Hilarious that people think AI will take over. AI can only be effective if our data is in good shape, which it is not. It is inconsistent, incomplete, error prone, and very much out of date. We’ve invested millions of dollars into planning and automation systems which don’t work as intended because the data assumptions are faulty. We need to spend the money to fix the data.
Couldn’t agree more, @fe+1jvksrfw2.
Amazon was caught lying that they had AI track their customers at their grocery stores. It turns out, they had people in India monitoring them.
https://www.theguardian.com/commentisfree/2024/apr/10/amazon-ai-cashier-less-shops-humans-technology
They market and publicize "AI taking over" when in reality it is offshoring to India and a few other low cost countries + lots of H1B1 foreign workers visas mostly from India. We see how bad the quality out of India is...
I have done a lot of research in AI. Data can be divided into categories and subcategories or clusters. Does there exist at least one field where a cluster of data is too small, limited, or corrupted to facilitate AI? Yes. Absolutely. However, to generalize that hypothesis to ALL data would be inaccurate. Then it would become an example of selection bias and observation bias.
There are cases where there is more than enough data to support AI but the wrong architecture is used. A primary example is a class imbalance using logistic regression. But that can be addressed by switching to a different data model using graphs with nodes and edges. So, instead of using pyspark with distributed data one could switch to a graph data model using Neo4J. The latter would address class imbalance more succinctly by switching from ANN to GNN. There are other approaches to address data issues. For example, heterogenous medical data sets can be aggregated with innovative labels - a common task in merging clinical data with machine learning models.
Like it or not, AI is coming. People resisting that technology remind me of the 1970s when some resistors wanted to keep using 8-track tapes. It reminds me of when Sears of the 1990s resisted the next innovative step of putting their classic, immense catalog online. That led to its demise. I know there is an empty Sears next to the capital at St. Paul, MN. I parked there for the Medtronic marathon before it was cancelled due to heat.
Instead of "resisting" AI it's best to become an early adopter. I could have helped Medtronic merge healthcare technology with AI. But they sent my job to India. However, there are other companies interested in those AI skills. So, I am still in demand despite the horrible job market and keep getting inquiries from recruiters. I learned to change with technology. I still remember when Fortran was the "great new language" when I was an undergraduate back in the 90s - lol. Currently, I've learned python, Java, C++, Matlab, Swift, SAS, Kotlin, R, and Java. Next up is cypher.
There are so many things I could say about this being a practitioner in AI, but all I will say is you have identified the key barrier that hardly anyone in management understands or will accept.