Can A Goldfish Drive a Car? Yes! But How? 🐠

Dear Fellow Scholars, this is Two Minute  Papers with Dr. Károly Zsolnai-Fehér. Today we are going to try to teach a  goldfish to drive a car…of sorts. Now,   you may be asking, are we  talking about a virtual fish,   like one of the virtual characters here?   No-no. This is a real fish! Yes, really.    I am not kidding. This is an experiment where  researchers took a goldfish, and put it into   this contraption that they call an FOV, a  Fish Operated Vehicle. Very good. I love it. This car is built in a way such that it goes  in the direction where the fish is swimming. Now, let’s start the experiment by specifying  a target and asking the fish to get there,   and give them a treat if they  do. After a few days of training,   we get something like this.   Wow, it really went there. But, wait a second. We are experienced  Fellow Scholars here, so we immediately say   that this could happen by chance. Was this  a cherry picked experiment? How do we know   that this is real proficiency, not just chance?   How do we know if real learning is taking place? Well, we can test that, so let’s  randomize the starting point,   and see if it can still get there. The answer  is yes… yes, it can! Absolutely amazing.    This is just one example from the many  experiments that were done in the paper. So, learning is happening. So much so,  that over time over time, our little friend   learned so much that when it made a mistake,  it could even correct it. Perhaps this means   that in a followup paper, maybe they can  learn to even deal with obstacles in the way. Now, note that this was still just a couple of  videos, there are many more experiments   reported in the paper. Now, at this point,  we are still not perfectly sure that learning   is really taking place here, so let’s run a  multifish experiment and assess the results. Let’s see…yes! As we let them train  for longer, all 6 of our participants   show remarkable improvement in finding the  targets. The average amount of time taken is   also decreasing rapidly over time. These two seem  to be extremely good drivers, perhaps they should   be doing this for a living. And if we sum up  the performance of every fish in the experiment,   we see that they were not too proficient in  the first sessions, but after the training,   wow, that is a huge improvement. Yes,  learning indeed seems to be happening here. So much so that…yes, as a result, I kid you not,  they can kind of navigate in the real world too. Now note that the details of  the study were approved by the   university and an animal care committee and  were conducted in accordance with government   regulations to make sure that nobody  gets hurt or mistreated in the process. So, this was a lot of fun, but what is the  insight here? The key insight is maybe navigation   capabilities are universal across species. We  don’t know for sure, but if it is true, that is   an amazing insight. And who knows, a couple papers  down the line, if the self-driving car projects   don’t come to fruition, maybe we will have fish  operated Teslas instead. What a time to be alive! Thanks for watching and for your generous  support, and I'll see you next time!