My very bold prediction is that AI's capabalities are not overblown, but the practical applications will be WAY more limited once they get these "products" into the marketplace. Every exec/director/middle manager overstates how integrated their systems are, and my experience is that right now at least, humans are making up that gap.
In order for AI to work in the workplace the way people are afraid it will and disrupt jobs the way people predict, system exchange and interoperability are going to have to make giant leaps forward. The big companies like Amazon, UPS, MS, Fed ex - the ones with investment in infrastructure - they will see benefit in the supply chain and delivery systems, but Guitar Center, Brinker, sprouts, etc, the middle guys who have not invested in the comms and operations of their various systems need to do that first and long before sticking any AI in there - it won't work.
Humans are making up that gap today. Too many companies have too many isolated and antiquated architectures to see any super benefit from this shit. If your retail and inventory systems are still running SW more than 2-3 years behind, well, you're way behind and AI won't get you there.
I do see some things eventually coming to pass. Docs, lawyers....the types of jobs where memorization and application of information is paramount to decision making....machines will do that sooner I think. Even checking out groceries will be automated, but I just don't have the same fascination with the demise of the workforce that others do. It will happen I'm sure, but as with all things, the people in charge have no idea how shit really works.
Re: Thing: Artificial Intelligence
262Real world example - I work on a lot of Cisco gear. That shit is expensive, and one of the ways to save money is to not pay for certain support packages, or allow working usits to go EoL without replacement.
In order for AI to replace my work, the gear in question would have to have some out of band comms to notify another system that the device has a problem, the diagnostics from the device would have to be enabled and would need compatible modules (not knockoff amazon parts like SFP or LTE mod inserts) in order to accurately report and diagnose the problem, then from there you'd need a tie in to a dynamic inventory system to order the replacement for the defective part.
Most of that chain of automation is broken in every env I've worked in. Whether the modules were installed by a contractor cutting corners and hoping to not be caught, or there was an SFP surplus from the last set of gear, or your out of band comms are down and you didn't know because it's not a managed service......in order for AI to work, your automation has to work, and like I said, most places are 5-10 years from that.
Add in the fact that you have to invest capital just to invest in the AI to then get an ROI......I dunno man, feels a bit like snake oil right now for many things.
In order for AI to replace my work, the gear in question would have to have some out of band comms to notify another system that the device has a problem, the diagnostics from the device would have to be enabled and would need compatible modules (not knockoff amazon parts like SFP or LTE mod inserts) in order to accurately report and diagnose the problem, then from there you'd need a tie in to a dynamic inventory system to order the replacement for the defective part.
Most of that chain of automation is broken in every env I've worked in. Whether the modules were installed by a contractor cutting corners and hoping to not be caught, or there was an SFP surplus from the last set of gear, or your out of band comms are down and you didn't know because it's not a managed service......in order for AI to work, your automation has to work, and like I said, most places are 5-10 years from that.
Add in the fact that you have to invest capital just to invest in the AI to then get an ROI......I dunno man, feels a bit like snake oil right now for many things.