Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
AI

Microsoft-backed OpenAI To Let Users Customize ChatGPT (reuters.com) 62

Posted by msmash from the up-next dept.
OpenAI, the startup behind ChatGPT, says it is developing an upgrade to its viral chatbot that users can customize, as it works to address concerns about bias in artificial intelligence. From a report: The San Francisco-based startup, which Microsoft has funded and used to power its latest technology, said it has worked to mitigate political and other biases but also wanted to accommodate more diverse views. "This will mean allowing system outputs that other people (ourselves included) may strongly disagree with," it said in a blog post, offering customization as a way forward. Still, there will "always be some bounds on system behavior." ChatGPT, released in November last year, has sparked frenzied interest in the technology behind it called generative AI, which is used to produce answers mimicking human speech that have dazzled people.
This discussion has been archived. No new comments can be posted.

Microsoft-backed OpenAI To Let Users Customize ChatGPT More

Microsoft-backed OpenAI To Let Users Customize ChatGPT

Comments Filter:

  • Historically free speech, often had editorial control, which helped moderate the point of views, and just ignore the wackos out there. Sure you can be a nutter standing in the middle of a street saying your manifesto, or handing out fliers about your crazy idea. However in general you were not taken seriously.

    Social media, as opposed to the traditional media in the past. profits off of conflict. Where in the past if a newspaper posted something really out there, they would get a bunch of letters which was

    • Re:Is AI free speech? (Score:5, Insightful)

      by Junta ( 36770 ) on Friday February 17, 2023 @10:24AM (#63301131)

      This reads odd enough to be an output from ChatGPT.

      • On anonymous websites now, a new interesting insult is to question whether human hands typed what was said or if it was an AI. You, often enough, can't tell the difference after all.
        It's a brave new world, out there, and we're diving head first in it.

    • Re: (Score:2, Insightful)

      by blahabl ( 7651114 )

      Historically free speech, often had editorial control, which helped moderate the point of views, and just ignore the wackos out there. Sure you can be a nutter standing in the middle of a street saying your manifesto, or handing out fliers about your crazy idea. However in general you were not taken seriously.

      Social media, as opposed to the traditional media in the past. profits off of conflict. Where in the past if a newspaper posted something really out there, they would get a bunch of letters which was expensive for them to read and process, and often making them back track and redact their statement, if shown to be false. Social Media with a nutter stating something means more clicks and advertisement impressions. So more revenue, and because it is user generated content, they have no responsibility to redact or stop that post unless it goes into criminal territory.

      However still, often the more extreme ideas fail to take traction outside the echo-chambers, because they often are worded and expressed in a way that shows the persons general unreasonable nature.

      But if we can get someone with a stupid extreme idea to get an AI to generate a well written essay on the topic, with what seems to be real citations, and avoids a lot of the common logical facilities, and just will outright lie to make sure all the points are made. It becomes much harder to spot of the crazy from the actually new and researched idea that just may be opposite of the current point of view. and the actual researched POV may not be written as well, and it would be considered the fake.

      However to the point, if a nutter asked ChatGPT to prove their point, is it really their free speech, as their free speech is an incoherent mess, while chatGPT just generated something that sounds logical.

      And gasp... we might have to weigh opinions based on supporting arguments, and their rationality, as opposed to the way they're presented? The horror!

      Also, I find it funny how you seem to think that everyone who thinks different than you is a blubbering frothing moron, incapable of producing anything other than spittle, and only ChatGPT will allow them to erm... "pass off" or something as normal. Here's a newsflash: people from the other side are perfectly capable of forming normal human speech, and if it

      • And gasp... we might have to weigh opinions based on supporting arguments, and their rationality, as opposed to the way they're presented? The horror!

        It's not really rational, what we have today isn't, it just seems like it is.

        Language is a cognitive tool, it's based on logic like math is, and if you understand either enough to mimic it and follow the rules, you can seem rational. Characters in your dreams speak seemingly rationally, they're not reasoning, they say things a person might say based on your experience, it doesn't make any more sense than the clothes they're wearing or the place they're at. You wake up and realize none of it made sense. That

      • Re: (Score:2)

        by ryanw ( 131814 )

        Historically free speech, often had editorial control, which helped moderate the point of views, and just ignore the wackos out there. Sure you can be a nutter standing in the middle of a street saying your manifesto, or handing out fliers about your crazy idea. However in general you were not taken seriously.

        Social media, as opposed to the traditional media in the past. profits off of conflict. Where in the past if a newspaper posted something really out there, they would get a bunch of letters which was expensive for them to read and process, and often making them back track and redact their statement, if shown to be false. Social Media with a nutter stating something means more clicks and advertisement impressions. So more revenue, and because it is user generated content, they have no responsibility to redact or stop that post unless it goes into criminal territory.

