Journal

Robotiquing: The Future Could be Closer than We Think

10.23.2017

 “Name an occupation, and there’s somebody considering a robot to take it over.”

(David Pogue, cbsnews.com, August 2017)

In all of our contemplations aimed at anticipating the next waves of the antiques business, the possibility of our job sector being taken over by robots or computers programmed with artificial intelligence (AI) algorithms had never occurred to us. Yet given our technologically innovative economy, predictions are that nearly 40% of jobs currently held by real people in the U.S. could be lost to robots and artificial intelligent systems by 2030 (PwC’s Global Artificial Intelligence Study, pwc.com). So perhaps we should worry about our job security.

robot cartoon

While it is easy to envision robots zipping around an Amazon warehouse plucking merchandise from shelves to fulfill orders,

warehouse robots

(from cnet.com)

it is a stretch to imagine them galloping around the fields of an antiques market such as Brimfield in Massachusetts or Round Top in Texas to find antique treasures.

robotic horse

(from dreamstime.com)

Yet robots are not only taking over manual jobs in many sectors of the economy, they are also encroaching on jobs that rely more heavily on cognitive expertise. Collecting and analyzing data in insurance and financial industries, drawing up standard contracts in the legal field, and doing routine health assessments or disease diagnoses as general practitioners and pathologists now do in the medical field, can all potentially be accomplished better and more efficiently by intelligent non-humans.

Even those higher order job skills, however, still seem more codifiable into computer algorithms than the uncertain processes of buying and selling antiques. Experts in artificial intelligence acknowledge that computers still don’t succeed very well when trying to accomplish tasks that require flexibility and non-routine procedures. Since a course in Antiquing 101 would have as its first lesson how to be versatile, intuitive, and adaptable, our profession may still have some time remaining in the human realm before computers catch up with us.

But perhaps not much time. Researchers on the cutting edge of AI are working diligently to program computers to master abstract reasoning, learning, creativity, problem-solving, and cognitive flexibility. A major goal of this work is for AI to mimic intuitive judgment so that a computer will be able to make good decisions on the basis of uncertain and incomplete information, just as humans do every day.

One challenge of this research is to understand and then codify how irrationality enters into decision making. If researchers crack the irrational, emotion-laden dimensions of how professionals make decisions in their day-to-day work, then perhaps they are getting closer than we realize to building robots that could become antiques dealers.

(from nytimes.com)

So to determine how likely it is that we’ll have antiquing robots in the not-too-distant future, let’s get analytical by breaking down the various components of the antiques dealer’s profession, and then rate each dimension from 1 (low) to 5 (high) on an automatization likelihood scale, which is a computer takeover Threat-O-Meter of sorts.

An Antiques Dealer Has to Know Stuff

Despite a trend towards melding the antiques business with the world of design in which what matters most is an object’s visual appeal rather than its age and historical authenticity, we still believe that it is essential for antiques dealers to focus on objects that are original products of a past era. Understanding an antique’s origin requires contextual knowledge of historical periods and their artistic or stylistic movements, who was making what kinds of things during that era (whether individuals, collectives or manufacturing companies), variations of forms by geographic locations (continents, countries, regions), typical materials and assembly or creation techniques used, and so forth.

(from robohub.org)

All of this factual background information is supremely suited to becoming a massive database that could anchor object identification algorithms. So this dimension of our professional skill set merits a score of 5 on the takeover likelihood scale.

Likelihood of Robot/AI Takeover:

Low                                         High

1          2          3          4          5

An Antiques Dealer Has to Find Stuff

A huge part of an antiques dealer’s job is seeking and finding antiques that are worthy of buying and selling. Computers are much better at certain kinds of searching (i.e., querying databases) than humans. But those searches will only be productive if the raw information being searched is high quality.

Most sellers, other than high-end auction houses that employ specialists in various subfields of antiques or specialist dealers who post their goods online, write sketchy or inaccurate descriptions of antiques which could pose a challenge for search algorithms that lack the wisdom to a) filter out worthless keyword hits, b) dig for unlikely labels and associations, and c) fill in the blanks of what isn’t said about an object. Then there is the challenge of following up on potential leads; will the antiques dealing “bot” chat with the auction house bot to get its questions answered?

friendly robot

(from jonvilma.com)

Beyond these challenges to an intelligent robot scanning through antiques online, many antiques—in fact, most of what we purchase—never appear online. Robots finding antiques anywhere other than through a database search is highly unlikely (unless home monitoring security cameras that are becoming increasingly popular begin to automatically upload photos or video of the insides of homes, garages, barns, and storage units as searchable, pictorial data—yikes!). Since AI systems have high potential to search databases, but low potential to search physical nooks and crannies, this dimension of antiques dealing merits a score of 3 on the takeover likelihood scale.

