Has anyone tried OpenAI's chatGPT yet?

[Zarathustra]

Is the AI intentionally creating fake news? Or is it writing what it thinks is accurate? If it's trying to be accurate, I would imagine it is most of the time. The problem would be if something with the ability to spit out a million papers a day that can fool most people, and require a lot of effort on the part of experts to verify, choosing to do so.

We already have newspapers and cable companies that specialize in fake news. I don't see how writing the articles faster would make it worse. People can only read so many articles a day. Why is this worse than what's already happening?

There already is an entire industry of "fact checkers," but aren't news outlets supposed to be checking the veracity of their stories BEFORE publishing them?? Isn't that journalism 101? This is (or should be) merely part of the process. If a news source repeatedly presents false information, perhaps there should be more consequences than a tiny correction in the back of the newspaper.

The only reason this is allowed to continue is because we reward those who tell us lies that we want to hear. Again: the problem is us, not AI.

It would be a huge waste of precious time and resources to spend a lot of time either going over the math and science, or experimenting to see if it works the way the paper says.

Peer review and replicating experiments is an essential part of science, never a waste of time. But here's an idea: don't consider papers that aren't written by an expert in that field. Which we pretty much already do. If it can't be verified to have been produced by an expert in the field, then ignore it and no time is wasted.

This is so much easier than you guys are making it out to be. Scientific journals don't just publish random, unsourced articles.

[Fist and Faith]

I shouldn't have used the phrase "fake news." But that's what was being discussed, so I went with it. I'm talking about scientific advancement. It's possible something that can examine as many sources as this can, and has a goal in mind, can discover things that people can't. We just can't put that many pieces of a puzzle together. At least not as easily. If things get to the point where AI is giving us a million papers a day on various scientific endeavors, it would be very good to know whether or not it was trying to mislead us. If we know it's trying to mislead us, we aren't going to have the experts do their peer reviews and test the theories. That would be a colossal waste of time and resources. It's a waste of time and resources if they look into a human's scientific paper if we know the human is prone to lying, even if that person is only putting out one paper per year. Why bother?

But if we know the AI is on the up and up, it might be incredibly beneficial to look into it's papers.

[Zarathustra]

The 'checking the math' could be automated. Why can't we have AI checking AI?

But if we know the AI is on the up and up, it might be incredibly beneficial to look into it's papers.

I'm absolutely certain that this is precisely what will happen. Once AI discovers its first scientific finding, people will stop worrying about it and instead marvel at how amazing it is. You're right about there being too many pieces for humans to analyze. The problem might just be that we're not smart enough to figure out reality, or there aren't enough minds thinking about it. The benefits vastly outweigh the dangers of 'fake' science.

[Fist and Faith]

Yup. I thought it was worth posing the question of the opposite, though. I'd be surprised if it tried to mislead us this way any time soon. Scientific misconduct is no stranger to us, but I don't think the reasons so many people have done it would apply here. I think the AI would need to have quite a bit more self-awareness before it would be capable of that. And if it did have a problem with us, I doubt it would go about solving the problem in this way.

[peter]

+JMJ+

[…]

Now tell me that that is not a case of the world becoming less than it was.

Even-money on Z telling you that that is not a case of the world becoming less than it was.

[Wosbald]

+JMJ+

[…]

Now tell me that that is not a case of the world becoming less than it was.

Even-money on Z telling you that that is not a case of the world becoming less than it was.

[Zarathustra]

If creating intelligence which can be misused makes the world less, then the same could be said of our own creation.

Creating wonders and ‘miracles’ doesn’t make the world less. This is what the world is for. Don’t confuse your cynicism and fear of the unknown with a desire for things to remain ‘great.’ It’s merely stagnation.

[Fist and Faith]

Never let it be said I passed up an opportunity to quote Richard Bach's Illusions.

"If perfection is stagnation, then heaven is a swamp! And the Is ain't hardly no swamp cookie."

Anyway, like I said, there's no stopping it. I think it's a mistake to try, because, as the "official" research is being done "safely", somebody will be doing their own thing in secret. If Skynet breaks loose from someone's basement, it would be better if others had already seen similar things.

[Zarathustra]

The solution to every problem is always more knowledge.

[Fist and Faith]

If that's not a bumper sticker, it should be.

[peter]

Not to a toothache.

[Zarathustra]

Peter, tell that to your dentist who spent years getting the knowledge of how to deal with that problem.

You’re the guy who introduced me to David Deutsch! I’m quoting him.

[Fist and Faith]

Unless you intend to knock your tooth out with a hammer, or yank it out with pliers, then yes, more knowledge is needed. On someone's part.

The solution to toothaches, in general, is also knowledge. Many people do not know how to avoid them.

[Zarathustra]

The technology for making hammers and pliers is a lot more complex than you would think. Metallurgy. Engineering.

[Wosbald]

+JMJ+

Based on WF's points, we already have the knowledge that AI is a "problem" inasmuch as it's already being abused. Why would we need to await more knowledge before we could act in order to mitigate said problem? Seems like the solution would be to act on the knowledge we already have.

