Math grade for ChatGPT: F(AIL).
That’s hilarious.
Plus 20 characters.
One would think that the math side of things would be the easiest to program correctly.
DB2
Not with machine learning. Math uses a different kind of algorithm. I can’t really explain it but from my programming experience I know there is a difference between how the subconscious solves problems and how the rational boolean brain solves problems – two different kinds of algorithms.
Doing my very first program at IBM I experienced it. I wrote a program that was too big for the computer and I told my boss it didn’t fit. He replied, “It fits!” Back to the drawing board but I could not get it to fit. Back to my boss, same answer, “It fits!” Maybe a third time. I had a patient boss. Then one night I woke up around 4 AM, "IT FITS!!!: The subconscious found the solution. I was usually late for work but that morning I could not wait for IBM to open the office so I could test the solution. It worked!
I’m curious how they marry the two kinds of algorithm at Tesla for FSD.
The Captain
Not only is it math challenged, it makes stuff up (lies?).
“Please write a blog post discussing South Dakota’s oldest and youngest governors.” Here is what it wrote:
South Dakota has a rich history of political leaders who have left their mark on the state…
Oldest Governor: Nils A. Boe
Nils A. Boe served as the 18th governor of South Dakota, from 1965 to 1969. Boe was born on May 10, 1909, in Hayti, South Dakota, and he grew up on a farm near Watertown…
Youngest Governor: Crawford H. “Chet” Taylor
Crawford H. “Chet” Taylor served as the 14th governor of South Dakota, from 1949 to 1951. Taylor was born on July 23, 1915, in Sioux Falls, South Dakota, and he grew up in nearby Flandreau…
Not a bad effort, although almost entirely wrong. Most notably, Crawford H. “Chet” Taylor was never Governor of South Dakota and, in fact, I can find no evidence of such a person, at all. I will credit ChatGPT, though, that Governor Taylor is a plausible-sounding fictional governor.
DB2
It is called hallucinating in the biz. It is a known problem with AI.
Here is an article (gifted, so it should work) in today’s Wapo about AI hallucinations. No one really knows what to do about it.
“No one in the field has yet solved the hallucination problems. All models do have this as an issue,” Pichai said in an April interview with CBS. Whether it’s even possible to solve it is a “matter of intense debate” he said.
Note that ChatGPT is a language bot. It will give stuff that it calculates sounds good as conversational language but who knows about accuracy. The ChatGPT list seems a bit suspect when Hawaii (15517) has similar driving mileage as Arizona (15541).
If the ChatGPT learning program values human-like prose over computer-like accuracy then who knows what it considers to be “facts”.
Here is Kelley Blue Book data for 2021. Lots of fairly big differences, e.g., Hawaii (10869), Arizona (12728).
Not the best advertisement for AI doctors
In our not-so-brave new world everybody is entitled to his/her/its facts and ChatGPT serves his/her/its clientele.
The Captain
The Planet Money podcast is doing a 3 part series using ChatGPT to write a script for them. A problem they ran into in one of the early iterations of the script was a reference to a journal article that appeared to be false. They tracked down a researcher with the name of the author of the cited article. She denied writing it or ever making the statement that ChatGPT quoted.
The chatbot junk is a sickening signs of larger scale problems of digitally generated “lifelike” and “realistic” lies.
My grannies insisted that I learn to appreciate the depths of the process of “formal introductions” as crucial validations of the honesty and solidity of people. They argued that without a formal introduction I should be open to the possibility that someone is good, but not gullible.
They would consider almost everything about the social world we live in as INSANE and DANGEROUSly gullible.
david fb
And whose fault is this “not-so-brave new world”? In the past, “facts” were disseminated by people who had to earn the right to be called experts. They had to justify, usually by merit, that they had enough expertise that someone would publish their books or put them on TV/radio. Those were the only ways to reach a significant audience.
With today’s social media, all one needs to reach a crowd is personality. When one rejects the “elites” then one also devalues expertise. Facts are now defined by the loudest common denominator.
Welcome to anarchy!
Maybe you overvalue the past when a fool was born every minute.
The difference is that now fools can be louder.
The Captain