Quote:
Originally Posted by chezlaw
None the less there are AI lorries on the road in sweden and cars can do all sorts of things that used to require human decision making. Mostly it's about reducing the amount of human hours/involvment required. Maybe I'm wrong but it seem to me like you think it's not working unless no human involvement is required
It's already a case of what can be done. The only faith (if we want to ignore the fact it's not faith) is how fast it it is happening.
Ya you are very wrong.
I have said consistently it is and has worked WELL within very confined limits to date. Yes AI is doing some of the more black and white, clear Yes/No type tasks. It is easy for AI to disqualify everyone applying for a loan who lives outside the service area by just looking at Zip Code. Check. And that does save time. Good.
But the next steps for AI where they can truly replicate and replace the human factor in driving and other areas is just not there.
And more importantly that next advance is no longer believed, as some theorized and hoped at the start just a function of 'Big Data' but will require new 'logic' pathways be discovered.
The biggest advancing in AI we are currently enjoying were due to Amazon, Facebook, Google et al opening up and utilizing their big data in public consumption ways. So much data eventually starts to create road maps that Machine Learning and AI can read and use as predictive. And early on the limits of where Big Data could take the world were unknown and many assumed it would be enough for most things like Autonomous cars. It is no longer believed that Big Data alone and the continued aggregation of it will solve these remaining issues.
To make it real simple it is easy for Big Data to predict the most efficient path between A and B by looking at the route millions of cars took and time they took. That path Big Data gives will be the best MOST times and if the cars stick to it they will get the best result most times. Big Data is very good at that.
Where Big Data sucks is when you have dynamic environment of pedestrians, cyclists and people getting in and out of cars into the flow of traffic. pulling out of parking spots or other such dynamic events that require adjustments. We humans, have a form of intuition for those events. An alert driver is estimating in advance (defensive driving) whether it looks like the person sitting in the car may open their door and step out, a cyclist might be doing a shoulder check as they need to get over to turn left, etc. We basically make instantaneous bets and start reacting to them in advance.
The belief is that for AI to get that 'last mile' it will come from Quantum Computing and perhaps AI eventually teaching itself in ways we humans do not even understand in the way some AI game engines now learn chess and other games. But that needed jump in AI tech may not ever happen. It is not a given and there is a good number of experts who do not think it will happen. We are talking an AI jump that would be a precursor for a Data like AI using logic and intuition and very much approximating sentience, if not achieving it under some definitions.