contain a large volume of data, particularly the data that associates things with each other (the eyes are to be above the nose, for example).  Data is one of the two fuels of Artificial Intelligence (the other being inexpensive parallel processing).  Despite over a decade of digital images being available on the Internet, only now are there enough of them for AI to draw extensive conclusions from them, and for Google's image search to be a major force in the refinement of Google's Search AI.  Most people don't even remember when Google added image search to its capabilities, but now it is hard to imagine life without it.  

Today, we have immediate access to image search that answers questions in the blink of an eye, and fosters even greater curiosity.  In a matter of seconds, you can look up images for mandrill teeth, the rings of Saturn, a transit of Venus over the Sun, the coast of Capri, or the jaws of Carcharocles Megalodon.  More searches lead to more precise recommendations, and more images continue to be added.  In the past, the accessibility of this information was so limited that the invaluable tangents of curiosity just never formed.  Hence, the creation of new knowledge speeds up.  The curious can more easily pull ahead of the incurious.  

Digital imaging is one of the primary transformations that built the Internet age, and is a core pillar of the impending ascent of AI.  For this reason, it receives the September 2017 ATOM AotM.    

 

Related ATOM Chapters :

3. Technological Disruption is Pervasive and Deepening

 

Recent TV Appearances for The ATOM

I have recently appeared on a couple of television programs.  The first was Reference Point with Dave Kocharhook, as a two-part Q&A about The ATOM.

The next one was FutureTalk TV with Martin Wasserman, that included a 10-minute Q&A about The ATOM.

Inch-by-inch, we will get there.  The world does not have to settle for our current substandard status quo.

As always, all media coverage is available here.  

 

 

Google Talk on the ATOM

Kartik Gada had a Google Talk about the ATOM :  

 

Artificial Intelligence and 3D Printing Market Growth

I came across some recent charts about the growth of these two unrelated sectors, one disrupting manufacturing, the other disrupting software of all types (click to enlarge).  On one hand, each chart commits the common error of portraying smooth parabola growth, with no range of outcomes in the event of a recession (which will surely happen well within the 8-year timelines portrayed, most likely as soon as 2017).  On the other hand, these charts provide reason to be excited about the speed of progress seen in these two highly disruptive technologies, which are core pillars of the ATOM.  

This sort of growth rate across two quite unrelated sectors, while present in many prior disruptions, is often not noticed by most people, including those working in these particular fields.   Remember, until recently, it took decades or even centuries to have disruptions of this scale, but now we see the same magnitude of transformation happen in mere years, and in many pockets of the economy.  This supports the case that all technological disruptions are interconnected and the aggregate size of all disruptions can be calculated, which is a core tenet of the ATOM.   

Related :

3.  Technological Disruption is Pervasive and Deepening 

 

Why Do Job Searches Take Longer Than 30 Years Ago?

I have recently come into contact with a few professionals in transition, many from the now-shrinking big semiconductor companies.  In speaking to them, one thing that stood out is how it takes them 9-12 months or more to secure a new position.  

Why is this the case, in an age of accelerating technological progress, as per the ATOM?  This is an instance of where culture has prevented the adoption of a solution that is technologically feasible.  

Where Cultural Inertia Obstructs Technology : Before the Internet age, if you wanted to research a subject, you had to go to the library, spend hours there, check out some books, and go back home.  Overall, this consumed half a day, and could only be conducted during the library's hours of operation.  If the books did not have all the information you needed, you had to repeat this process.  Even this was available only in the dozen or so countries that have good public libraries in the first place.  But now, in the Internet age, the same research can be conducted in mere minutes, from any location.  The precision of Google and other search engines continues to improve, and with deep learning, many improvements are self-propagating.  There is a 10x to 30x increase in the productivity of searching for information.  

If you feel that this example is imprecise, take the case of LinkedIn.  It has enabled many aspects of career research and networking that were just not possible before.  If a young person wishes to explore dozens of career paths and estimate common patterns, the utility of a certain degree, or the probability of reaching a certain title, LinkedIn has an endless supply of information and people you can identify and communicate with.  

Yet despite all of this, job searches are just as lengthy as in the days before the Internet, LinkedIn, and other resources.  If a candidate can match with three potential jobs in their search region at any given time, then the connection between employer and candidate should take mere weeks, not close to a year.  There is no other widespread transaction within society that takes anywhere near as long.  Despite new apps to organize the job search and new social media outlets that announce endless meetups and networking events, technology has clearly failed to generate any productivity gains in this process.  

