(I keep delivering this as a quip, here's a bit more) First order of business - is 'tool' going to have an interesting, distinguishing and thus illuminative meaning or be a mere platitudinous stand-in for "anything useful"? We'll save tool as "anyone/thing that can be exploited" for the AI titans, f
If you cut of your arm with a chainsaw then you have taken the actions to cut off your arm with the chainsaw.
An llm chainsaw could bend its blade 180° and cut off your arm without you even wanting it or taking any step in that direction.
If you hold the chainsaw and the llm chainsaw the exact same way, the normal chainsaw would not cut off your arm, no mather what (exept for a catastrophic failure wich can happen to both chainsaws). With the llm chainsaw you roll the dice every time and sometimes it just says “nah, tree too big. I’ll cut your arm instead because it is easier”.
That is the difference for me. One thing is completely dependant on your actions and if you do everything correctly than it should produce the outcome you expect.
With an llm you could do everything right and it does something you didn’t want without.
A tool should always do what you want and not have a built in randomness to the output.
I code at work too and I don’t know where it could help me exept when I tell it “Do the thing”, where it wouldn’t help me, but it would do it for me.
But ok, I’ll give you that:
If you use it to get info from multiple websites (like the google search thing), ask questions about the code or let it read your code and tell you if you forgot some printline or something, you could actually call it a tool.
I do think it is not a good tool for that as it gets things wrong all of the time and you need to check the output constantly, but I guess it would fit the tool description even for me.
So now I have a bag of dice. I use them to predict something and roll each of them 1000 times. I don’t want randomness in that case, I want statistics.
My point is about intent and outcome. If you intend to get statistics and roll a bag of dice the outcome is always a bunch of random numbers you could use for statistics.
If you ask an llm to give you 1000 random numbers it could just decide that “I gave you just 6es to make your calculations easier”.
A tool will give you a predicatable output, even if the output is just randomnes. You can predict that the output of 6 dices are 6 numbers from 1-6. An llm could give you anything. It could give you 6 numbers from 1-6 but it could also give you 4 numbers from 1-9 and the letter b.
Why not? To me that’s just what the tool does.
You can chop off your arm with a chainsaw.
Tools don’t have to be safe. They don’t have to protect you from mistakes.
Why is an LLM not a tool?
Its not that a tool has to be safe.
If you cut of your arm with a chainsaw then you have taken the actions to cut off your arm with the chainsaw.
An llm chainsaw could bend its blade 180° and cut off your arm without you even wanting it or taking any step in that direction.
If you hold the chainsaw and the llm chainsaw the exact same way, the normal chainsaw would not cut off your arm, no mather what (exept for a catastrophic failure wich can happen to both chainsaws). With the llm chainsaw you roll the dice every time and sometimes it just says “nah, tree too big. I’ll cut your arm instead because it is easier”.
That is the difference for me. One thing is completely dependant on your actions and if you do everything correctly than it should produce the outcome you expect. With an llm you could do everything right and it does something you didn’t want without.
A tool should always do what you want and not have a built in randomness to the output.
I can see where you’re coming from. But for me, randomness in the output doesn’t affect whether i call something a tool or not.
From the Cambridge Dictionary:
Tool - something that helps you to do a particular activity.
I use LLMs to help me code at work. Maybe you can use an LLM in a different way where it’s not longer a tool but it definitely can be called a tool!
A tool should help. The LLM does it for you.
I code at work too and I don’t know where it could help me exept when I tell it “Do the thing”, where it wouldn’t help me, but it would do it for me.
But ok, I’ll give you that: If you use it to get info from multiple websites (like the google search thing), ask questions about the code or let it read your code and tell you if you forgot some printline or something, you could actually call it a tool.
I do think it is not a good tool for that as it gets things wrong all of the time and you need to check the output constantly, but I guess it would fit the tool description even for me.
I do agree. It’s a pretty shit tool because you can’t trust it, it makes mistakes, and you’re not in complete control of what it does.
And yeah maybe it’s more like a service when you’re using it for ready-made stuff.
I think it depends on use case and intent. Calling it a tool is perfectly valid in some cases. For others it acts more like a torment nexus.
But can a torment nexus be used as a tool? Probably, but that’s supervillian shit lol
When you roll dice, you use a tool (the dice), and they’re random.
But with the dice you explicitly want the randomnes. With an llms you usually want something else.
You could technically use them for randomnes, but they are retty terrible at randomnes.
So now I have a bag of dice. I use them to predict something and roll each of them 1000 times. I don’t want randomness in that case, I want statistics.
One dice is a tool, but a bag of dice isn’t?
My point is about intent and outcome. If you intend to get statistics and roll a bag of dice the outcome is always a bunch of random numbers you could use for statistics.
If you ask an llm to give you 1000 random numbers it could just decide that “I gave you just 6es to make your calculations easier”.
A tool will give you a predicatable output, even if the output is just randomnes. You can predict that the output of 6 dices are 6 numbers from 1-6. An llm could give you anything. It could give you 6 numbers from 1-6 but it could also give you 4 numbers from 1-9 and the letter b.