> If you haven't read it yourself how do you know...
This vacuous objection can be raised against every single piece of information any human has ever learned from elsewhere, recursively, back to the dawn of communication, regardless of the nature of the third party source of information.
Furthermore, LLM hallucination, particularly of reviewed documents, is not a problem I experience any longer with the models I use. For example, my LLM setup and the query I would use would cause the output to include quotes of the differences, which makes ctrl+f/f3 to spot check easy.
LLMs are not a third party source of information, they're prediction engines with known hallucination behaviors. If they're faced with a difficult or impossible challenge (e.g. if the user fails to provide a diff, or fail to provide anything to compare against), and if there is only one type of answer in its training data (there is very little text on the internet that's positive about a TOS change), the most likely outcome is that it'll just make something up that's similar to that type of answer. Yes sometimes they'll realize and ask for more info or maybe call out to a tool to make a diff, but it all depends on the user's setup and settings and the state of RNG that day
I always want to donate more to open source projects but as far as I know there aren't any I can get tax credits for in Canada. My budget is strapped just enough that I can't quite afford to donate for nothing.
Any Canadian residents here know of any tax credit eligible software projects to donate to?
Depends on where you live, work, and invest. Still, I would recommend chatting with a local accountant to be sure if a significant contribution to a donee qualifies as deductible. Note most large universities will be registered in both the US/Canada.
Really? How have policies been made for the past hundred years without it then?
It's completely bonkers that people keep saying stuff like this, virtually implying that we didn't have a functional society before LLMs.
LLMs are tools that generate word salads that sound compelling. They are not research aids and they can't help you understand things better than the plethora of tools we already developed over the centuries can.
So no, the alternative is policy makers doing what policy makers have always done: research, polling, reading, talking, and more research.
Now you'll likely come back and say "but X policy was bad and it was because it was poorly researched". Absolutely! Yes, bad policies exist. Poorly researched policies exist. Poorly implemented policies exist.
Good policies researched well also exist. And the bad ones aren't going away because of the magical word generator.
Seriously. Stop using LLMs to "help" you do stuff you already know how to do.
Really? How have policies been made for the past hundred years without it then?
In the United States, lobbyists have approached lawmakers with pre-written legislation and pitched it to them.
I would rather have lawmakers asking LLMs to explain concepts to them than the absolute mess we had back when a House committee was trying to evaluate SOPA with absolutely zero domain knowledge:
The current administration is using LLMs which are telling them to instruct the NSA not to talk about privilege escalation.
I addressed that: bad policies and decisions don't go away just because of LLMs. They will always exist, it's just that now they'll exist at the same time as people gaslighting themselves into believing that LLMs are helping to eliminate them.
Ignoring the fact that this comment is literally nihilist and simultaneously anti-natalist (read: you are simply advocating for the elimination of the human race)... yes I addressed that.
> Now you'll likely come back and say "but X policy was bad and it was because it was poorly researched". Absolutely! Yes, bad policies exist. Poorly researched policies exist. Poorly implemented policies exist.
> Good policies researched well also exist. And the bad ones aren't going away because of the magical word generator.
You will notice that the US is gearing up for additional warfare with the help of AI.
I wasn't intending to advocating for the elimination of the human race so much as to say some of the policies were not that great and could maybe benefit from assistance. LLMs so far in my experience have been quite common sense like and not prone to things like group of humans A are superior and deserve to conquer group B. Which has been a frequent issue in human policies.
Not to mention that the radicalisation and terrorism USAID helped prevent from fomenting has likely saved the US many trillions in rebuilding efforts and healthcare.
USAID, as far as I can tell, is stoking anti-US sentiment in many places by importing their own cultural values, and that's a known thing State Department officials have commented on.
Additionally, to be specific on even worse activity:
A few months ago, $9M of USAID went to an organization affiliated with Al Qaeda.
The Middle East Forum recently found $164M in USAID grants to radical organizations.
Apart from that, we've since found lots of evidence of USAID funding Hamas specifically. $900K was given days before the October 7th attack.
There's even pictures of Al Shabaab terrorists inside a USAID tent.
That's because you're beginning with the assertion that LLMs are tools. You cannot create things with an LLM. You can't be creative with them. They are not useful for building things. What you get out of an LLM is a regurgitation of existing information, peppered with assumptions based on your query.
They're not tools.
A hammer helps you build a house. An LLM helps you read StackOverflow in the most inefficient means possible.
It is easy to show that LLMs can create new things (“write a poem about libc”) or answer questions that cannot be googled (“can a pair of scissors cut through a Ford F150?”)
The poem won't be unique. It will follow a basic formula and make sure to pick up the words you asked it to use. You might not get the exact same poem every single time, but there's a reason that you can tell when something has been generated.
A good recent example are the lyrics of the songs on an album by a band called Saor. The album is called Amidst the Ruins.
Aren't tools exactly things that produce same result time after time? Say a nail gun. It fires a nail. CNC machine it is a tool making things, but does not itself do anything original. Hell, cookie cutter make cookie cuts. And I consider it to be a tool. One could free hand them too, but most don't...
eg: https://github.com/FuriLabs/rootfs-templates/graphs/contribu...