Will Neurosymbolic AI Change Autonomous Vehicles? It was done in GTA.
<p>A few weeks ago, a grad student named Wout Vossen in Belgium posted a video where he create a self-driving car in <a href="https://en.wikipedia.org/wiki/Grand_Theft_Auto" rel="noopener ugc nofollow" target="_blank">Grand Theft Auto (GTA)</a> using <a href="https://neurosymoblic.asu.edu/" rel="noopener ugc nofollow" target="_blank">neurosymbolic AI.</a> This video touched on two areas that I am currently working on: self-driving cars and neurosymbolic AI. I was fortunate enough to chat with Wout about his experience. To me his project illustrates some ideas about both neurosymbolic AI and self-driving vehicles.</p>
<p><img alt="" src="https://miro.medium.com/v2/resize:fit:630/1*Xs4Dva33-GKggO5e0sXEjA.png" style="height:380px; width:700px" /></p>
<p>Wout’s self-driving vehcile in Grand Theft Auto</p>
<p><strong>Road rule can be directly enforced during training. </strong>This is one of the main advantages to current neurosymbolic approaches — the addition of rules. In Wout’s experiment, he encoded traffic rules directly and these rules were enforced during training using. The result is the autonomous agent is more likley to adhere to the rules.</p>
<p><a href="https://medium.com/towards-nesy/will-neurosymbolic-ai-change-autonomous-vehicles-it-was-done-in-gta-26d50ca4e3f9"><strong>Click Here</strong></a></p>