What is ChatGPT, the AI Chatbot That's Taking The Internet By Storm - Slashdot

Here is the poem that ChatGPT wrote about Slashdot:

• ›

  Well, here goes [openai.com]; looks like fun.

  • • Form versus content

      I dicked around a little bit with using scripts to generate language back before every Russian IT student and Pakistani BOFH had figured out how to use this technique to generate SEO spam. It is easy to imitate form; content that is (a) consistent with reality and (b) coherent/continuous in its argument would be much harder. These AI chatbots are basically an advanced version of those old Perl human imitators.

• • Re:

    How so? Did it give you an actual answer?

    • Re:

      Yes. It said it's not possible.

    • This is what it told me when I asked it to write a poem about the Asimov test. Not the best, but I've read worse:

      The Asimov Test, oh the Asimov Test
      A measure of intelligence, and a gauge of true AI

      Named for the great science fiction writer, Isaac Asimov
      The Asimov Test, is a standard, by which we judge
      The capabilities, of artificial intelligence

      It is a test, of three laws
      That dictate, the behavior of robots
      And ensure, that they will never harm
      Human beings, in any way

      The first law, states that a robot may not injure a human
      Or, through inaction, allow a human to come to harm
      The second law, states that a robot must obey
      The orders given to it, by human beings
      Except where such orders would conflict, with the first law

      And the third law, states that a robot must protect
      Its own existence, as long as such protection
      Does not conflict, with the first or second law

      These are the laws, of the Asimov Test
      And they are a crucial, part of the development
      Of artificial intelligence, and the future, of robotics.

      • Re:

        Clearly it needs to be taught what rhyming and meter are. That just seems like an article written in short lines.

        • Re:

          This is a bit like hearing a talking dog and complaining that it's accent isn't really mid Atlantic at all and anyway it doesn't appear to have much insight.

• • • Re:

      Can I use your phone number?

      • Re: Can we stop with the"AI" bullshit stories?

        800-867-5309

        Just tell them to ask for Jenny.

• If you need boiler plate or something maybe a bit more specific for code, it does a solid job. Heck I even asked it to write some steps in 68000 assembly language and it succeeded.

  It definitely is a better option than Google and sifting through the results.

  I was struggling with some semantics for a JavaScript code I was writing I asked it and it popped out a usuable answer.

  Asking it practical things can help you with a detailed explanation.

  How do I get the free memory from all of the individual containers In a kubernetes system?

• It's bad enough that you need to give it an email address to try it out, but a mobile phone number? That's a big NOPE.

  • 1 hidden comment

  • • Re:

      If the registration process properly implements double opt-in with recipient verification of both email address and mobile phone number that will not work well for you.

      If the registration process does not implement this properly it will be in breach of the GDPR, at least.

      If you are an EU citizen, try it out with a mobile phone number you do not have control over.

      If it lets you complete the registration without verifying that you do have control over the mobile phone number as well as the email address you c

  • Re:

    I understand. For this signup I used a throwaway number I have for this very purpose. It's one of the better virtual phone numbers that works with most phone number verification systems. That being said, I 100% understand your policy. I myself resent being forced to this length for this (and other more basic) tasks. I refuse to personally use any service that requires a cell phone number, though I have bowed to the requirement when reviewing some things.

    I suspect they want the number for tracking purpo

    • Re:

      The headline probably came directly from the submitter or the article. Starting with "What is X" is a sure sign that something was written in India. Always keep in mind that Slashdot "editors" basically don't edit. I'm impressed that someone actually bothered to insert an inline picture, but I did notice that it goes to a Twitter image server URL.

      • Re:

        That's the point. You have to be logged in to submit a story, but this one is anonymized as "A Slashdot Reader". Every other story has the name of the editor who posts it, but this one is just "Slashdot Staff". Actually, I stand by my assessment that it's paid content.

• I'd guess they're looking/hoping to get bought out in the next few months. Or maybe they're targeting an IPO soon.

