Monumental AI advances, I couldn’t resist writing a blog on this topic, partially because it is indicative of the positives and negatives in the process of Artificial Intelligence development, partially because it made me laugh, and partially because I have been a victim.
Hiccups in the development of new technology are inevitable, and artificial intelligence is such a radical step forward (we all hope) that it is bound to have more than its fair share of issues. The example in this blog is petty, serious, funny and a good example of that process.
Many of you will possess, or certainly will have seen or heard about, the small round robot that cleans the floor of your home, your office and other larger locations. It seems to work well, and most people report they are very happy with it. However, there have been some glitches. It doesn’t handle obstacles like wires very well, and gets confused when encountering down staircases and a few other things.
One of the companies that makes these robots has just upgraded its capabilities in its new model……and that announcement was the driver for writing this blog.
The new Roomba robot vacuum cleaner is now programmed/trained to dodge pet feces. The latest model recognises and avoids cables and “pet waste”, the manufacturer says. Owners have complained that the devices run into, and then mindlessly spread, pet feces all over the house as they “clean”. Yes, I’m one of those victims, and it may be funny in retrospect but it certainly isn’t at the time.
“Quite honestly, we see this a lot,” a company representative told the Guardian Newspaper, in 2016. (And they are only just bringing out a new model that doesn’t do this!!)
Apparently, rival devices already claim that their products avoid pet mess. A reviewer was “fairly confident” that a new machine from Roborock, for example, would avoid excrement. (Fairly confident? What on earth does that mean?)
iRobot, which makes the new Roomba, promised it would replace any of the new machines, the Roomba j7+, that failed to avoid “solid” pet waste. (A bit late if you have feces spread all over your home, not to mention the disgusting effort necessary to clean the machine after its escapades – personal experience!)
The company said the algorithms controlling the robot had been trained on “hundreds of thousands of images of real, fake, and simulated dog and cat waste. (I’m not sure I want to visit their laboratories!!)
BBC News apparently has asked the manufacturer if its guarantee will apply to less-solid pet waste. In response an iRobot spokesperson said the promise, “does not apply to the non-avoidance of materials other than solid cat or dog waste”. (and they’ve spent 5 years developing this aspect of the robot’s capabilities!!)
We have a major problem here. Understandable, perhaps, that the development companies would not have anticipated this dilemma but you would think there was some urgency to fix it now they do know – apparently not!
The story I read had a followup, which I also found intriguing. The development of a robot lawn mower that avoids hedgehog corpses. Now, who would have thought of that as a requirement for lawn mower development?
Dr Sophie Lund Rasmussen, of the Wildlife Conservation Research Unit, at the University of Oxford, is working with robot-lawn-mower manufacturers to minimise their risk to hedgehogs (the lawn mowers, I assume, not the Wild Life Conservation Research Unit!). She told BBC News her research, using hedgehog corpses collected from local rescue centres, showed “the sensor technologies used in the 18 different models tested were not able to detect the hedgehogs at a distance” – though some robots had stopped or changed direction when in contact with the corpses. (The mind boggles……how about avoidance of small children?).
In conclusion, the report said that the challenges facing lawn-mowers and vacuum cleaners were not directly comparable “as they serve different purposes and may not use the exact same technology”. (Well! Thank goodness for that).
It’s amazing what goes on in research labs without our knowledge!
What is concerning about all this is that this small, and somewhat amusing, example of AI development could easily be replicated in much larger, and more dangerous, AI projects.
Food for thought, as we embrace the concept of AI more and more.
