Attention: Please take a moment to consider our terms and conditions before posting.

Chat GPT

2

Comments

  • Ridiculous twaddle.
  • It is twaddle. It’s not AI. It picks up information that’s already out there and tries to make it relevant to the question. Whatever wordage that is put out will have been programmed somewhere and the source data is extracted from some database or other. 
  • JiMMy 85 said:
    A good rule of thumb is that you shouldn't accept the first response. You have to keep iterating responses until it gets to where you want it to be. 
    Aye.  Mucked about with it on emails to prospects as I am in sales.  There was nothing wrong with them as it got all the benefits of the product that I sell into the body of the email and it was slightly tailored to who I asked it to write the email for, but it read as robotic.  I know you can ask it to write emails in the context of a poet say, or as if I was a friend.  It will do some amazing things and continue to get better.  

    It will be interesting to compare it to other iterations of AI and GPT as they develop get released over time 
  • Interesting chat on Wake up to Money re AI. 
    One woman describes Otter AI where the transcripts of telephone conversations are provided including bullet points, time stamps and the forwarding to email / groups. 
    https://www.bbc.co.uk/sounds/play/m001kg86    (Conversation from 27 minutes to 41)

    The question about regulation was raised and who will be held accountable as AI will interact with a myriad of different apps.  I think it's gonna be an absolute minefield.

    Mention made of the Goldman Sachs report where they claim 300 million jobs could be displaced worldwide.
    https://www.bbc.co.uk/news/technology-65102150


  • Anyone using this much, day to day life? any tips / hacks, i have been watching a few videos today and listening to a podcast, so just seeing how it would benefit the average Joe.
  • AI is all fun and games at the beginning until...


  • Sponsored links:


  • Anyone using this much, day to day life? any tips / hacks, i have been watching a few videos today and listening to a podcast, so just seeing how it would benefit the average Joe.
    I use it when I can't think of a word. Just describe the word and it nails it every time. It's fucking great for tip of your tongue stuff. 

    I use it when I am writing jokes for work. If the line is "He's put in more shots than..." I will ask CGPT to give me a list of people associated with 'shots'. I'll get a list of people, and if I don't like it I will tell it to give me some more, or tell it to avoid war-related stuff or whatever. Eventually it will give me something that makes me laugh. The more obscure the better. It's just given me the name of the guy who invented the polio vaccine. 

    "The Dorking striker has put in more shots than Jonas Salk" - Thanks CGPT! 

    I used it to write a brief for my work too. It asked me some questions and then spat out everything I needed. 
  • I am currently in the middle of a programme of work to educate staff in a 1200 person business the benefits of using ChatGPT (before Microsoft copilot is launched at which time we will move to that exclusively). 

    It really depends on your role, but I have been educating others on 

    1. Training and development. 
    It is very good at creating training plans, quizzing you on a specific topic and offering direction. 

    2. Mundane tasks 
    Ask it to take a list of emails, separate into first name, surname and order alphabetically it will do a good job. 

    Ask it to write an excel formula for a specific task. 

    3. Plug-ins for specific tasks 

    with paid version you have a wealth of free plugins that can perform specific tasks. Some notable plugins 

    askyourpdf : I use this every day to load a pdf into ChatGPT and then ask questions on the content. Load in a set of company accounts and ask ChatGPT any questions on the data within (e.g. create a table with all references to PENSION). 

    Wolfram alpha : computational platform for complex maths questions. Process and then interact with results. 

    Plug-ins are being added daily. 
  • I use it most weeks to write pub quiz's for me, however at least 20% of the answers are completely wrong!!! Still 80% I don't need inspiration to write
  • A bit of a technical one for work.

    I asked it:

    "Write c++ code to determine if a user belongs to azure active directory group."

    It gave me an impressive looking answer, but it didn't work, as it assumed I had particular set up, so it would have been just as difficult to get it to work as it would be with the examples from Microsoft.
  • edited July 2023
    Last month I had to submit an RFP response for a large kiwifruit packhouse business in NZ, as they’re evaluating their current print fleet.
    I had a list of all their current printer brands and models and asked ChatGPT how many different toner models that all these printers used (as they want to consolidate the number of toner models).
    Would’ve taken me a very long time to work this out myself, as some of the models use the same toners and some don’t.

    I also used it to write the executive summary for the RFP response. Hugely timesaving and a superb tool (as long as you check all the info, as it does make mistakes).
  • Been playing with the 4o model over the last few days. Particularly impressed with the language translation. The latency of responses is usually very good and could be used when travelling abroad. 

    Anyone interested in more use cases with the new model, this tweet has video examples. 

  • Bacon ice cream and nugget overload sees misfiring McDonald's AI withdrawn

    https://www.bbc.co.uk/news/articles/c722gne7qngo
  • AI and voice recognition? That's not a recipe for disaster at all
  • Sponsored links:


  • aliwibble said:
    AI and voice recognition? That's not a recipe for disaster at all
     ChatGPT voice recognition on iPhone is excellent. You can use it for live translations from English to most languages with low latency.  

    I suspect it’s more IBM tech is shit rather than the potential.  drive thru orders is one of the first things that could be automated with ChatGPT.  