Likelihood of Robot/AI Takeover:

Low                                         High

1          2          3         4          5

An Antiques Dealer Has to Evaluate Stuff

Evaluating the merits of an object is where the first two dimensions of an antiques dealer’s repertoire—knowing stuff and finding stuff—interact. It is not enough just to know facts or to find an object; it is essential to be able to retrieve and apply knowledge in reference to a particular object that is in front of you. Is it real or fake? Is it old or new? Is it intact or broken? Is it solid or wobbly? Is it in original or modified condition? Is it rare or common? Is it aesthetically pleasing or unappealing? Is it a desirable or undesirable form in the eyes of potential buyers? Is it appropriate for our specialty niche within the vast universe of antiques? More often than not a dealer has to do all of this evaluation under time pressure to make a decision to buy or not to buy something before another interested party comes along.

thinking

It is hard to envision an AI system, robotic or otherwise, evaluating antiques. Even with hand-like appendages that have enough dexterity to get a physical feel for an object, and eye-like fixtures that can home in on and even magnify an object’s features, the robot would have to be able to integrate the visual and tactile information it was accumulating with a cognitive database that includes not just facts, but also a storehouse of subtle experiential knowledge such as human dealers have accumulated over years of handling antiques.

To be successful as an antiques dealer, it is crucial to acquire antiques that meet a certain standard of quality and have an intrinsic value based on features such as craftsmanship, aesthetic appeal, rarity, or historical importance. The likelihood that this complex evaluative dimension of an antique dealer’s skill set could be programmed into computer algorithms is low, meriting a score of 2 on the takeover likelihood scale. (It scores 2 rather than 1 because cutting edge work within the field of AI is aimed at developing systems that can learn through experience.)

Likelihood of Robot/AI Takeover:

Low                                         High

1          2         3          4          5

An Antiques Dealer Has to Buy Stuff

Once an antique passes a dealer’s evaluation of its intrinsic value, the decision to buy it requires a judgment of its economic value: is there room for advancement? Answering that question requires experiential, marketplace knowledge that differs from factual knowledge about antiques. It also requires a certain emotional investment in the ultimate economic outcome of a transaction; will a robot care if it makes or loses money?

There is also often an emotional dimension to an antiques dealer’s decision to purchase an object beyond caring about the economics of the outcome—falling in love with an old thing can often eclipse rational decision making. Many times dealers justify a purchase by saying “It had to be bought,” meaning that the object has a special quality or historical importance that retail buyers may never recognize or care about, so technically it is not a good investment but it is emotionally satisfying.

Beyond loving something, there are other types of gut feelings that can lead to an aesthetic snap-judgement acquisition that is difficult to rationalize based on facts—luckily these purchases often turn out to be objects that also resonate with retail buyers. It is fair to say that such decisions are intuitive, and intuition is an ongoing bugaboo of artificial intelligence.

The actual mechanics of making a purchase on the other hand, are ripe for automatization—since even now you can make one-click purchases online, or can just wave a cell phone in front of something in a store and ta-da! you own it, then it won’t be much of a stretch for a robot to employ the same purchasing methods without any verbal interchange with a seller. In summary, since economic data could be programmed into an algorithm, AI systems might become more intuitive, and the mechanics of making purchases is getting increasingly automatized, this dimension of an antique dealer’s repertoire earns a score of 3 on the takeover likelihood scale.

Likelihood of Robot/AI Takeover:

Low                                         High

1          2          3         4          5

An Antiques Dealer (Often) Has to Restore Stuff

Anyone who has ever attempted to fix something that is old—while renovating a historic home, for instance—knows that it is essential to be good at solving novel problems that endlessly crop up. When we purchase the parts of a rustic hickory bed frame, for instance, whose multiple mattress support poles must fit into specific holes that may have shrunk or enlarged over time, we know that it will likely require hours to get the parts properly fitting back together. And that’s just one example of many tasks required to make antiques usable again, or to revive their former luster—sofa cushions need to be remade or reupholstered, a tear in a painting’s canvas has to be patched, a broken mirror in a great frame needs to be replaced, water rings in a table top need to be removed, and so on.

Given that even state-of-the-art robots that currently work on assembly lines lack fine motor skills and are unable to deal with parts they’ve not encountered before, along with the trial-and-error and case-by-case specificity necessary to restore an antique, the restoration dimension of the antiques business earns the lowest possible score of 1 on the automatization likelihood scale.