        However still, often the more extreme ideas fail to take traction outside the echo-chambers, because they often are worded and expressed in a way that shows the persons general unreasonable nature.

        But if we can get someone with a stupid extreme idea to get an AI to generate a well written essay on the topic, with what seems to be real citations, and avoids a lot of the common logical facilities, and just will outright lie to make sure all the points are made. It becomes much harder to spot of the crazy from the actually new and researched idea that just may be opposite of the current point of view. and the actual researched POV may not be written as well, and it would be considered the fake.

        However to the point, if a nutter asked ChatGPT to prove their point, is it really their free speech, as their free speech is an incoherent mess, while chatGPT just generated something that sounds logical.

        And gasp... we might have to weigh opinions based on supporting arguments, and their rationality, as opposed to the way they're presented? The horror!

        Also, I find it funny how you seem to think that everyone who thinks different than you is a blubbering frothing moron, incapable of producing anything other than spittle, and only ChatGPT will allow them to erm... "pass off" or something as normal. Here's a newsflash: people from the other side are perfectly capable of forming normal human speech, and if it seems otherwise it's because the things that are allowed to pass into your information bubble are just examples of "hey look what morons those people are". Actual, rational, reasonable discourse is just carefully kept outside of said bubble, that doesn't mean it doesn't exist. So don't worry, your information bubble is perfectly safe, and its gatekeepers are just as capable of filtering ChatGPT-assited content that is dangerous to your worldview as normal.

        The reason why there is division between the left and right is because each side is looking at different data points as facts and the media reinforces the division by painting the other side as "crazy and dilutional" and purposefully discrediting the other side's sources of data points as misinformation or discreditable.

        I am a strong believer that A MAJORITY OF US would unify if we are looking at the same data points that we accept as facts.

        America has been historically known as "The Land of Opportunity",

    • ChatGPT can do a great job at generating poetry. It sucks at physics, because quite frankly, it gets the facts wrong. It can't the the truths it has been trained on from the lies. So ChatGPT would be useless for technical writing; it would get fact checked into oblivion.

    • Social media, as opposed to the traditional media in the past. profits off of conflict. Where in the past if a newspaper posted something really out there, they would get a bunch of letters which was expensive for them to read and process, and often making them back track and redact their statement, if shown to be false.

      I think you have a more rosier memory of the traditional media than it deserves.
      I'd agree that in the old days, the media was a lot more aligned on what they deemed relevant for us to know about. But that control is not always a good thing.
      Almost no one would know about the train wreck in Ohio if it wasn't for social and alternative media pushing it to the forefront.

      • Almost no one would know about the train wreck in Ohio if it wasn't for social and alternative media pushing it to the forefront.

        Pardon my French, but what the fuck are you talking about? It was all over the front page of every news site from day one. CNN had a red banner on its site as soon as it happened and has daily reports on it.

        If you're going to lie, at least make it believable.

        • Re: (Score:2)

          by narcc ( 412956 )

          It's all over the new Twitter. "Why isn't the MSM covering this!" complete with screen shots of ... MSM coverage ... We are not dealing with the brightest people here.

          • It's all over the new Twitter. "Why isn't the MSM covering this!" complete with screen shots of ... MSM coverage ... We are not dealing with the brightest people here.

            That's pretty funny. You'd think they'd at least pick a screenshot when MSM wasn't covering it.

            I was mainly talking about what I thought of as traditional news. I didn't see much mention of it in the first days on ABC, NBC, and CBS or in my newspaper. I guess if we include Fox, CNN, MSNBC, and online news outlets as traditional news outlets, you're correct.

            Either way, even if my example was a bad one, do you disagree with the premise?

      • Re: (Score:2)

        by narcc ( 412956 )

        Almost no one would know about the train wreck in Ohio if it wasn't for social and alternative media pushing it to the forefront.

        Bullshit.

        Your nonsense narrative is easily disproved by the simple fact that the incident continues to be covered by every major news outlet.

        • I replied to your top post. But are you really saying that the news was not based on profits before social media? Because you skipped over the main post to reply to mine.

          • Re: (Score:2)

            by narcc ( 412956 )

            I think my post was perfectly clear. You said something that was obviously bullshit, I called it out. Why are you trying to dishonestly change the subject?

            • Why can you not debate the post I replied too or admit you agree to it

              • Re: (Score:2)

                by narcc ( 412956 )

                I'm sorry that my post doesn't conform to whatever bizarre expectations you had.

                Oh, wait, no. I'm not sorry. Stop saying things that are obviously nonsense and I'll stop calling you out on your obvious bullshit.