    However, I am not convinced it is that useful as ordering from a person regulates the queue and offers upsell opportunities. 
  • aliwibble said:
    AI and voice recognition? That's not a recipe for disaster at all
    It's here to stay and is only going to become more and more prevalent, with some amazing and absolutely awful consequences. There is little to no regulation, and even if there were, those who want to abuse it will simply do so from other countries.

    Interesting and worrying times ahead.
  • edited June 19


  • Natural selection in play for anyone who follows that.
  • A central London cinema has cancelled a private screening of a film which was entirely written using artificial intelligence (AI) following a public backlash.

    The Prince Charles Cinema in Soho was due to host the world premiere of The Last Screenwriter, which was created by ChatGPT, on Sunday.

    However, when concerns were raised by people about "the use of AI in place of a writer", the cinema announced that the screening had been axed.

    https://www.bbc.co.uk/news/articles/cjll3w15j0yo

  • edited June 19
    Good, easy to follow video on the state of AI. 


     AI has exploded very quickly compared to where it was two or three years ago. But relative to six or even 12 months ago, progress has been far more incremental. Have we already reached a point of diminishing returns? 


  • Good, easy to follow video on the state of AI. 


     AI has exploded very quickly compared to where it was two or three years ago. But relative to six or even 12 months ago, progress has been far more incremental. Have we already reached a point of diminishing returns? 



    Good video. I think the basic argument it makes is that as you get more specific concepts, say animal -> cat -> tabby cat -> tabby kitten -> fat tabby kitten, and so on, you have less and less training data and so your classifications become less reliable, unless you add a prohibitively large amount of training data.

    Certainly my experience is that the more specific you are the more likely you are to get a generic answer or a no comital response from Microsoft Copilot. My latest failure was: "plan a car journey from London to York with stops every 1 hour" (my back and legs start aching after that time), and Copilot's answer was effectively to use Google Maps or the AA route planner.

  • Good, easy to follow video on the state of AI. 


     AI has exploded very quickly compared to where it was two or three years ago. But relative to six or even 12 months ago, progress has been far more incremental. Have we already reached a point of diminishing returns? 



    Good video. I think the basic argument it makes is that as you get more specific concepts, say animal -> cat -> tabby cat -> tabby kitten -> fat tabby kitten, and so on, you have less and less training data and so your classifications become less reliable, unless you add a prohibitively large amount of training data.

    Certainly my experience is that the more specific you are the more likely you are to get a generic answer or a no comital response from Microsoft Copilot. My latest failure was: "plan a car journey from London to York with stops every 1 hour" (my back and legs start aching after that time), and Copilot's answer was effectively to use Google Maps or the AA route planner.

    Microsoft copilot is many multiples behind ChatGPT and its 4o model (despite Microsoft having access to use it). MS are concerned about the performance (and cost) of unleashing OpenAI’s models in copilot and therefore seriously limit the features available. 

     ChatGPT has given me a decent answer to a questions regarding travel and potentials stops (EV charging in my case). 

  • BalladMan said:
    Good, easy to follow video on the state of AI. 


     AI has exploded very quickly compared to where it was two or three years ago. But relative to six or even 12 months ago, progress has been far more incremental. Have we already reached a point of diminishing returns? 



    Good video. I think the basic argument it makes is that as you get more specific concepts, say animal -> cat -> tabby cat -> tabby kitten -> fat tabby kitten, and so on, you have less and less training data and so your classifications become less reliable, unless you add a prohibitively large amount of training data.

    Certainly my experience is that the more specific you are the more likely you are to get a generic answer or a no comital response from Microsoft Copilot. My latest failure was: "plan a car journey from London to York with stops every 1 hour" (my back and legs start aching after that time), and Copilot's answer was effectively to use Google Maps or the AA route planner.

    Microsoft copilot is many multiples behind ChatGPT and its 4o model (despite Microsoft having access to use it). MS are concerned about the performance (and cost) of unleashing OpenAI’s models in copilot and therefore seriously limit the features available. 

     ChatGPT has given me a decent answer to a questions regarding travel and potentials stops (EV charging in my case). 

    Interesting, yes, I get a reasonable answer from 4o for "Plan a driving route from London to York with stops every hour" and "Value after 3 years of saving £200 a month when the interest rate is 5% a year", but for "All Nobel prize winners born in the London borough of Greenwich" I get two names but neither was born in Greenwich according to Wikipedia.

    I'll keep trying to use ChatGPT but I'll still need to cross check some of the answers against Wikipedia, stackoverflow, reputable news organizations, etc.
  • Questions for those of you who are comfortable with using the freely available platforms.

    I have to write a fairly formal report on a serious topic. My style has always been a bit racy. What would I say to ChatGPT to make my text more suitable for a conservative business audience? And what is the difference between standard ChatGPT and GPT4, and is the latter a subscription platform, and if so, is it worth it for the occasional user? Are there now better platforms for “improving” authored texts? I have been using DeepL to edit blogposts or give them a snappy headline but it tends to make them a bit bland, maybe thats inevitable?
Sign In or Register to comment.

Roland Out Forever!