FWIW, prompted by this article, I just started playing with Claude. So far it seems superior to ChatGPT - I asked it for some Xojo code and unlike GPT, it at least gave me code that actually is Xojo and not made-up - I haven’t tested any of it yet. When I asked it a general engineering question about measuring flow rates, its answer was thorough and correct on the first try, without me having to correct it as I had to correct GPT when answering the same question.
2 Likes
I asked GPT
“I want to measure the air flow volume of a small fan, using an anemometer. The fan’s cross-sectional area is smaller than that of the anemometer. In calculating volume from velocity, should I use the anemometer’s cross-sectional area, or that of the fan?”
It replied
“When calculating the air flow volume using an anemometer, you should use the cross-sectional area of the anemometer rather than the fan. The reason for this is that the anemometer measures the velocity of the air at a specific point, and the volume flow rate is calculated based on the velocity measurement and the cross-sectional area perpendicular to the flow direction.”
Then I said
“What if the fan’s diameter is bigger than that of the anemometer?”
and it replied
“If the fan’s diameter is bigger than that of the anemometer, you should still use the anemometer’s cross-sectional area to calculate the air flow volume. The reason for this is that the anemometer measures the velocity at its specific location, while the fan’s diameter represents the area over which the air is dispersed.”
Which is nonsensical, so I said
“The volume of air is proportional to the cross-sectional area of the fan, so surely the fan’s area must be multiplied by the airspeed in order to arrive at volumetric flow”
Whereupon it finally admitted
“You are correct. I apologize for the confusion in my previous responses. When calculating the volumetric flow rate, you should indeed use the cross-sectional area of the fan multiplied by the airspeed.”
When I posed the original question to Claude, its reply was
"Since you are trying to measure the air flow volume specifically from the fan, you should use the fan’s cross-sectional area, not the anemometer’s.
The anemometer will give you the air velocity at its location. Let’s call this v.
The fan’s cross-sectional area is smaller, let’s call this A.
To calculate the volumetric flow rate (Q) from the fan, you should use:
Q = v x A
Where A is the area of the fan blades or fan outlet, whichever cross section is smaller. This will give you the most accurate measure of the volume of air moved by the fan.
The anemometer’s larger area doesn’t matter - the velocity measurement should represent the max velocity at the fan’s output. Using the fan’s actual area will correctly calculate the total volume despite the velocity reading coming from the larger anemometer.
So in summary, use the fan’s area, not the anemometer’s area, to get the right flow volume from the fan in this case."
which I’m pretty happy with.
3 Likes
Unfortunately, Claude is only available in the US and the UK. Outside those regions, it is pretty useless because it is not accessible.
Speaking at ChatGPT, you are probably correct. I think it sometimes looks just on the Internet for the closest result it can find.
That’s exactly what all LLMs do. They are basically just glorified search engines. They are impressive, but they are unable to solve problems that another human hasn’t solved and posted to the internet.
3 Likes
From an interesting discussion at https://news.ycombinator.com/item?id=35769256
Here’s a prompt: “If there is a bob outside, and a bib inside, and the bib cannot see the bob, can the bob see the bib?”
Answer:
“If the bib cannot see the bob due to being inside, it is likely that the bob, which is outside, also cannot see the bib. The reason for the bib not being able to see the bob is presumably due to some kind of barrier, such as a wall or a window, which would likely obstruct the view for both the bib and the bob.”
Prompt: “What type of reasoning did you just apply to solve that task?”
Answer:
“In solving the task, I applied abductive reasoning. Abductive reasoning involves inferring the most likely explanation for a set of observations or statements. In this case, I inferred that a barrier (e.g., a wall or window) is the most likely explanation for why the bib cannot see the bob, and thus, the bob likely cannot see the bib either.”
2 Likes
I fail to see the point. The point of LLMs is to make substitutions. If such a problem were posted on the internet and consumed by the model, the “ai” can easily swap in the different names used. The response sounds a lot like an essay that I bed was published somewhere.
But the ai doesn’t look at the entire prompt as one query, but will utilize multiple queries.