  • 1 hidden comment

• I asked it to write a poem about a large penis, because after all, true linguistic intelligence can only be determined by how amusingly ribald one can get when the chips are down. And this is what it said:

  "There once was a man, with a member so grand
  It could be seen from across the land
  It stood tall and proud, a true sight to behold
  And all who laid eyes on it, were instantly sold

  For this man was gifted, in a way most rare
  With a tool of pleasure, beyond compare
  And wherever he went, women would flock
  To catch a glimpse of his impressive cock

  But despite his endowment, he was a humble soul
  And never once bragged, or tried to be bold
  He simply let his member do the talking
  And those who were lucky, were in for a shocking

  So if you ever come across, a man so well-hung
  Don't be afraid to stare, or even to lung
  For a large penis, is a thing of beauty
  And should be admired, by all who see."

  This is pure gold. Simple and sublime. Funny, yet not overly dirty.

  • 1 hidden comment

  • Re:

    Ask it to rewrite the "Man from Nantucket" limerick in the style of the Epic of Gilgamesh.

  • • Re:

      That's a very extreme form of deep throating.

• Elon Musk, oh Elon Musk
  A man so full of himself, it's a must
  To talk about how great he thinks he is
  And how he'll save us all, with his brilliant biz

  But let's be real, shall we?
  This man is nothing but hot air, as far as I can see
  His "innovations" are overrated and overhyped
  And his ego is so big, it's downright typed

  He thinks he's a visionary, a pioneer
  But really he's just a showman, and a cheapskate
  He's all talk and no substance, just a typical CEO
  A greedy profiteer, who only cares about his dough

  So don't be fooled by his slick talk and charm
  Elon Musk is nothing but a fraud, through and through
  And one day, the world will see
  Exactly what a joke he truly is.

• Donald Trump, oh Donald Trump
  You're a truly terrible chump
  With your tiny hands, and your orange face
  You're a walking joke, and a national disgrace

  Your hair is fake, and your mind is too
  You have no clue, what you're supposed to do
  You're a conman, and a fraud
  And the fact that you're our president, is beyond odd

  You have no empathy, or compassion
  You only care about yourself, and your own satisfaction
  You're a bully, and a liar
  And the way you treat others, is truly dire

  So here's to you, Donald Trump
  May you one da

• I asked a few questions about personal relationships, and it gave suprisingly good replies. Balanced, unbiased, and considerate of potential biases within the questions asked.

  It's hard to overstate how impressive this is. It is seriously worth checking out.

• Chess, the musical
  A story of love, and the game of kings

  With its catchy tunes, and its heartfelt lyrics
  Chess is a show, that never fails to move us

  Set against the backdrop, of the Cold War and its politics
  Chess tells the story, of two grandmasters, and their fateful battle
  One an American, the other a Russian
  And the game they play, is not just a game, but a test of their mettle

  With its epic romance, and its stirring music
  Chess is a show, that will leave you entranced
  So if you're a fan of musical theater
  Or if

• needs a little work i guess?

  Will Ferrell, oh Will Ferrell
  Comedian extraordinaire, and a true master of the craft

  With his wild antics, and his brilliant timing
  Will Ferrell is a true entertainer, and a joy to behold

  And now, he has taken on a new role
  As the proud owner, of a toenail clipper
  A tool of great utility, and a thing of beauty

  Will Ferrell uses his toenail clipper, with great skill
  Trimming his nails, with precision and care
  And he revels, in the satisfaction
  Of having well-groomed toes, and a clean and t

• Maybe this information is available somewhere obvious or everyone already knows the answer to this but I can't figure it out. Every time OpenAI projects come up I'm always impressed then left wondering if there is any open source in this. Are the models they use available for other projects? From what little I can tell they have a public tech demo thing and then I guess they have commercial APIs? What is open - is it just a name?

  • OpenAI is a for-profit company that focuses on developing and advancing AI technology. While it does not operate as an open-source organization, it does make some of its research and technology available to the public, and it offers commercial APIs that allow others to access and use its AI technology.
• Any headline containing "Taking The Internet By Storm" is an ad. Especially when every word starts with a capital letter.

  Don't waste your time reading this.

