Anonymous at Fri, 14 Mar 2025 16:44:40 UTC No. 16618844
>>16618841
It is impossible to believe that Meta would stop a bad idea within 48 hours. They usually run with them for years.
Anonymous at Fri, 14 Mar 2025 16:47:26 UTC No. 16618847
>>16618841
>it spewed misinformation
so it became the average scientist?
Anonymous at Fri, 14 Mar 2025 16:49:21 UTC No. 16618848
>>16618841
piping hot AI news from 2022 I see
Anonymous at Fri, 14 Mar 2025 17:10:00 UTC No. 16618855
LLMs are bullshit machines. Which works fine for some purposes but when they have to really reason and draw fine distinctions they tend to suck a lot. Moreso in 2022 when that was written.
Anonymous at Fri, 14 Mar 2025 17:12:44 UTC No. 16618858
Anonymous at Fri, 14 Mar 2025 22:17:53 UTC No. 16619095
>>16618841
If "misinformation" in reality was pointing out scientific fraud, currently estimated at 1 in 7 papers, the sure, it is quite plausible.
Anonymous at Fri, 14 Mar 2025 22:22:14 UTC No. 16619104
>>16619095
>currently estimated at 1 in 7 papers
that's just the minimum they'll admit to
Anonymous at Fri, 14 Mar 2025 22:28:20 UTC No. 16619111
>>16619095
It's not a truth-telling machine. It's not an oracle. GIGO.
Anonymous at Sat, 15 Mar 2025 07:50:33 UTC No. 16619561
>>16618847
If that were the case, they would have made millions of clones of it. More likely is that it "spewed" politically incorrect information.
Anonymous at Sat, 15 Mar 2025 08:43:55 UTC No. 16619573
>>16618847
stop watching sabine retard
Anonymous at Sat, 15 Mar 2025 13:38:23 UTC No. 16619724
>>16619111
>It's not a truth-telling machine.
It doesn't have to be. You can do a two-step process wher you first use AI to look for well known patterns that indicate fraud. Retraction Watch covers this. Then you eliminate all authors that have even once been involved in a fraudulent paper. The rest is probably honest enoug to build the LLM.
The manual process is very slow and the Alzheimer scandal has been going for nearly 20 years while the "scientists" use their personal networks to remain in positions and out of jail. Meanwhile, academic libraries are being filled up with lies, fabrication and fraud.
>>16619104
Get Elon involved and he will bring the big axe to bear on this.
Anonymous at Sat, 15 Mar 2025 14:11:48 UTC No. 16619739
>>16619724
>Meanwhile, academic libraries are being filled up with lies, fabrication and fraud.
Seems crazy that when a paper is retracted that other papers that cite it aren't automatically marked as at least being questionable. Ideally it would filter down through the tree of citations with some kind of uncertainty score. If a paper ten levels of citation up from your paper gets retracted, only a small amount of uncertainty is increased but if your paper heavily relies directly on a retracted study, the uncertainty score increases by a large margin. Any paper that cites your paper gets points added to its uncertainty score.
Anonymous at Sat, 15 Mar 2025 18:27:52 UTC No. 16619946
>>16619739
That's what you get when all the shit is behind a publishers paywall. A publisher has no interest in such a system.
Anonymous at Sat, 15 Mar 2025 18:36:22 UTC No. 16619954
>>16619739
This doesn't really respect the actual process. So most references are just nods. Pretend references. The reader skims the paper looking for x,y,z to shove in the doc.
You have to really take it up a to a higher level to talk about the accountability you want. What is another area where this happens in?
Law. Lawyers do it when looking for relevant cases to the one at hand. And they have a much higher bar given through discursive, adversarial, or even contrarian argumentation by another professional. Could this be implemented? Yes, yes it could. But then for every paper you would need the anti-paper.
There is more to say on this topic, but it is treading to an area I don't want to talk about because I don't want the noosphere stealing my IP - unironically. But there are many more solutions.
So in general, a follow methodology should only be theoretically linked to fraudulent conclusions unless the method is fraudulent itself. And in this sense, the theory is background and not so much a premise.
Anonymous at Sat, 15 Mar 2025 18:38:05 UTC No. 16619956
>>16619739
>Crazy
That's by design.
Anonymous at Sat, 15 Mar 2025 23:04:17 UTC No. 16620139
>>16619739
Yes, it is even a cascading problem. Some tools such as Zotero helps in this. ALso, Retraction Watch now has a database over retyracted papers, the volume is just too huge for a simple text file.
Anonymous at Sat, 15 Mar 2025 23:23:47 UTC No. 16620153
It said that humans are not exempt from the evolutionary pressures and subsequent outcomes that we have observed and measured in all other forms of life, aka, we are not all the same
Anonymous at Sat, 15 Mar 2025 23:27:51 UTC No. 16620155
>>16619724
Elon is already on it, Anon. Just look at what's going on at NIH and NSF. The bugs are scurrying after Elon lifted the rock and started stomping.
Anonymous at Sat, 15 Mar 2025 23:29:15 UTC No. 16620157
>>16619954
LLM generated anti-papers will not solve the problem.
Anonymous at Sun, 16 Mar 2025 00:05:11 UTC No. 16620178
>>16620155
Good. I hope he also hunts down the publishers. Their business model is crazy as it is destructive.
Anonymous at Sun, 16 Mar 2025 00:15:23 UTC No. 16620189
>>16620178
Publishers are not at fault for any of this. Their business model is successful. If the community did not like the way publishers ran things then they would start their own journals. And they do! The existing structure is successful because it works, so Elon does not have to do anything about it. The rule is if it makes money then it is correct. The market can never be wrong.
Anonymous at Sun, 16 Mar 2025 00:42:20 UTC No. 16620205
>>16620189
>Publishers are not at fault for any of this.
According to Retraction Watch, the publishers drag their feet on retracting papers. They also paywall papers long after they are supposed to be available.
>Their business model is successful.
Oh yes, definitely. They are raking in huge sums.
Anonymous at Sun, 16 Mar 2025 16:08:50 UTC No. 16620872
>>16619573
stop watching professor jew explains, and I'll think about it.
Anonymous at Sun, 16 Mar 2025 22:46:45 UTC No. 16621192
>>16618841
Imagine the shock and anger when the head poly-sexual developer at Meta asked the AI to properly name all 58 genders and it refused saying there are only two genders.
Of course they had to shit it down.
Anonymous at Sun, 16 Mar 2025 23:12:33 UTC No. 16621210
>>16621192
Can't even list dumb things in bulk...
Why we need ai for then?
Anonymous at Sun, 16 Mar 2025 23:19:37 UTC No. 16621213
>>16618841
They probably didn't prioritize research recency or filter out Zhang et als
🗑️ Anonymous at Mon, 17 Mar 2025 00:49:13 UTC No. 16621271
>>16621192
if this were real, i bet the devs got fired and had to move to another company
Anonymous at Mon, 17 Mar 2025 02:05:02 UTC No. 16621314
>>16618847
This anon gets it.
>>16619573
Don't know who that is and I don't give a fuck.