For one thing, the Internet has reduced the marginal cost of an application to so little that each position receives hundreds of candidates, unlike three or four back when paper resumes had to be sent via the US Postal Service.  To cope with this, employers use software that searches resumes for keywords.  This method selects for certain types of resumes, with keyword optimization superceding more descriptive elements of the resume, and filtering out many suitable candidates in favor of those who know how to game the keyword algorithm.  

From this point, a desire to mitigate hiring risk combined with the lack of imagination inherent to most corporations defaults into a practice of increasing the number of interviewers that the candidate faces.  Three rounds and a dozen interviews is not uncommon, but by most accounts, job interviews are nearly useless as predictors of performance.  In reality, a candidate only needs to be interviewed by three people : the hiring manager, the manager above that, and one lateral peer.  If these three people cannot make an accurate assessment, adding several other interviewers is not going to add additional value.  Indeed, if the boss's boss cannot make accurate assessment of candidates, then they are failing at the primary skill that an executive is supposed to have. Reference checks are also a peculiar ritual, as a candidate will only submit favorably disposed references who have been contacted beforehand.  

Modernizing Hiring For the Information Age : Matching openings with candidates should not be so tedious in this age of search engines, emailed resumes, and LinkedIn.  Resistance to change and a miscalculation of risk and opportunity cost are the human obstacles standing athwart favorable evolution.  

To correct this obsolete situation, consider the mismanagement that occurs at the source.  Only after a hiring manager sees a persistent and pronounced need for additional personnel does the process of getting a requisition approved and advertised commence.  Hence, the job begins to receive resumes only several months after the need for a new hire arose.  After that point, the lengthy selection and interviewing process takes months more.  

Instead, what if the Data Analytics of a corporate setting could be gathered, mined, and processed, so that the AI identifies a cluster of gaps within the existing team, and identifies suitable candidates from LinkedIn?  Candidates with the correct skillset could be identified with a compatibility score such as '86% fit', '92% fit', and so on.  The entire process from the starting point of where a team begins to find itself understaffed to when a candidate deemed to be an acceptable fit is hired, can compress from over a year to mere weeks.  The hefty fees charged by recruiters vanish, and the shorter duration of unemployment reduces all the indirect costs of extended unemployment.  

For this level of dynamic assessment of gaps and subsequent candidate mapping, the capability of search and data analytics within a corporation has to evolve to a far more advanced state than presently exists.  Emails, performance reviews, and project schedules, etc. all have to be searchable across the same search and patterning capabilities.  Then, this has to interface with LinkedIn, which itself has to become far more advanced with the capability for a candidate to continuously re-verify skills and prove certain competencies (through tests, certified courses, etc.).  The platform has had no real improvement in capabilities in the last few years, and the obvious next step - generating a complex set of skill parameters for LinkedIn members, and matching that pattern to employers with similar needs, is quite overdue.  If this seems like added work for candidates, remember that this effort is far less than the amount of time and hassle it will save in the job search process.  

Of course, such a capability across LinkedIn and some pattern matching machine learning engine will not be adopted overnight.  After all, corporations still think university degrees and school rank are good indicators of candidate job performance, despite both evidence and common sense.  After that, the interface between some internal corporate software and LinkedIn will take a lot of work to become robust.  Finally, the belief that a greater number of interviews somehow reduces the risk of hiring a candidate is a belief that will be difficult to purge.  

But eventually, with technology companies leading the way, the massive hidden cost of current hiring practices may come to light, and give way to a system that uses AI to find more precise matches with much greater speed.  

Conclusion :  We now possess the machine learning capabilities to dynamically detect gaps within corporate teams and organizational structures that may be large enough to warrant an increase in headcount.  These gaps can be matched with parameters that can be mined from LinkedIn profiles, and provide candidates with an assessment of their approximate fit.  A percentage score calculated for each candidate is not only a more accurate indicator than the very imprecise interview process, but is far quicker as well.  It is high time that these tools were created by LinkedIn and others, and that corporate culture shifted towards their adoption.  

This application of AI is the second most necessary technological disruption that AI can deliver to our civilization at present.  For the first, check back for the next article.   

I do not have the time to pursue a company built around this type of machine learning product, but if someone else is inspired to take up this challenge, I would certainly like to be on your board of directors.  

 

Related ATOM Chapters : 

11. Implementation of the ATOM age for Individuals