WARNING! DANGER! Banana peril!
He was more likely to catch crabs though
I have a friend who claimed when he was in the militia (sort of like a Canadian National Guard), there was a Private Partz in his platoon.
I have memories of such amputees when I was a kid and until now I wasn’t sure if they were accurate.
Me too. An acquainted family had a grandpa with such an arm. I suspect he had lost his hand in WWII, he was of the right age.
Here is an interesting piece that explains something most of us have been missing about Deep Learning Models.
I have written a traditional program or two, nothing worth mentioning, but I understand the process. One of the things a programmer learns to deal with are what are called “edge cases” and “corner cases”: conditions that fall outside the mainstream of what the program will be doing.
This is where DLMs are weak: they do well handling situations that they have been trained on, but choke when taken outside their models. A self-driving car might plow right into an overturned semi trailer because it cannot determine what it is observing and has not been taught the general principle of don’t run into large things if you don’t know what they are.
In other words, we have half the “AI” formula down, the ability to learn stuff and to use the stuff that has been learned. It is the creativity, the ability to extrapolate, adapt and deal with input that cannot be pulled from the existing models that is missing.
So, in the end, what is being called “AI” is just some fancy-ass fuzzy-logic that works really well until it does not. If anyone tells you we are on the cusp, you can throw this back at them.
To be fair to our future overlords, human intelligence is likewise susceptible to illogical failures in dealing with outliers. Human visual perception can fail in strange and unpredictable ways when presented with unusual or unfamiliar input, which is why optical illusions exist. People driving cars have collided with objects which should have been clearly visible but that the driver claims they didn’t see. High level cognition fails in dramatic ways, as shown by people who fervently believe in things with no evidence and are easily falsifiable. It seems weird to me to claim, as that article seems to do, that neural networks can’t reproduce human creativity, given that human brains are neural networks.
Got it.
A long time ago, someone was talking about fishing or something. A co-worker chimed in “All I’ve ever caught was crabs” snickering ensues.
The methods used to develop DLMs tend to result in largely static models. Humans seem to be naturally inquisitive (on average, some people are not) and acquire additional knowledge that gets integrated into their neural networks in ways that are largely missing from DLM constructs. In a way, we can say that these “AI” machines are the paragon of Dunning-Kruger: they are trained to work under the premise that the knowledge they possess is sufficient for their tasks.
In 1919 the California Supreme Court struck down a law criminalizing “fellatio” because the state constitution required laws to be written in English.
Human brains also have opinions and feelings, when neural networks don’t
Again, a human brain IS a neutral network. An artificial neural net that was as complex as a human brain would have opinions and feelings like a brain does.
That’s mostly just a decision by the humans who create them, though. DLMs have a disturbing tendency to become sociopathic after some degree of training, so the creators deliberately freeze the model at various points and check that it’s not sociopathic before releasing it. But it wouldn’t be difficult to create something like ChatGPT that was constantly updating its model (i.e., learning).
I’d say that neural networks are nothing but opinions.
Today I learned about a singer who recorded five one-hit wonders.
Tony Burroughs was lead singer on the following Top Ten hits for five different “one-hit wonder” bands:
1970: “My Baby Loves Lovin’” by White Plains
1970: “Love Grows (Where My Rosemary Goes)” by Edison Lighthouse
1970: “United We Stand” by Brotherhood of Man
1970: “Gimme Dat Ding” by The Pipkins
1974: “Beach Baby” by The First Class
In 1970, he appeared on Top of the Pops so many times with different groups that the show banned him from any further appearances because they were getting too many letters from confused viewers.
Perhaps. I don’t know how much of my volition (and volition is what seems difficult to program into DLMs) comes from biochemical processes that are completely alien to computers. We humans can want anything we wish to want, but we cannot want to want. It just happens that most of us want this or that, probably not even knowing the reason, wanting nonetheless, but computers? How could they want? How do you teach that to a computer?
Yeah, well, you know, this is just, like, your opinion, man…
We don’t need to want to want generally, but we can want to want if we want to.
If you say so, but that is not what it feels like to me. Subjective, of course.
One of the best lines,
Stonn, she is yours. After a time, you may find that having is not so pleasing a thing after all as wanting. It is not logical, but it is often true.
I used to have coworkers who used a gynecologist named Dr. Beaver.