What if the problem is not that we overestimate LLMs, but that we overestimate intelligence? Or to express the same idea for a more philosophically inclined audience, what if the real mistake isn’t in overestimating LLMs, but in overestimating intelligence itself by imagining it as something more than a web of patterns learned from past experiences and echoed back into the world?
I think AI skeptics have a strong bias to assume that human intelligence fundamentally functions differently from LLMs. They may be correct, but we don't have a strong enough understanding of human cognition to make the claim in as uncertain terms as the skeptical argument is unusually made. The training methods between human learning and machine learning are obviously fundamentally vastly different as are the infrastructure-level mechanics. These elements are likely never going to align, though with time the machine infrastructure may start to increasingly resemble human bio hardware. I bring this up because these known vast differences may account for a significant portion of the differences in expected output from human and machine processing. We don't understand the fundamental conceptual "black box" portions of either form of processing well enough to state definitely what is similar or dissimilar about those hazy areas. Somewhere within that not-well-understood area is what we collectively have vaguely defined "intelligence." But also within that area are all the other aspects that both humans and now machines are quite good at - prediction, fluency, translation. The challenge of lexicon and definition is potentially as difficult a task as is sharpening the focus of our understanding of the hazy black-box portion of both machine processing as well as human processing. Until all those are better defined I don't think we have a good measure for answering the question of machine intelligence either way.
LLM's fail because their input data is limited in dimension compared to humans (text, pictures, audio, video) and because their capacity to rewire their own brains is limited to the transformer architecture.
They pivoted to regular deep learning when Jeff stepped away from the company several years ago. It does not appear they're doing much of brain modeling these days. Last publication was 3 years ago.
I could not find any official English translation of the report, but Romanian can be easily translated with online tools. This link (found in the article) seems to have more technical details https://www.bleepingcomputer.com/news/security/romanias-elec...
“The Romanian intelligence agency says that the 85,000 attacks continued until November 25th, the night after the first presidential election round, and the goals ranged from gaining access to the election infrastructure and compromising it to altering election information for the public and denying access to the systems.
SRI notes in the declassified report that the threat actor tried to breach the systems by exploiting SQL injection and cross-site scripting (XSS) vulnerabilities from devices in more than 33 countries.
The agency is also warning that Romania's election infrastructure is still affected by vulnerabilities that could be exploited to move laterally on the network and establish persistence.”
So, they can't even get the Romanian Intelligence Service to make a specific claim that some aspect of voting was affected, instead that they were "targeted by" cyberattacks? And this is enough to cancel already-held elections?
>TikTok favoring one candidate even after the campaign was over
This shouldn't be an issue because, the mainstream candidates where heavily favoured by the conventional media. If TikTok isn't allowed to publish media supportive of one candidate then the mainstream media shouldn't be either.
That's not how the law works in Romania. As a political candidate, your paid ads have to marked accordingly with a number that can be traced back and all donations and spending must be reported to a government authority.
One of the issues with this TikTok business was that many ads for Georgescu were not marked correctly, and thus in violation of electoral law.
The underlaying news is the same, but my link focuses more on the technical reason (IT system breach), while the link you posted emphasizes the political reasons
Do they vote on paper ballots or on a machine, that then logs the vote somewhere?
If they have paper ballots a recount can be trusted. If they vote on a machine they would be trusting the machine that they say was targeted by those attacks. In general my opinion has always been that voting on a machine is a bad idea.
Paper ballots, and a recount was already issued. There were differences between the two counts but they were minor enough; the recount was actually done because some other guy didn't like the fact that 2nd and 3rd place were only 2000 votes apart.
The breaches apparently didn't do much, the big reason why they cancelled the election was because the leading candidate declared 0 spending for his campaign, but it was proven he used russian money to fund countless tiktok videos that got him popular overnight. This is illegal under romanian law.
Personally, I don't think they should've been cancelled. It's a dubious thing to do under a democracy, and the runner-up was pretty decent anyway and had a chance for the second tour. I guess we'll see how it goes.
If people can only cast one vote, and you take the candidate with the most votes out of circulation (for whatever reason, campaign fraud in this case it seems), then you are discarding the votes of a large contingent of voters. Doing this and letting the results stand is wrong for the same reason why simply distributing the votes pro-rata among the other candidates is wrong: it is more likely that the candidate pulled votes from closely-aligned candidates than candidates on opposing ends of the political spectrum, so you get a skewed representation of the voting distribution.
Whichever method you choose to redistribute (or ignore) the votes cast for that one candidate doesn't matter: you will always end up in election-doctoring territory, even if you do everything by the book and in the open.
Other systems, like ranked-choice voting, might not need a do-over because relative preferences are already expressed on the ballot. But in this case, canceling the entire vote and re-doing it is the only sensible solution.
