Chat GPT & Generative AI tools

[Kai]

I wish we could hit pause and enjoy it slower, stage by stage. Even though it still makes errors it's like passing your driving license and suddenly you're driving an F1 car. Who needs an electrician, a conveyancer or a coder?

Same here, but time seems to speed up as we get older, so seems a no-brainer to use any tool that can greatly accelerate our learning and help us keep up with the mad pace of the world today.

Gone are the days of posting relatively basic queries on forums and waiting in hope that someone eventually provides an adequate reply.

But the vid I posted earlier about the "attempted murder" highlights exactly which issues we need to solve before we give it more power.

[Euler]

I worry that by not embracing AI now, I risk getting left behind in tomorrow's technology. Like the way my dad could never set the timer on the video recorder.

How do you guys use AI? Voice, text? Phone, PC, other device? Which AI system do you use? What do you actually use it for?

I had solar installed recently, and the inverter came with an API. So I can control the inveter and tell it what to do and when.

I used AI to write code to look up the weather forecast by hour and change the behaviour of the inverter to maximise my ability to harvest solar, or store power at optimal moments.

Today I had the smart meter upgraded. As soon as the guy left, it was clear something was wrong, either that or the sun had gone super nova.

Took a picture of the wiring, of the data from the inverter and asked AI what was wrong.

A few seconds later, it had identified the problem, and five minutes later, the sun calmed down to a normal state. Or I fixed it, one of the two.

[weemac]

CLOCKS BACK
In the UK, clocks go back one hour at 2 a.m. on Sunday, 27 October 2025. This marks the end of British Summer Time and the return to Greenwich Mean Time (GMT).
You said:
27TH?
Yes. In 2025, the clocks go back at 2 a.m. on Sunday, 27 October.
You said:
26TH
No. In 2025, the last Sunday in October is the 26th, not the 27th.
So clocks go back at 2 a.m. on Sunday, 26 October 2025.
You said:
YOU JUST LIED
Correction: you are right.
I misspoke earlier — the clocks go back at 2 a.m. on Sunday, 26 October 2025

[firlandsfarm]

CLOCKS BACK
In the UK, clocks go back one hour at 2 a.m. on Sunday, 27 October 2025. This marks the end of British Summer Time and the return to Greenwich Mean Time (GMT).
You said:
27TH?
Yes. In 2025, the clocks go back at 2 a.m. on Sunday, 27 October.
You said:
26TH
No. In 2025, the last Sunday in October is the 26th, not the 27th.
So clocks go back at 2 a.m. on Sunday, 26 October 2025.
You said:
YOU JUST LIED
Correction: you are right.
I misspoke earlier — the clocks go back at 2 a.m. on Sunday, 26 October 2025

[Euler]

I've been using Atlas this week, it's really good. It feels like the start of a reinvention of the internet.

I see Wikipedia traffic is falling away as people use AI instead. Ironically, a free encyclopaedia has hastened its downfall.

[firlandsfarm]

I've been using Atlas this week, it's really good. It feels like the start of a reinvention of the internet.

How to decide which AI to use? It's a bit like the land grabs of the wild west!

One may be good at some things while another is good at others. Is there a website that tries to denote which AI is good for what?

[weemac]

One may be good at some things while another is good at others. Is there a website that tries to denote which AI is good for what?

Whatever you do, don't ask GPT5.

[Euler]

I've been using Atlas this week, it's really good. It feels like the start of a reinvention of the internet.

How to decide which AI to use? It's a bit like the land grabs of the wild west!

One may be good at some things while another is good at others. Is there a website that tries to denote which AI is good for what?

https://llm-stats.com/benchmarks/llm-leaderboard-full

I think the gains are coming weekly. So whatever model you choose, it will be beaten next week, but will be upgraded itself so will be ahead of the one that beat it last week etc.

I've found more payoff by focusing on a couple and learning how to get the best out of them.

[firlandsfarm]

One may be good at some things while another is good at others. Is there a website that tries to denote which AI is good for what?

