4 ms·
I think the Docklands Light Railway (DLR) model is fine. The author seems to be as ignorant of "machine learning" as he claims the general public to be of the c
by davidhyde 5y ago
I think the Docklands Light Railway (DLR) model is fine. The author seems to be as ignorant of "machine learning" as he claims the general public to be of the concept of "driverless trains". For those unaware of what the DLR is, it's a 35 year old driverless train system in London with a human "train captain" who is basically like any other passenger but that can also operate the doors and override things if needed. You don't need machine learning and AI to drive a train on a one dimensional track, just a half decent team of Engineers to design something properly.
Yes the "train captains" can strike but they are far more replaceable than highly trained and very highly paid human train drivers. The evidence is clear, the DLR is far less affected by industrial action than regular tubes. Industrial action is so rare on the DLR that it becomes delicious news when it does happen: https://www.independent.co.uk/voices/comment/strike-proof-docklands-light-railway-has-gone-strike-a6720706.html https://www.independent.co.uk/voices/comment/strike-proof-do...
- Silhouette 5y agoYou don't need machine learning and AI to drive a train on a one dimensional track, just a half decent team of Engineers to design something properly. Most of the article was about the reasons "designing something properly" is an almost impossible task in this situation, once you take into account the reality of what is there right now, the goal(s) you might actually want to achieve, and the likely costs of doing so.