Right now, none of us can predict what the impacts of AI will be and, rather concerningly, what mix will occur between the benefits and costs.  One can only hope that the potential dangers are contained.

For auctioneers such as Manshed Auctions, Australia’s premier auction house for railway memorabilia, AI is now bringing very direct impacts.

Positively, consignors and customers can undertake very rapid research into the past prices achieved for auctioned items as well as information about the items they have and are thinking about consigning.

There are downsides, though.  Research about the recent sales of a locomotive builder’s plate, such as one from Sydney’s Clyde Engineering, may not be particularly helpful since several variables will affect the price achieved at auction.  The particular locomotive from which the plate was derived will have a significant impact on price.  The level of supply of the plates (rarity) will affect the market, as will the level of interest from the relatively small pool of potential buyers at any particular auction.  Who “attends” is incredibly important.

What is becoming very apparent is that AI cannot control for these variables.  It can only use existing information and previous sales to make its conclusions.  If an item is a “one off” rare item, the more inaccurate AI is likely to be.  We experienced this a few days ago in relation to a customer enquiry about a steamer trunk which a potential consignor is looking to sell.  The following two pictures highlight the problems here.  AI has decided that having a person’s name and address on the trunk are a massive contributor to its crazily-high valuation of $5,000 to $7,000!  The trunk appears to be from France and has travelled to Australia from Paris’s Opera building.  There appears to be no indicators of “brand” in the photo and more extensive research tends to indicate that it is not dissimilar to trunks which sell on eBay or Etsy and elsewhere for a few hundred dollars.

The result?  A disappointed (or deluded) trunk consignor/seller and the creation of unrealistic expectations for all.  For an auction house, it means drilling down by having to check AI’s conclusion against market realities.  As I have written before, to the extent that there is a gap between consignor expectations and market reality, the more likely the whole thing will end in frustration and tears.  The greater the gap, the greater the disappointment.

So, for common items (say signed memorabilia part of a large edition of the particular item such as the thousands of AFL/NRL jerseys released by the licensee after each Premiership), AI is certainly likely to assist.  For rarer items, it may provide wildly inaccurate results.  Such items require more homework and wider research.