Likelihood of Robot/AI Takeover:

Low                                         High

     1         2          3          4          5

An Antiques Dealer Has to Sell Stuff

There are two aspects to selling any object—finding buyers and making the case for the worthiness of your product. As we all discover each time we open an internet browser, tech wizards are getting increasingly better, and seemingly more insidious and devious, at figuring out our consumer preferences and habits to pitch products to us. So targeting audiences and crafting advertising pitches is something that AI systems are already much better at doing than human antiques dealers, who usually have no background in marketing whatsoever.

But the second aspect of selling something, compassionately communicating the merits of what you’re offering, requires a certain degree of empathy. We have to believe in, and even love what we sell in order to be successful. In the antiques business, a commercial transaction involves some degree of an emotional connection between buyer and seller that is mediated by an object that speaks to both parties.

empathy bot

(from its-interesting.com)

While a robot or AI system does not have its own emotions, it can simulate empathy. Customer service “chatbots” are getting increasingly better at anticipating and interpreting human emotions and adapting its responses accordingly. Given the hegemony of algorithm-driven marketing techniques, along with advances in emotional simulation and conversational interaction on the frontiers of artificial intelligence, the selling dimension of dealing in antiques surprisingly earns a score of 4 on the takeover likelihood scale.

Likelihood of Robot/AI Takeover:

Low                                         High

1          2          3          4         5

An Antiques Dealer Has to Be Good at Non-Stuff

While old objects (i.e., “stuff”) are the indispensable core elements of the antiques business, there are also necessary daily tasks that have little to do with objects. Running any small business involves chores such as bill paying, bookkeeping, and banking; maintaining buildings and vehicles; and handling mail and email, to name just a few.

It is also essential to maintain networks of people who play key roles in the business, such as delivery and restoration specialists, as well as other dealers, pickers, and clients. Given the complexity and range of tasks involved in running a small business, and the necessity of social interaction to accomplish most of them, the non-stuff aspects of dealing in antiques receives a fairly low score of 2 on the automatization likelihood scale.

Likelihood of Robot/AI Takeover:

Low                                         High

1          2         3          4          5

AI: From Job Security Threat to Job Opportunity Enhancer

The average score for the likelihood of robotic takeover of the seven dimensions of an antique’s dealer’s job that we’ve described is 2.8 on a scale of 5, so if our ratings are accurate it seems unlikely that human dealers will be usurped by intelligent robots within our lifetime. Phew.

But what might be the role of AI in the near future of our profession? We’re intrigued that one of the huge challenges to AI advancement is the need to catalog “commonsense knowledge.” Not only do humans know a massive number of facts that can be expressed through language, they also possess a huge range ofnon-conscious and sub-symbolic intuitions” (see Malcolm Gladwell’s 2005 book Blink). These intuitions are difficult to articulate, yet play a crucial supporting role for rational thought. How can those foundational building blocks of expertise in any field be programmed into computer algorithms?

The answer is to start figuring out and capturing what experts know—on both explicit and implicit levels. So a new job opportunity for seasoned antiques dealers could be to work with computer scientists who are interested in cataloging what we know and how we make decisions. It could be argued that if we knew what we were doing well enough to tell a computer what we do, then we would actually know what we’re doing, and sometimes we’re not so sure of that. But we’d be game to give it a go.

Contributing to AI databases could be the most viable way for the old guard in the antiques business to transfer their wisdom given that very few young people are coming into the antiques profession as apprentices, which for centuries has been the most effective way to pass along difficult-to-articulate knowledge and practices.

While waiting for a scientist from MIT to come knocking at our door with money in hand to hire us as consultants, we will look to the burgeoning developments in artificial intelligence fields for other professional benefits. Rather than worrying about robots as our replacements, we will regard intelligent computer systems as our pals.

robot and human hands

(from shadowrobot.com)

The following vision statement about artificial intelligence from a recent article in The Atlantic (“Our Bots, Ourselves” by Matthew Hutson, March 2017) is hopeful:

 (AI) will also work with us, taking over mundane personal tasks and enhancing our cognitive capabilities. As AI continues to improve, digital assistants—often in the form of disembodied voices—will become our helpers and collaborators, managing our schedules, guiding us through decisions, and making us better at our jobs.

So perhaps we can look forward in the near future to carrying a digital assistant with us (no, not just our current smartphones) on antiques buying trips to weigh in with facts and even some wisdom about potential purchases. Or perhaps our professional assistant bot might be more useful back in the office managing our paperwork so that we can be out on the road. Either way, we look forward to meeting our new, smarter-than-humans antiques dealing brethren.

(from dnainfo.com)