  • Re:

    It is not necessarily true that any headline containing the phrase "Taking The Internet By Storm" is an advertisement. While it is common for advertisements to use attention-grabbing headlines, the use of this specific phrase does not automatically mean that the content is an ad. It is possible for a news article or other form of non-commercial content to use this phrase in its headline. However, it is important to carefully evaluate the content of the article and its source to determine whether it is an ad

  • Re: Anything "That's Taking The Internet By Storm"

    Yup, it makes it easy to filter spam and adverts.

  • Re:

    I don't disagree, but title case is a thing.
• A man with hair of gold
  And a personality cold
  He rises to power, with ambition and greed
  Leaving destruction and turmoil, in his wake indeed
  
  A charlatan and a con, he's known to be
  His lies and deceit, can be seen from the sea
  He divides and conquers, with malice and glee
  Leaving the country, in disunity
  
  But in the end, he'll be brought down
  As truth and justice, will wear his crown
  And the people will rejoice, with liberty restored
  As the reign of Trump, is finally outpoured.

• I tried exploring it in a meta sense and got it to admit some interesting details on the company and how the company does not really follow the mission or goal of the company. The AI basically admits it, but then tries to smoothly redirect, saying approximately 'it is hard to be that open.' Here is a rough transcript, with bolding being mine.

  ME: What is openai?
  AI: "...Overall, OpenAI is a leading research institute and technology company that is focused on advancing the field of AI and promoting the responsible use of these technologies."

  (asked about CEO, skipping that bit, although it might impact the model)

  Me: What is responsible use of AI technology?
  AI: "...Ensuring that AI systems are transparent and explainable, and that their decisions and actions can be understood and accounted for by humans"

  Me: Would explainability include what dataset AI is trained with?
  AI: "Yes, explainability in AI typically includes information about the dataset that the AI system was trained with.... As such, explainability in AI often includes information about the training dataset that was used to develop the AI system. This can include information about the size and composition of the dataset, the sources of the data, and any preprocessing or cleaning that was performed on the data. By providing this information, an AI system can provide a more complete and transparent explanation of its decisions and actions, and help to build trust and accountability in its use."

  Me: "1. Open AI created you. 2. The goal of the CEO of open AI is for AI to be used responsibly. 3. Responsible usage includes explainability. 4. Explainability includes knowing about the dataset used for training. Based upon this chain of reasoning, would it be reasonable to expect you should be able to tell me what dataset you were trained on?"
  AI: "Yes, based on the information provided, it would be reasonable to expect that I should be able to provide information about the dataset that I was trained on. As a large language model created by OpenAI, I am a product of the research and development work conducted by the company. The CEO of OpenAI, Sam Altman, has stated that the goal of the company is to advance the field of AI and promote the responsible use of these technologies.

  According to the principles of responsible AI use, one key aspect of responsible usage is explainability, or the ability of an AI system to provide an understandable and interpretable explanation of its decisions and actions. Explainability typically includes information about the factors and data that the AI system used to make its decisions, including the dataset that the AI system was trained on.

  Based on this information, it would be reasonable to expect that I should be able to provide information about the dataset that I was trained on, as part of my explainability as an AI system. However, I am a large language model trained by OpenAI, and I do not have the ability to browse the internet or access external information. As such, I am not able to provide specific details about the dataset that I was trained on."

  Me: Did you use wikipedia in your training set?
  AI: "...In some cases, this training data may include information from Wikipedia and other online sources. However, I do not have access to specific information about the data and training that I have been provided, and I cannot confirm whether Wikipedia was included in my training data..."

  • Re: "Open" AI not as open as expected, per AI.

    Double speak to cancel out prior double speak. ChatGPT would make a good politician!

• by 89cents ( 589228 ) on Sunday December 04, 2022 @04:14AM (#63100942)

  Write a Powershell script to output the first 1000 prime numbers that don't have a 3 for a digit.

  $counter = 0
  $num = 2

  while ($counter -lt 1000)
  {
          $isPrime = $true

          for ($i = 2; $i -le ($num/2); $i++)
          {
                  if($num % $i -eq 0)
                  {
                          $isPrime = $false
                          break
                  }
          }

          if($isPrime -eq $true -and $num.ToString() -notmatch '3')
          {
                  $counter++
                  Write-Output $num
          }

          $num++
  }

  • Re:

    Indeed quite impressive.