Whatever you do, don't ask GPT5.

I've not had any serious problems with GPT5(thinking). My biggest frustration with it is if you question it's response it will react and explain/correct but also repeat all the stuff you didn't challenge which is boring by repetition!

[weemac]

One may be good at some things while another is good at others. Is there a website that tries to denote which AI is good for what?

Whatever you do, don't ask GPT5.

I've not had any serious problems with GPT5(thinking). My biggest frustration with it is if you question it's response it will react and explain/correct but also repeat all the stuff you didn't challenge which is boring by repetition!

I'm definitely into double figures of the times I've accused it of error/misinformation, and it invariably backs down. It's especially bad with anything requiring recent or current knowledge e.g. the clocks back example.

[firlandsfarm]

I'm definitely into double figures of the times I've accused it of error/misinformation, and it invariably backs down. It's especially bad with anything requiring recent or current knowledge e.g. the clocks back example.

My us is basically about coding.

[ShaunWhite]

and it invariably backs down.

It's heavily biased towards agreement, as 'we' discussed the other day, subscriptions are more likely to come from it making people feel smart rather than it being smart. And as it's an LLM with zero 'intelligence' it's unable to distinguish between a preponderance of bad evidence vs a lesser amount of good evidence. So it will bow to your judgement rather than insisting you're wrong.

And that's a big problem with something like betting, internet content is predominantly wrong, naive or plain bs, so it's biased towards giving incorrect information rather than delivering any sort of useful truth. It's just reciting what losers talk about rather than validation or confirming what the fewer number of knowledgeable articles are saying.

On the plus side, if you approach it from first fundamentals and the data science, then those are topics that are irrefutable and well documented and not polluted by 'sports' subjectivity.


While I'm here, this was a good watch... https://youtu.be/A5w-dEgIU1M?si=Ti5r6QWYkJE18FVy

Screenshot_20251026_181638_YouTube.jpg

[wearthefoxhat]

and it invariably backs down.

And that's a big problem with something like betting, internet content is predominantly wrong, naive or plain bs, so it's biased towards giving incorrect information rather than delivering any sort of useful truth. It's just reciting what losers talk about rather than validation or confirming what the fewer number of knowledgeable articles are saying.

I've seen a quite a few trying to use GPT to predict the outcome of a horse race that way, asking it to evaluate form, and produce an outcome. Good luck with that, if used for that purpose, then should used only for fun and not taken too seriously.

The effort should be put into asking GPT to "assist" with either formulas or coding, that can, in turn, be used to evaluate/interpret data sets. The rest is up to the user. Any user of A.I is ultimately responsible, but we tend to blame others instead of admitting to messing it up ourselves.

I'd say the same applies to any sport that has a wealth of data, ie: Football, Cricket, Tennis...etc.. (more liquidity too).

[Big Bad Barney]

I'm definitely into double figures of the times I've accused it of error/misinformation, and it invariably backs down. It's especially bad with anything requiring recent or current knowledge e.g. the clocks back example.

My us is basically about coding.

IMO Claude Code (anthropic) is the go for coding atm. It's good because it has access to your filesystem. (Some people might care about privacy...I don't...)

You obviously wanna know what you are doing or it'll make a mess...but if it fks up, ya just revert and try again... It's exceptionally good at writing things from scratch, but gets less good as your app gets more complex or ambiguous. One ends up spending their time documenting coding standards, rather than actually coding....then it gets a bit boring repeating, 'follow this directive you...suck...get it right... ok thx'

But still 50% faster than doing it yourself....

ChatGPT is cheaper because one quickly ends up on the max plan with anthropic.

Edit: The real advantage of these tools is learning... learning to code is never just the syntax, it's the patterns and practices and architectures that REALLY take ages to learn.......IMO one would do well to short places like Pluralsight for learning those things.... that used to be the goto... LLM's do it better.

[firlandsfarm]

and it invariably backs down.

Hey Shaun, bit of a missed punctuation/quote there ... I didn't say that, it was weemac!