On September 4, the British magazine The Economist revealed some rather surprising information. In fact, it revealed that self-driving cars are no smarter than 7-month-olds. The reason is that, unlike children, these vehicles do not integrate the notion of permanence of objects.
Unsurprisingly, many automakers have entered the field of autonomous driving. But although some have already managed to get their vehicles to a certain level of autonomy, few manufacturers still dare to launch them on the road.
Regardless, among the most popular and followed manufacturers in the industry, Tesla has already promised a 100% autonomous car without a steering wheel or pedals by 2023.
Also read: A YouTuber changed a Tesla Roadster for an NFT
AI and 7-month-old baby have a different view of reality
In reality, autonomous driving continues to cause a lot of doubt and skepticism among most people. The Economist expressed this by stating that despite the observed advances, autonomous cars still do not know how to understand the environment and, therefore, how to anticipate human behavior.
The article takes as an example that “a 7-month-old child from whom a toy is confiscated is able to understand that the toy still exists.” However, this is not the case for an autonomous car. For the latter, “a bicycle that momentarily disappears behind a van is a bicycle that ceases to exist.” Here is one of the fundamental differences between human and machine intelligence.
To read also: Tesla Bot: Elon Musk just introduced a humanoid robot, but you will run faster than him
A “software block” could solve the problem
Also, you should know that artificial intelligence uses machine learning, which is based on many images and videos. Despite this, the fact that the AI does not have a fundamental understanding of its environment is the source of the various obstacles encountered, such as the consideration of raindrops and insects as obstacles.
To solve this problem, CoDesign Lab company founder Mehul Batt suggested designing algorithms for autonomous cars by adding a software block after the usual results. This “software brick” is software that applies physical concepts rather than addressing the environment in a probabilistic way. The first tests have already been carried out and have proven to be encouraging, although much remains to be done.
Despite everything, many doubt that one day artificial intelligence in autonomous driving will ever reach the human level. However, as the saying goes, never say never …