Sit back and let the machine figure it out.
A growing amount of scientific research involves using machine learning software to analyse data that has already been collected. This happens across many subject areas ranging from biomedical research to astronomy. The data sets are very large and expensive.
But, according to Dr Allen, the answers they come up with are likely to be inaccurate or wrong because the software is identifying patterns that exist only in that data set and not the real world.
“Often these studies are not found out to be inaccurate until there’s another real big dataset that someone applies these techniques to and says ‘oh my goodness, the results of these two studies don’t overlap‘,” she said.
“There is general recognition of a reproducibility crisis in science right now. I would venture to argue that a huge part of that does come from the use of machine learning techniques in science.”
Why do you need reproducibility in science? The left wants science to support their social justice constructs. Machine learning lets the left cherry pick results and lets them hide inconvenient data.
Reproducibility is essential to science but the left doesn’t want real science.
This is a recurring and systemic problem because:
1) the younger scientists/engineers do NOT fully grasp the knowledge needed to actually work in their discipline and lack any grounding in the science of common sense to deal with it, and compounded with a level of hubris that makes trudeau look like humble pie.
2) the boomer generation currently handing out the funding and grants seek primarily to a) help their kids, b) preserve their legacy (i.e. disallow any work that would negate theirs), and c) eases their transition to a comfy retirement. They also don’t have a clue what AI or deep learning actually is but can’t admit it.
Particularly for medical science, where we are working with an already engineered unit designed by rules we can’t fathom, we really aren’t much beyond putting stuff in the fuel tank and seeing what happens to the engine.
I can’t wait to see the results of using aborted baby parts mixed in with mice and grown in a lab. I am sure that will yield nothing but good results. Maybe we can power them with wind turbines made from green lithium and sprinked with fairy dust.
Green lithium and fairy dust?
I didn’t know our Minister of the Environment posted here at SDA. Welcome!!!!
Lol, but what Frenchie says is about the size of it. The other evening Climate Barbie was literally shrieking about we only have twelve years left.
She did not however elaborate on whether that twelve years was tacked onto the previous twelve years that was predicted by around the year 2,000. I can’t remember who by, was it Al gore or the Fruit Fly guy?
Climate Barbie is so dumb other blondes have noticed it.
It’s the need to constantly publish something, anything.
Sounds just like the Climate Change data BS.
Insects are smarter than so-called self-driving cars but human hubris can’t accept that.
good grief dont let THAT factoid out. the ‘experts’ are liable to hook up the steering wheel to the cockroach’s brain. excluding the ones that had it bitten off by a rival.
apparently cockroaches can live without one:
https://www.scientificamerican.com/article/fact-or-fiction-cockroach-can-live-without-head/
and, additionally dont let THAT factoid get out either or the LIEberals are apt to pick one
to replace the TURDoo.
Or, as we used to say in the old days of punch card programming, “garbage in, garbage out”.
Unfortunately that still applies……but is ignored…….Steve O
Yes, B A, whenever I hear a ‘scientist’ exalting the results of a certain computer model, what I actually picture in my mind is a phony fortune teller sitting in front of their crystal ball.
Nothing new. Back in the early 1980s I used U of Alberta statistical software that turned multiple long columns of numbers into formulas and did statistical analyses on the numbers. It beat the hell out of trying to determine relationships using my hands and my head. If I remember correctly, we were taught that there could be many problems with data. Haven’t done any statistical research in at least 35 years and I was always more of a qualitative than quantitative guy. I preferred bullshit to numbers although I managed to pass 5 math and stats courses without understanding one damn thing. Repetition, repetition, problems, problems, problems.
FAMOUS Scientist quote by Alexandria Occasio Cortez
“We need to invent a technology that hasn’t been invented yet”
Well at Advanced AI Analytics they can crank that out for you in like 45 minutes….
Define “patterns that exist only in that data set and not the real world.”
I suspect that here they include differences between the sexes and breeds of mankinds—which are self-evident in all social scientific data sets as well as the real world, no matter how much the left pretend otherwise.
If a computer algorithm ISN’T racist and sexist, it probably wasn’t trained properly.
An example can be data that is sampled twice per day, say 3am and 3pm. If you only examine that data, you can find patterns like “afternoon samples are warmer than morning samples”. Of course, if your samples are taken at 11am and 11pm, then that pattern would be reversed. This is explained by the Nyquist sampling criterion; we cannot legitimately recognize actual patterns unless they occur at less than half the frequency of our sampling, or alternatively we cannot make predictions further into the future than half the time we’ve been keeping records. That’s why climate predictions a hundred years into the future make no sense, when we’ve only been keeping temperature records for a hundred and fifty years, and high-precision satellite measurements for less than fifty years.
All models are wrong. Some are useful.
kinda looks like the first time a computer verifiably mimics the humahn brain will be that of a . . . . . .
schizophrenic.
(pssst, all part of the plan! aka watch how quick they ‘blame the computer’ when THAT happens )