Chat GPT is amazing

[Tickets25]

I know AI can be scary but I've struggled with something for over 30 years and last night chat gpt framed it for me in a way that years of therapy never has and it's given me peace.
It's fabulous.

No point to this post, just wanted to say!

My chat GOT had turned Welsh. It only responds to me in Welsh these days even when I ask it not to.
Ive uninstall ed znd reinstalled it but whether I speak to type, it responds in Welsh.
I now use co pilot.

I have tried to use it for lots of things - most unsuccessfully because it is factually incorrect so often.

I use it almost exclusively to draft letters and emails now. My brain really can’t get over starting them so I create a rough template with some basic info them tweak and send. It also helped me ID a pen I wanted to replace but couldn’t find the make or model of.

Let's not forget the entire setup is rooted in stolen intellectual property and is run by some of the most unpleasant people on earth. Elon Musk keeps tweaking the parameters of Grok to reflect his own views so that it made claims of white genocide in South Africa for instance. Plus it reflects and reproduces all the unpleasant social biases of the internet and is sexist and racist. It will for example tell women to negotiate lower salaries than men.

Oh and I'm using it to teach me farsi, which is not available on Duolingo. It's fantastic.

My employer has been encouraging us to use copilot for a specific task to 'save time' and yet it's so unreliable that we have to double check the output, every line of it, because any gaps have a wider impact. It takes more time to use it this way. It's actually not even always that helpful for functional tasks.

The woman who used it to look up statistics, was given some, asked for a link or source, it said it didn't have any, she asked where the stats came from, and it said it has just answered in a way an expert might, is mental - you can't use it as a search engine and people are. That's where the not thinking originates. It was bad enough when people cited Wikipedia as sources without checking where those sources originated. Chat GPT is like Wikipedia.

It’s like any assistant. You need to give it the right instructions and sometimes it messes up. Unlike an assistant though, you can tell it it’s wrong and it won’t get upset or take you HR. Also like any new employee you need to teach it your house style and preferred use of language. And just like your human assistants, you are responsible and accountable for the work it produces and its mistakes.

It’s perhaps worth saying that none of the errors I have had from ChatGPT even come close to the howlers made by humans.

That said, I do get annoyed at AI CVs and poorly worded AI emails. It’s easy to edit an AI draft to give the right tone and impression, so I know you are a lazy fucker if you haven’t bothered.

I guess those people who say 'it's amazing for work' are doing very different jobs to people who say 'it's absolute bullshit that lies'.

It's great at drafting documents per clear instructions, or polishing emails.

But as a maths teacher it is absolutely shit at producing resources.

To demonstrate, I just asked it to produce a worksheet on finding the surface area of a hemisphere with a worked example.

A few more prompts later about needing more realistic diagrams which are labelled it has lost the worked example reasoning and in fact actually lost the questions.

At this point you give up and do it yourself.

Horrible story on Kuensberg this morning about a young teenager from Orlando who took his own life, encouraged by an AI chatbot over many months.

I don't know how I forgot about this but a couple of months ago a guy told us that he'd been profiling his friends and family using it. Of course he said it was telling him what he already suspected to be true about those people but he was absolutely insistent that he was giving it objective information and that the whole thing was free of bias. It wasn't nice to read and I wonder now how it has affected his relationships with the people he was 'profiling'.

I haven’t used Facebook but doesn’t this kind of thing happens there also? I have been told of people silently stalking people on Facebook.

I used to work for a company that made report software. Yes, it was 'impersonal' in that teachers didn't write a unique sentence for every pupil, but they did select all the statements that applied to each child in a "does well at/needs to work harder at/enjoys" type matrix.

I don't know why parents get so wanky about this, the personal bit is understanding what your child is good at, which isn't that different across a class of 30 all doing the same topics FFS!

I agree.

I used it to chat through some upset when I was scammed by a cab driver on holiday a few months ago. It was really bugging me, but Chat really helped me think it through and let it go.
I've also used it to think through a few niggly issues with some friends, and I thought it was really good.

Like with all technology, if you use it sensibly, with an open mind, it can do great things. I love it.

Just saw this 😂

A fitness chat group I belong to has many chat gpt devotees. They love it for making meal and fitness plans. One woman was told by the program to eat 750g of Greek yoghurt for breakfast. She shared this as a funny, ridiculous aside... as if the ridiculous thing weren't trusting a non-sentient and error-prone primitive pattern-matching computer program with your health and nutrition.

I think this might be another untested area for AI, bespoke software.

If an error is spotted in "traditional" software, there will be a document trail of design specs, decisions, plans, who done what etc. So should be ok to trace and fix ( erm, Post Office excluded).

If a school uses AI to create it's own system, none of this will exist. Liability issues might arise.

Schools are buying the report software you worked on, because it is documented and tested.

It was a fantastic help with travel arrangements on a recent holidag , took a lot of stress away

There have been loads of instances of it messing up people's holidays by telling them to go to places that don't exist or the wrong airport etc

I never had any problems .

Imagine just blindly trusting it without fact checking!

I actually really enjoy researching holidays myself. Feels so lazy and unimaginative asking Chat GPT

Yeah, I really don't get this thing with using AI to plan trips.

Years ago, in the days of landline telephones, I would spend weeks pouring over maps, youth hostel book in hand, trying to work out how to get to a particular thing I wanted to see, or a road I wanted to ride on my bike.

As I got older, I moved onto Lonely Planet guides, travel agents and the phone to book stuff, on the landline.

I still basically do the same, but now use the net.

Because if I want go somewhere, it's because there is something I want to do there. And realistically, there are not a million options on how to do it.

If I want to go to the Empire state building, no need for AI. Just go online, book flights and an hotel.

Then spend hours reading up on what else to do in NY.

The research is half the fun.

I do find it useful for some things e.g. "please walk me through how to get my spicy cat into a cat carrier to take it to the vets in simple steps".

What it was not useful for was researching a new oven. I wanted a double oven that also has a steam function, within a certain budget. Instead of just telling me "no mate that doesn't exist", we went round and round the houses with it specifying all these different models but when I actually looked those ones up on line they either were no longer in manufacture, didn't meet my criteria (e.g. they were a single oven only, or they didn't have the steam function etc etc). It was so unhelpful and actually just made the whole process more confusing. I wish it would have just straight up told me it couldn't help!

If I'm having a break, a few days away from home, I often ask one of the AI engines if there are any interesting buildings, monuments or graves - anything with a historical story. Several times I've been told something was in the area I would be staying but on a few occasions I've known that, for instance, a writers grave (which I have already visited), the AI engine was giving me completely false information.

On another occasion, I was searching for a poem - I knew the writer as I had about 7 of his books but I'd flicked through and couldn't find it but I knew exactly what the poem was about - the AI engine responded by making up a poem in the style of the writer and even claimed it was in one of the books I owned - if course I didn't recognise the poem or the made up title and checked in the actual book to confirm that what the AI engine had given me was wrong - when confronted - it immediately confessed to the fabrication - something I never asked for. It's a phenomena well known as "AI hallucination".

Musk's Grok now straight up saying Trump won the 2020 election. GenAI is a massive problem for disinformation.

I didn’t ask it to plan trips . It was used for things like which platform do I need for Kanazawa . Japanese train stations are like cities in size . It was never wrong .