Issue 123 - December 2nd 2021
This issue of That Space Cadet Glow talks about the failure of Zillow's house valuation algorithm, AI-generated artwork making millions, and the best ever use of robots. Image credit Ryan Nagata
Death of an Algorithm
It is not often that you hear of AI algorithms failing in the market, but that is essentially what happened in the case of real estate company Zillow and its house-buying estimator algorithm, Zestimate. The headline writers and tweeters were quick to feed off the wounded animal, asking whether this was the beginning of the end for AI (typical tweet: "A significant strategic blunder! Everything in our world cannot be managed with algorithms parsing big data! They failed to think through the implications of buying all those homes. A perfect HBS case study on bonehead! @zillow heads should roll!"). But, like the informed readers we are, we need to look behind the headlines to see exactly what happened with Zillow. The algorithm was part of a business that bought and sold (flipped) houses in the US using price estimates provided by the AI model. It looked for properties that were under-valued, bought them, did them up, and sold them off at higher price. At least that's what was meant to happen. What did it for the algorithm was the massive uncertainty that entered the market with the pandemic. Back in March 2020, I wrote about how the huge skewing of trend data that the pandemic brought will play havoc with predictive algorithms, and this is a prime example of that. The predictions simply become less accurate. But there is also another line to this story, because blaming the algorithm (which, remember, is just some clever maths) is the easy thing to do. What is not often considered is the role the humans played in all of this (which is, of course, everything). Another tweet that asks the right sort of questions was from a property entrepreneur: "It would be fascinating to find out where in the stack Zillow's failure lives. Was it incorrect use of ML? Too much trust in ML? Aggressive management that wouldn't take 'we aren't ready' for an answer? Wrong KPIs?". Notice how none of these questions are blaming the algorithm? A piece from Wired manages to take a deeper dive and concludes that it was probably parts of all of these things. For example, the heart of the buying activity was in Phoenix, where there are plenty of 'cookie cutter' homes that are easier to price. Once you try and take that model to New York then many of the assumptions and training data are no longer relevant - the variability in the types of homes makes predicting much, much harder, before you then introduce pandemic impacts. It also seems that Zillow expected the algorithm to do everything, including estimating the amount of capital required for the business, when all it did was estimate prices of individual houses. So there should probably be a Harvard Business School case study made out of this, but the lessons are clearly about how to use the algorithm in the right way - understand what it can and cannot do, model extreme events to know its limitations, don't try to extend it to areas where it will not perform, and, most importantly, never trust it too much.
Scraping the Botto of the barrel
If you are reading this on the 2nd December, then today I will be presenting a keynote talk to artists and art market professionals on behalf of DACS. We will be looking at the role of AI in art, particularly where AI-generated art has been derived from existing works (this is in response to the UK Government's consultation on AI and Intellectual Property). One of the examples I will be talking about (as well as the now-infamous Edmond de Belamy portrait) is Botto, an AI system that creates artworks (and their titles) that are then sold as Non-Fungible Tokens (NFTs). Apparently, the works (like the one shown here) have netted more than $1m so far. Botto is very much a team effort, bringing in a number of different algorithms as well as real-life people: the process starts with a random string of words that are fed into a Generative Adversarial Network called VQGON that creates an image. Another algorithm called CLIP then judges whether the image reflects the words well enough - if not then VQGON tweaks its image until CLIP is satisfied. CLIP also creates a two-word title for the image that it believes resonates with the image. Then a third algorithm, GPT-3 (which I've mentioned in this newsletter a few times before) creates an 'abstract poetic' description of the image. Amongst all of this algorithmic fervour are some humans which help judge the artistic merit of the images (short-listing the 2,000 weekly images down to 350) and tweak the descriptions, mainly to remove any obscenities (GPT-3 was trained on the whole internet so includes more than enough rude words for anyone's liking). And by 'listening' to the human feedback the algorithms are able to do a better job next time around. The Botto system highlights a number of challenges from AI that are facing the art world, none more so than the use of other artwork (some of which may be copyrighted) to derive the new pieces. One could argue that all art is effectively derivative, but Botto is able to do this at scale and with relative ease. We do not get any sense of the effort or the motivations behind the Botto art (because there really is none) which is surely a major element of any work of art. If Botto's role is to generate some nice images then fair enough, but let's not call it art.
Working from home
Most articles about robots show a shiny machine with a smiley face trying very hard to provide some trivial service to a human. Most examples end up just being marketing hype and novelties that have little practical use. But I did come across an example where shiny robots with smiley faces were actually doing something really, really valuable. A restaurant in Tokyo called DAWN uses the robots to serve customers, but the amazing part is that the robots are controlled by disabled people who are largely bedridden or house-bound. The cafe gives opportunities, employment and a sense of purpose to people who would normally struggle to find work. Through the robots they can interact with the customers and, importantly, feel that they are contributing to society and the economy. After some 'pop-up' trials, the initial restaurant site is now permanent. This is such a great model on so many levels - imagine the same idea being applied to other industries including manufacturing and retail. At last, a really good use for all those shiny, smiley robots.
Curtis Harding - If Words Were Flowers
There is something about Curtis Harding that makes him stand out above all of the other soul singers right now. His music is so much more than just retro-rehashing that it is able to stand by itself, sounding contemporary and fresh (and he doesn't particularly like the genre-packaging). It certainly helps that he has style that oozes from every corner (and is actually a Gucci model in his spare time). Each one of the tracks on this LP is a winner - luscious, smooth, velvety. The video for 'Can't Hide It' brings humour to the retro feel, with a guest appearance from Anthony 'Captain America' Mackie in a faux music show called, appropriately, the Velvet Touch. You can listen to the whole LP on Spotify and Apple Music.
AJBurgess Ltd
ajburgess.com
thegreenhouse.ai
Andrew Burgess is the founder of Greenhouse Intelligence, a strategic AI advisory firm.