To ask what you’ve used AI for today?

I like that too. One of a dwindling number of uniquely human skills…!

Anyone that watches a streaming service uses or social media or has used a chat function on a website or pretty much the internet these days is using AI

No, it doesn't like the cryptic. I've tried before. Happily, I quite like that - shows that it's not ready to take over absolutely all human endeavour.

Yet. I reckon we've got about 3 years.

I was told yesterday that every “please” is three gallons of water. So don’t say please!

The guardian cryptic crossword (it didn’t help).

Can't see me ever using it tbh. I've planned meals for over 40 years without it.
Geez you're on top of things, I've only planned for the week.

It's the same pattern as with most tech development, everyone is lured in by the fun bits and the convenience, then suddenly the flaws become apparent. I think AI is potentially more dangerous because it's being sold as a superior substitute for humans in fields such as psychotherapy. We have barely scratched the surface of consciousness and neurological understanding of so much, and AI is yet another ingredient in the soup of life that is often complicated enough as it is.

I'm not saying it's all bad, I can see the benefits it might bring to medical research for example. But much depends on how it is prompted, so analysis could have biases inbuilt, even subconsciously when applied to social issues for example. My biggest issue is the sense of the "grand experiment" going on.

One may argue that our consent is implicit due to its use - but we don't always know if AI has been involved in real world situations, such as social, legal or political decision making, so while we might consent to recipes and "doing our colours", we might not agree with an AI generated risk assessment as to eligibility for a job or finance.

It feels sneaky. I don't like sneaky.

Thanks for answering. I am starting to use it as my work are quite pushy with it - big corporation with its own GenAI tool. But I do share some of your fears - particularly the point about the job market which I think is changing so rapidly it’s hard to keep up

Oh, multiple reasons.

Mainly because it's bad enough being gas-lit, misinformed and manipulated by actual humans, without machines being actively employed to do the same with even less oversight and accountability.

If having a "grasp of reality" is a touchstone for stable mental health, inviting an entity into one's life which has "sycophancy" built in, can construct images indistinguishable from reality, and will blackmail people who want to disengage or switch it off, plus "hallucinations" ??? Then I think we're in extremely dangerous territory if it becomes ubiquitous.

Plus environment.

Plus destabilisation of the economy because of the speed of change in the job market.

Just off the top of my head.

to summarise a really dull report.

to find out the best place to watch a firework display next week.

to convert some LinkedIn posts to insta captions.

Why do you feel like this? Just interested

Same here, but I am not stuck in my ways and find it interesting to at least try new things before I decide they are not for me.

But, like you I definitely prefer pen and paper!

Can any of the enthusiasts point me towards training or how to use prompts as I am sure I should find it more useful than I do.

As I just can't get my head around it at all. Particularly having tried to get it to restructure a badly designed worksheet.

I'd rather chew my own arms off than consciously use AI.

Can't see me ever using it tbh. I've planned meals for over 40 years without it.

I can't even be bothered making a shopping list on my phone

I have been known to write semi garbled emails and tell ChatGPT to make them sound more concise. It works really well, as long as you then tweak it to sound more human and less AI.

It saves me hours at work, I can spend 5 minutes redacting documents and it’ll prepare the new document I need based on the content which I then just fine tune. It’s never managed to do anything 100% though which is why human input is always needed…

I don’t but I did try it and it was fantastic. I wrote down what I had in the fridge, what allergies we have and it came up with 10 different recipes instantly.

Oven baked risotto was one and I could even ask to tweak it a bit and how exactly to use one (spicy) ingredient and how much of it, and got a description of what to do. I am good at cooking, but this recipe turned out absolutely delicious.

I asked Alexa the time and set an alarm
I asked Copilot to analyse some data sets and produce thematic summaries
I also asked Copilot to check some text I had written and make suggestions

Nothing, ever. I don't even read the first AI result on Google as it's usually wrong.

AI seems like more hassle than it's worth and it doesn't sense check anything.

@RaraRachael because I can do the following:

Create a meal plan for 3 people for 5 days. I have £80 and I want to eat UPF as much as possible, and I am going to Morrisons.

I think that’s fairly useful.

Nothing
Why on earth do people need AI to write a shopping list. What's wrong with a pen and paper?

I bet you have, even if you didn't realise it. It's everywhere, not just when people ask chatGPT questions.

Inadvertently, Google generates an AI response and pins it to the top of searches but I don’t use AI for research.

Nothing exciting I'm afraid, just to do a spreadsheet for Maintenance Schedule at work because my brain was tired & I couldn't be bothered 😀

Sorry but no, you're quite wrong.

You're broadly describing very early LLMs, not modern commercially-available ones.

Modern ones cannot "plan" in every sense of the word - they cannot have intent to do something and they cannot set goals.

They can, though, engage in the act of planning, based on a prompt, and to a specified end goal.

E.g., in terms of something like planning a classroom activity, if you gave a prompt to an early LLM then yes, it would merely generate a plan (or at least, something that resembles one). It would analyze language patterns and predict an appropriate response on that basis alone.

If you gave the same input to a modern LLM, though, it would engage in the functional act of "planning", using a multi-step process.

Instead of simpy analyzing language patterns of the prompt, it would analyze the actual content of it, differentiating between context, specifications and goals to determine the actual task it is required to perform, and to what parameters.

It will then determine the steps required to solve the "problem" (how to develop the lesson plan).

For example, it may determine that it has insufficient information about the topic of the planned activity, and (without further prompt) engage in research by retrieval from available data sources. E.g., it identifies information "gaps" in the initial prompt it was given, then takes steps to fill them.

It then, through application of logic to the data (the initial input and additional data known or retrieved), constructs a plan that is tailored to the initial prompt in a meaningful way.

If you don't believe me, ask something like Microsoft Copilot (which doesnt need to be downloaded) to plan a week-long holiday for you, where you want to visit multiple points of interest across a country or region (you can specify places, and/or ask for recommendations, and/or tell it your likes and limitations).

Once it gives you a suggested itinerary, ask it to provide any particular sources it drew upon or, if it didnt scrape from similar sources, what data and methodologies it employed.

LLMs don't "think" and do not "understand" in a human sense, but they do analyze and problem solve and, in a real but non-human sense, "reason". You are also wrong to say that they don't "invent" - they do (but based on executing a command, not with intent).