Slashdot Mirror
IBM Aims To Meld AI With Human Resources With Watson Suite (zdnet.com)
PolygamousRanchKid shares a report from ZDNet (with some commentary): IBM has launched a unit designed for human resources to better find talent and recruit using artificial intelligence. The company is wrapping its latest HR effort, dubbed IBM Talent & Transformation, which includes select Watson services. According to IBM, its suite of AI tools can help HR become a growth engine to enable digital transformation. AI can be used to revamp workflow, employee engagement, recruitment and retention while providing a more diverse workforce. (I can still program Fortran; I learned it from Forman S. Acton -- does that make me diverse enough?) Big Blue's Talent & Transformation suite includes a Watson Talent Suite that rolls up behavioral science, AI and psychology and applies it to HR. (Sounds like the recipe for The Apocalypse to me.) IBM Garage, which serves as a test bed to meld HR, AI and culture, will also be available. (Garage? It sounds like the creepy CRISPR basement of a mad scientist to me.)
... unemployed and has been fired too many times.
It little behooves the best of us to comment on the rest of us.
The medical one failed pretty spectacularly, with Watson recommending treatments that would have killed people, as far as I remember. I also know of an attempt to use in in IT security, but basically it ended up being a kind of news-compiler.
Seems to me that while Watson is a nice demonstration about the state-of-the-art in NLP, that state is still sorely lacking and may continue to sorely lack for a long, long time and possibly forever. (And don't give me that nonsense that "science" would be claiming humans are just computers on legs because everything is known Physics. Science claims no such thing. Science very much says that we have no clue how humans do it. Incidentally, known Physics is known to be wrong, unless somebody solved quantum-gravity while I was not looking.)
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Basically, Watson can help in data-mining in data encoded in natural language. It does not do it very well though and a human expert checking the results is critical. All Watson can really give you is hints. Incidentally, to expert audiences, the IBM folks are not really claiming more than that. I have heard members of the Watson team speak to expert audiences several times now. But those claims are very far from what the general public can understand and and what Watson can do is not easily and directly applicable to problems.
As to what AI is today, it is not even simulated general intelligence. It is basically some pattern matching and some statistics and some simple automated deduction. There are of course the clueless that think things like recognizing a street-sign needs intelligence, but all we are finding is that doing is badly and in a way that is easily fooled, does not actually require intelligence. Or that think a piece of software playing Chess or Go must obviously be intelligent. That is a fallacy. Just because you see some black box perform a specialized task that can also be performed using general intelligence (i.e. humans) does not mean that box is intelligent in any way. What we are actually finding is that quite a few tasks or parts of tasks humans are used for today do not actually require intelligence, but that dumb automation can do it. So yes, absolutely no insight, no understanding, no knowledge, no most certainly no independent thinking (although most humans are basically unable to do that one too). And yes, we have absolutely no clue how humans do it. Some reputed Neuroscientist (they are not all hacks) recently said "the closer we look, the more mysterious it becomes" (cannot find the source of that anymore, sorry).
Sure, some parts of what humans do (motor functions, e.g.) are basically also just dumb automation. But when you look at true feats of general intelligence, humans are leaving machines completely in the dust. For example, automated theorem proving can theoretically find all theorems of a mathematical theory, given some upper proof and theorem size boundaries and given enough (but finite) computing power and memory. However in actual reality, mathematicians find things that the machine would not find if the whole universe gets converted into a computer for its use. Still, the algorithm has the same potential as a mathematician in theory. But that does not make that algorithm intelligent, because what is extremely obvious from the performance is that the mathematician and the algorithm are using two completely and fundamentally different approaches and the machine can basically fake it for small problems (what you call "simulated").
Now, these proving algorithms are still extremely useful. Because if a competent mathematician takes it by the hand and _guides_ it through a proof the machine could never find by itself, it can still verify mechanically that the proof is correct. That tells us that finding mathematical proofs of significant size likely requires general intelligence, but verifying proofs does not and is something dumb automation can do.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
In a field where personal opinions are as important as they are in HR, its going to be difficult to find an unbiased training set. The resulting AI could easily make strongly biased decisions.