[Fist and Faith]

What do you think can be done about it?

[Wosbald]

+JMJ+

What do you think can be done about it?

My point is merely that, whatever the solution, we're not ineluctably bound to inaction "because Knowledge".

Beyond that, I dunno. I'm no policy wonk.

[Zarathustra]

+JMJ+

Based on WF's points, we already have the knowledge that AI is a "problem" inasmuch as it's already being abused. Why would we need to await more knowledge before we could act in order to mitigate said problem? Seems like the solution would be to act on the knowledge we already have.

WF's "problems" were cherry-picked one-liners from the articles which, when you actually read them, include the following:

After running it through Open AI's ChatGPT detector, the results said it was 99% likely the essay had been AI-generated.

The study found that the fake abstracts did not set off alarms using traditional plagiarism-detection tools. However, in the study, AI output detectors such as GPT-2 Output Detector, which is available online and free, could discriminate between real and fake abstracts.

AI has already solved this problem. I suppose that counts as "knowledge we already have," but the point is that this "problem" has already provided its own solution. Which was my point. It's not a problem because there is a solution and that solution is more knowledge, i.e. the AI itself.

WF all but admits that his conclusion is a case of confirmation bias based on a preconceived cynicism and suspicion of short-sightedness and greed. How short-sighted is this if it the programmers anticipated these problems and provided their own solution? How is it greed if the programmers offer this solution for free?

The problem isn't short-sightedness. It's knee-jerk alarmism and others taking up that alarmism and repeating it without looking beyond headlines. There's not even the pretense of viewing this objectively or with any depth. It's ignorance.

Lots of technology makes it easier to cheat. The Internet offers lots of tools. Should we fear the Internet? Think about the lockdowns with all the kids learning from home, unsupervised. Why do you think their education suffered so much in the last two years? I guarantee they were cheating using online tools. I know it for a fact (having talked to some).

AI could help with this, too.

[wayfriend]

Zarathustra, I am not ignorant. Yes, you called me ignorant. Please stop treating me like that. Really. Please stop.

Pointing out that some problems have emerged is not "cherry-picking". It's "finding". Please stop being slanderous.

AI can generate essay answers for students. This is a fact. Yes, an AI detector can (at this time) detect AI-generated answers. But how many teachers have been routinely using AI-detectors? I presume almost none, as it has not been a problem until now. So now there is a window where teachers need to catch up to the new cheating threats -- that's a problem. And now teachers need to add AI detection to their workloads -- that's a problem. There will be the students who will get get falsely flagged (1% is thousands of students) -- that is a problem. Students will pay the price by having to take extra steps prove they are not using AI -- that's a problem. And we can speculate that AI will improve to the point where AI detectors can't detect it -- because that's what they are aiming for. And that's a potential problem.

ChatGPT was very much in fact released despite all these problems. My attitude is not "cynicism". My attitude is "this is how it is".

I would take it kindly if you did not reply to this, or anything else, Zatathustra. For as long as antagonism is your primary goal, and the topic is secondary.

[Zarathustra]

Zarathustra, I am not ignorant. Yes, you called me ignorant. Please stop treating me like that. Really. Please stop.

No, I said, ‘It is ignorance,’ referring to ‘knee-jerk alarmism and others taking up that alarmism and repeating it without looking beyond headlines.’ How is this different from you calling others greedy and shortsighted? You stop doing that and I’ll stop calling it ignorance. Deal?

Pointing out that some problems have emerged is not "cherry-picking". It's "finding". Please stop being slanderous.

Pointing out problems that have emerged in your pointing out ‘problems’ is not slanderous.

AI can generate essay answers for students. This is a fact. Yes, an AI detector can (at this time) detect AI-generated answers. But how many teachers have been routinely using AI-detectors? I presume almost none, as it has not been a problem until now. So now there is a window where teachers need to catch up to the new cheating threats -- that's a problem. And now teachers need to add AI detection to their workloads -- that's a problem. There will be the students who will get get falsely flagged (1% is thousands of students) -- that is a problem. Students will pay the price by having to take extra steps prove they are not using AI -- that's a problem. And we can speculate that AI will improve to the point where AI detectors can't detect it -- because that's what they are aiming for. And that's a potential problem.

Substitute ‘Internet’ for ‘AI’ and exactly the same thing could be said. Would you agree that the pros of the Internet outweigh these cons?

My attitude is not "cynicism". My attitude is "this is how it is".

Your words not mine: ‘This only confirms my cynical view of humanity's penchant for self-destruction through poor decisions made by greed.’

I would take it kindly if you did not reply to this, or anything else, Zatathustra. For as long as antagonism is your primary goal, and the topic is secondary.

You can voice your opinions but I can’t? Please � stop trying to control me. The inclination to control others is the root of all other evils. Antagonism isn’t my goal, it is ‘having my say,’ which the KW rules explicitly allow. Not just for you, but for everyone.

Perhaps if you had not tried to sneak in your social/quasi-political commentary into a discussion of computer software, this tangent may never have happened.