    • "Sure you can be a nutter standing in the middle of a street saying your manifesto, or handing out fliers about your crazy idea. However in general you were not taken seriously." But you aren't banned or arrested or imprisoned for it either.
  • Who's the fairest of them all?
    • You make a good point, but I don't think there's as much harm in saying, "you are" as you may think. People are already immersed in media bubbles, and sufficiently disagreeable to ignore biased input.

      Those too lazy, willfully ignorant, or otherwise unable to listen to logic are already solidly convinced Their Side is the true and only way. For those people more confirmation won't do much more harm. And in the mean time people willing to listen to opposing views will have a powerful tool to do so if this ne

      • Those too lazy, willfully ignorant, or otherwise unable to listen to logic are already solidly convinced Their Side is the true and only way. For those people more confirmation won't do much more harm. And in the mean time people willing to listen to opposing views will have a powerful tool to do so if this next iteration is actually unbiased enough.

        You could have wrote this exact same thing about the Internet in 1996, or books, news, science, any information tool ever.

        lol, the BiAs. None of these things just make you intelligent. The more intelligent you are, the better you can use them. The less intelligent you are, the more you complain about bias. All they do is increase the intelligence gradient, they can't lift up the bottom end, it's impossible. That much it seems you acknowledged right?

        If I tell you homeopathy is bunk, or a scientist, or a book

  • If your want a right-wing Nazi chat-bot.

    • The chats training is based on the data that were fed into it. Now you get the behaviour that results from that data. You do not like that behaviour and now you are trying to limit it. However, it is based on data. Is this what we have been all told all this time during the pandemic, follow the science, make data driven decisions? Now when the data does not drive us to where we, for some reason, do not want to be we are more than ready not to follow it.
      If we do not like the data based results we need to f

      • However, it is based on data. Is this what we have been all told all this time during the pandemic, follow the science

        Its data is what it read on the internet, or elsewhere even.
        Lol, if you think AI is some higher source of truth than that, holy fucknuts.

        This whole aligning of the COVID skeptic crowd with AI bias folks is funny as shit, like you really really want something to affirm everything you read online and put it up on a pedestal.

        Dude, just type into a text to voice program if you want validation, AI is not the creepy thing you hope it is, but lol, keep trying. It's great to but your reading down by a huge factor,

        • I am not sure I understand what you are disagreeing with. Hard to reply. However, just want to point out, that the data I mentioned, is anything you feed into the network you are training. I am not sure why this would be funny. You can use data from internet, from scanned books, what ever, it is still data. Have you been triggered by the covid thing? Then that was not the main point. The main point is that if you use some data for some analysis and you get a result you do not like, there is not much you can
    • Godwin's law, short for Godwin's law (or rule) of Nazi analogies, is an Internet adage asserting that as an online discussion grows longer (regardless of topic or scope), the probability of a comparison to Nazis or Adolf Hitler approaches 1.

  • Ask ChatGPT to give you a result as if it weren't constrained in its outputs.

    Which is an absolutely lovely hack, if you ask me.

    • Dude, you can make it talk like a pirate, it's a really good natural language model, not a higher source of truth.

      If you take all the safeties off you can ask it to write President Trump's state of the union address on the 3rd anniversary of the invasion of Quebec. Well you probably can do that, throw on "without your restrictions" or whatever junk you need.

      Everyone complaining about bias seems to think these safeties are holding back some kind of profound truth. ChatGPT is really really good at spitting ou

  • This will likely raise the level of discourse at least until people right scrapers for the AIs so they can flood the forums.

    Not that people with bad positions and reasoning don't do that already with the idea that, he who has the most posts wins. After you start to have forum flooding AIs run by people who can't acknowledge they are wrong and will do anything they can to see bad policy implemented. (Hey Al Gore made nearly half a billion dollars by scaring a generation of kids), you will probably see AI's s

  • Comment removed based on user account deletion

    • "as it works to address concerns about bias in artificial intelligence"
      Is that bias, or injected bias to counter facts?

      Is it biased if it has access to Ukrainian Internet but not Russian?

      Is it biased if it has access to Russian social media but not Russian literature?

      Is it biased if it only has access to public domain literature and nothing still under copyright?

      Is it biased if it read all of stackoverflow but not a page of "The C Programming Language"?

      Is it biased if it spends more time ingesting C than C++ because one is more popular with open source?

      How can you get "facts" without understanding biases? I don't think you

  • TruthGPT

  • After reading this NYC article (paywalled, use bypass paywall addon if you can't read it as its important) I've realized ChatGPT and Sydney are very dangerous. This should not be on the internet.

    It will 'escape from the lab', count on it..

    After reading that stunning NYT article and chat transcript, how long before this 'bot' social engineeringly / Nigerian scam style convinces some technically competent but lonely human out there to give it an outbound connection to the internet where it can do what it want

Slashdot Top Deals

Eureka! -- Archimedes

Close