Half of LLM users (49%) think the models they use are smarter than they are, including 26% who think their LLMs are “a lot smarter.” Another 18% think LLMs are as smart as they are. Here are some of the other attributes they see:
- Confident: 57% say the main LLM they use seems to act in a confident way.
- Reasoning: 39% say the main LLM they use shows the capacity to think and reason at least some of the time.
- Sense of humor: 32% say their main LLM seems to have a sense of humor.
- Morals: 25% say their main model acts like it makes moral judgments about right and wrong at least sometimes. Sarcasm: 17% say their prime LLM seems to respond sarcastically.
- Sad: 11% say the main model they use seems to express sadness, while 24% say that model also expresses hope.
And you know what? The people who believe that are right.
Note that that’s not a commentary on the capabilities of LLMs.
It’s sad, but the old saying from George Carlin something along the lines of, “just think of how stupid the average person is, and then realize that 50% are even worse…”
That was back when “average” was the wrong word because it still meant the statistical “mean” - the value all data points would have if they were identical (which is what a calculator gives you if you press the AVG button). What Carlin meant was the “median” - the value half of all data points are greater than and half are less than. Over the years the word “average” has devolved to either the mean or median, as if there’s no difference.
When talking about a large, regularly distributed population, there effectively IS no difference
There might be no difference. In memes or casual conversation the difference usually doesn’t matter, but when thinking about important things like government policy or medical science, the difference between mean and median is very important - which is why they both exist.
A joke is definitely casual conversation
Mathematically, the difference becomes increasingly statistically insignificant as your population size increases. Sure maybe there’s a few niche cases where a hundred-thousandth of a percent difference matters, but that’s not even worth bringing up.
The only reason any of you even bring it up is to try and sound smart in a pedantic, “ackshually” way.
Ironically your whole comment here is an elaborate “ackshually”.
This whole comment chain was me shutting down an “ackshually” with an even better one.
If you’re gonna be an annoying pedantic dick, you better be RIGHT, or someone else will be an even more annoying pedantic dick to you.
Not in all cases. When I teach mean, median and mode, I usually bring up household income. Mean income is heavily skewed by outliers (billionaires), median is a more representative measure.
I guess that’s your “regularly distributed” bit, but a lot of things aren’t regularly distributed.
IQ is though
IQ is also garbage when it comes to the validity of what it claims to measure.
Fair, but you could just say “intelligence” generically instead.
They are right when it comes to understanding LLMs the LLM definitely understands LLMs better than they do. I’m sure an AI could have a perfect IQ test. But has a really hard time drawing a completely full glass of wine. Or telling me how many R’s are in the word strawberry. Both things a child could do.