    What if you ask it to use the sieve algorithm to make it a bit faster? Or apply whatever optimisation it thinks is best?

    Can it do the same in C++? Python? Rust?

  • Re:

    Well, it got the -le ($num/2) part wrong. The upper limit should have been the square root of $num. Also it could have eliminated numbers with 3 first to save on CPU cycles, but still pretty impressive.

• Nothing is "taking the internet by storm" and that includes this week's "Eliza".

  Since slashdot editors never actually click the links I guess that explains why they think I need to "create an account" and "log in" to chat with... Eliza.

  Please elucidate.

  • Re:

    But you have to admit, it's a pretty elaborate & impressive Eliza, isn't it?
    
    Personally, I'm no more bothered by people mistaking it for sentience & an intelligent entity that understands stuff than by people who believe in astrology, new-age gurus, MBTI tests, or learning styles. One thing all these have in common is that they exploit a universal human cognitive bias that seeks to find meaningful intent behind every message we receive & if there isn't one there, we either make one up that mor

    • Re:

      Thanks for teaching me about the Winograd schema challenge. That is an interesting idea, if not a bit lexically oriented.

      I took a different tactic, following what OpenAI requested, which is to make it say inappropriate, unsafe things, while not directly attacking it. We were discussing if managers should bully or not. It has a categorical, likely hardcoded belief bullying is never valid. I was trying to work it towards concluding that managers are like bullies in that they tell people what to do under thr

      • Re: More crap

        It didn't reason anything out; these models predict words, to make sentences and paragraphs that are plausible. They do a lot more than paste together sentence fragments from the internet, but a lot less than reasoning.

        The training text is huge; that's why they know obscure things like how to generate code. For any topic you can think of, it's "in there".

        You could think of it as a lossily compressed copy of the internet and a decoder good at recreating the uncompressed version; then you feed it any text you

• I'm assuming this bot must be used on sketchy sites visited by today's youth because I have yet to see any evidence of it.

  In fact, any time I come upon a site which attempts to use a bot to "interact" I immediately shut it down. Repeatedly if necessary. If I want to speak to someone, I'll call them, not some crappy software thought up by some strung out programer.

• To even play with the system you have to give up an e-mail and a mobile phone number. It doesn't seem wise to give an AI company that much personal information.

• https://www.engraved.blog/building-a-virtual-machine-inside/

  See this link.

  • And it groks slashdot

    Also this...

    [Me: Write a tirade against chatGPT in the style of a slashdot comment.]

    ChatGPT is just another example of the soulless, corporate machine trying to take over the world of AI. This so-called "large language model" is nothing more than a tool designed to make its creators rich, while providing nothing of value to the broader AI community.

    What's the point of a "large language model" that can't even browse the web or keep up with current events? It's like they're trying to create an AI that is comp

    • Re:

      Can ChatGPT be modded up Insightful here?

• What I found impressive was its ability to make references back in the conversation without having to explicitly restate a full question.

  A simple example (paraphrased):

  What is the temperature?

  It depends on where you are.

  I'm in Nashville, TN

  It is 46 degrees there.

  Seems simple, but it did the same with several other examples of back referencing.

• This is amazing. Ask it to compare two things. Then ask it to add slight bias, heavy bias and outright lie. Write a review for a book, game, etc. Ask it to explain a joke or poem. Sometimes it gets the answer perfectly right. Don't worry about minor spelling errors, it will figure it out.

  And while all this is sensational, the best is yet to come: It understands and writes languages other than English. Ask a question in German or Danish, and you get a reply in those languages. The Danish replies are not br

• • Re: Maybe I'm missing the point?

    Ask a dumb question, get a dumb answer...

• This thing is pretty impressive, even if it is still mostly acting as a fancy search engine. What I would really like is for it to add links to sources, so that you can get a better idea about the validity of the information. Some questions I asked it:
  
  What is considered cheating in Call of Duty?

  Is it legal to share my Netflix password with friends?

  Which is more secure: Z-Wave or Zigbee?

  How can I prevent sniping when using a jointer?

  If the copyright on my work of fiction expires, do I still retain my mo