AI = LLM
LLM = hallucinations, model collapse, extraordinary energy consumption (Training and Inference), regulatory risk, copyright risk, legal risk, societal pushback.
the amount of money invested vs the actual predicted potential return is wrong.
for every 1 experienced software developer who benefits from using ChatGPT, 10 are using it wrong and entering poorly written insecure code into the wild.
cleanup will be a full time job
Bumped up from @djv+1uNPJZR6. All so very true.
If Cisco is going to step up its traffic management with AI it's not going to be doing it with artificially generated book reports or fake images of tiny Asian women with ginormous backsides.
For a very basic perspective of AI start with:
https://en.wikipedia.org/wiki/Artificial_intelligence
AI is a hot mess. It is good at making odd and slightly off looking art and stealing work from humans. It is good at getting driving to the 98% (But the 2% long tail tends to make mistakes that would certainly ki-l people (https://www.quora.com/What-are-the-typical-scenarios-where-self-driving-cars-machine-vision-AI-fails-to-interpret-scenes-or-roads-correctly-despite-training).
If you use AI outputs, to inform new AI the entire system breaks by the third or fourth interation. (AI > AI > AI = Trash / nonsense... Human > AI = stolen half decent work). Without human creativity to source from, the third or fourth generation AI tends to just make garbledy go-k that has no meaning.
It'll be great at:
But the costs outweight the profit in most cases. Tier 1 support will probably be all AI, Robo-calls for spam will probably be AI, search engines will likely get updates, and self-driving taxis will probably work on limited tracks inner-city.
But most of the other applications are too costly to be functional.
This may be one time in Cisco's history that it pays to be late the the game.
Wasn’t this already discussed in @OP+1uNPJZR6 post?
Massive LRs Loom from AI Ponzi