Existing AI and Machine Learning is stuck at doing literal interpretation and lacks any common-sense, which bodes for great concerns and especially when it comes to the advent of self-driving cars.
I think it safe to say that we would be surprised to find any sentient human that would act in this manner.
In this imaginary scenario, you simply saw the word “stop” and that was sufficient to get you to bring your car to a halt. We can slightly recast things and contemplate that you will at least first visually examine the stop message and only come to an actual stop if the signage sufficiently resembles a proper standardized stop sign.
A common means these days of doing this type of image processing consists of using AI-based techniques and technologies such as Machine Learning and Deep Learning . Those capabilities are neither magical and nor sentient. They are simply souped-up computational pattern matching algorithms. In addition to trying to get the ML/DL to find cars in an image, you would also seek to have the AI capabilities to find various traffic signs. This would consist of first establishing a database of traffic signs and use those during an ML/DL training stage. After doing that, you then would place the ML/DL devised image processing software into a self-driving car.
That is not something that today’s AI systems are able to readily do. There is a monkey-see-monkey-do sense of what the ML/DL is doing, though even that is an overstatement since a monkey has a semblance of sentience that none of today’s AI does .Does this troublesome problem of literal interpretation bode for issues about the advent of AI-based true self-driving cars?
Meanwhile, the Level 4 efforts are gradually trying to get some traction by undergoing very narrow and selective public roadway trials, though there is controversy over whether this testing should be allowed per se . With that clarification, you can envision that the AI driving system won’t natively somehow “know” about the facets of driving. Driving and all that it entails will need to be programmed as part of the hardware and software of the self-driving car.We will continue with the stop sign exemplar as a means of digging deeper into the existing limitations of AI techniques such as Machine Learning and Deep Learning.
This consisted of having the self-driving car drive around an area and see how well the stop sign detection functioned. To their unpleasant surprise, there were quite a number of misses such that the image processing was not signifying stop signs that undisputedly did exist. Meanwhile, the human backup driver present in the vehicle was able to readily see those missed stop signs and opted to momentarily take over the driving controls due to the AI ML/DL oversights.
The ML/DL mathematically calculated that stop signs must include two boltheads. That was computationally decreed as an essential ingredient of a stop sign. This makes sense because the provided set of stop signs all had that feature., it seems appropriate to ascertain that all stop signs have those two bolts.Of the stop signs that weren’t being detected, they all had some subtle differences about the bolts.
Norge Siste Nytt, Norge Overskrifter
Similar News:Du kan også lese nyheter som ligner på denne som vi har samlet inn fra andre nyhetskilder.
Confusion Matrix in Machine Learning: Everything You Need to Know | HackerNoonConfusion Matrix: Tabular representation of an ML classifier's performance. Evaluation metrics: accuracy, precision, recall can be found from confusion matrix.
Les mer »
Privacy policies are four times as long as they were 25 years agoA machine learning analysis of 50,000 privacy policy texts shows they have ballooned to an average of over 4000 words since 1996 – and one researcher says they are “fundamentally broken”
Les mer »
San Jose: Carjacking suspect forces woman to drive car with her family insideA carjacking suspect was arrested Tuesday after he forced his way into a woman’s vehicle and had her drive to a different location, San Jose police said.
Les mer »
A road in Detroit will charge an electric car while drivingEddy Grant sang about an 'electric avenue' in the 1980s. Detroit plans to bring that vision to life next year.
Les mer »
Battery Electric Cars Climbed to 9% of Sales in 2021, Driven by EU CO2 TargetsOne in every 11 cars sold in the EU last year was fully electric as EV sales were boosted by EU CO2 targets for the second year running.
Les mer »