A.I. Check

cfw1994

Scarum said:

AI engines are a great tool for learning how something works, finding a way to do something that is not overly complex, or solving a problem - but it always serves to question anything it comes up with as it will make mistakes; The day you don't question and rationalise anything you hear or read, you've lost the plot.

Absolutely.
BUT….a useful tool.
A bit off topic (but on AI): we did two months on Interrail in 2023.  Lovely way to travel 🙏

We walked a lot.  Over 800,000 steps - we like wandering a city, popping into the odd museum, gallery, cafe, etc.

I was using ChatGPT between some places to come up with a walking tour for the place we were heading. 

It worked brilliantly: like having our own personal mobile travel agent 😎

I used it to create a volleyball club constitution.  Of course I tweaked it a little, but we needed one for some new Volleyball England registration (despite the club running for 40 years 🤣), & within an hour on a Sunday morning, it was done. Ran it past the committee and everyone as very happy with it 💪

A local council proposed to splatter double yellow lines through an area in a small village we know.  Totally unnecessary.  Used ChatGPT to sketch up a solid defence against it.  Again, tweaked it a bit, sent, shared with others to tweak and send: Council backed down.

Nebulous2

cfw1994 said:

Scarum said:

AI engines are a great tool for learning how something works, finding a way to do something that is not overly complex, or solving a problem - but it always serves to question anything it comes up with as it will make mistakes; The day you don't question and rationalise anything you hear or read, you've lost the plot.

Absolutely.
BUT….a useful tool.
A bit off topic (but on AI): we did two months on Interrail in 2023.  Lovely way to travel 🙏

We walked a lot.  Over 800,000 steps - we like wandering a city, popping into the odd museum, gallery, cafe, etc.

I was using ChatGPT between some places to come up with a walking tour for the place we were heading. 

It worked brilliantly: like having our own personal mobile travel agent 😎

I used it to create a volleyball club constitution.  Of course I tweaked it a little, but we needed one for some new Volleyball England registration (despite the club running for 40 years 🤣), & within an hour on a Sunday morning, it was done. Ran it past the committee and everyone as very happy with it 💪

A local council proposed to splatter double yellow lines through an area in a small village we know.  Totally unnecessary.  Used ChatGPT to sketch up a solid defence against it.  Again, tweaked it a bit, sent, shared with others to tweak and send: Council backed down.

I use copilot, which I understand is based on chatGPT. I was quite suspicious initially as I'd read stories about hallucinations etc. 

I had 10 days to travel from the Mediterranean in the South of France to the channel. I'm not getting any younger and am trying to limit the big drives that I've previously done. Regard the journey as part of the holiday, rather than an endurance test. 

I gave copilot detailed instructions. Find a route from where I am to Calais, maximum of 450 km to drive in a day. Stay at least 2 nights, preferably 3 on each campsite. Interests are Unesco world heritage sites, woodland walks, cycle routes. 

It made a cracking route, found some very interesting places to see / cycle and gave me a choice of campsites with phone numbers and web links to book the site. It was also readily adaptable. "I don't need to be so close to Calais on the last stop, as our crossing isn't until mid-afternoon." Then it suggested a site further away for our final stay. 

The level of detail would have taken me hours of research and adjustment, and I'd probably have compromised on some of it. 

Rose1509

Yeah, I’ve definitely plugged my financial info into AI tools to see what insights I could get. They’re super handy for organizing data and double-checking things, but I always cross-check what they say, since they sometimes mess up on basic stuff or make weird assumptions. For me, it’s just another way to look at my finances from a different perspective, not something I fully rely on.

nicknameless

Lots of scepticism here.  Might be n=1 and a fluke but I produced a retirement model using voyant a few weeks ago and just out of interest chucked the same data into chat GPT (in much quicker order).  Was just curious as to what it would throw out in comparison and had used it to date like other posters e.g. walking and travel itineraries and the like needed in quick order.  Anyway answers from the AI (yearly spend, probability of success etc.) were extremely similar.  Was mildly reassured.  Perhaps I should be questioning voyant and chatGPT?

NormalNorman

Recently used AI for many financial questions and so far been happy with the answers which I already mostly knew. Just looking for some extra titbits really. Also used to it to go through my budget and categorise things better by displaying the data in more helpful ways. Being proactive spurs you on to get jobs done. For example, even on PAYE got three claims in with the revenue now. Cheers!

GenX0212

I will give everyone a stark example of why you need to be VERY CAREFUL 

I put some of my figures into CoPilot (DB, SP, Gilt Ladder values and a starting pot of £650k) and told it I wanted a post tax income of £50k.
I asked it to model growth rates of 0%, 3%, 5% and 10%
I then asked it to factor a market crash of 30% into the 1st year and produce a side by side comparison table:

Without checking the actual number values themselves someone tell me what is fundamentally wrong with the table above???

SarahB16

GenX0212 said:

I will give everyone a stark example of why you need to be VERY CAREFUL 

I put some of my figures into CoPilot (DB, SP, Gilt Ladder values and a starting pot of £650k) and told it I wanted a post tax income of £50k.
I asked it to model growth rates of 0%, 3%, 5% and 10%
I then asked it to factor a market crash of 30% into the 1st year and produce a side by side comparison table:

Without checking the actual number values themselves someone tell me what is fundamentally wrong with the table above???

I haven't got the time (or the inclination) to check where the error is however whenever I run my forecasts I have them run so that every year is shown so any errors can easily be spotted.  I recognise what you are displaying as Co-pilot and I use it a lot as when I use ChatGPT I hit my daily usage limit quite quickly when doing detailed calculations.  

I therefore do my background work in Co-pilot and then pose the question in ChatGPT where I wish to have the outcome in Excel and also ask for it to be produced in Excel (and for every year) then I can scrutinise the forecasts. 

If I were you you could have asked the same question in ChatGPT and for every year but with the outcome in Excel then it would be very easy to spot the error.    

michaels

GenX0212 said:

I will give everyone a stark example of why you need to be VERY CAREFUL 

I put some of my figures into CoPilot (DB, SP, Gilt Ladder values and a starting pot of £650k) and told it I wanted a post tax income of £50k.
I asked it to model growth rates of 0%, 3%, 5% and 10%
I then asked it to factor a market crash of 30% into the 1st year and produce a side by side comparison table:

Without checking the actual number values themselves someone tell me what is fundamentally wrong with the table above???

Dunno what is wrong but I sure am hoping for a 30% crash in the first year of my retirement based on those numbers....

GenX0212

OK, so here it is. Look at the numbers in the 'Crash at 57' columns vs the 'No crash' columns.
Obviously the pot values should be consistently lower if the starting point is 30% lower but what CoPilot has actually given:

I didn't try to deliberately confuse the AI in any way. I gave it the one scenario of wanting to draw £50k a year and then asked it to adjust for a market crash and compare the results side-by-side.

The numbers it has calculated in the 'Crash' column are clearly nonsense, but it hasn't just got an assumption wrong somewhere and carried it forwards in the projection or else the results in the'5% age 75' and '10% age 90' columns wouldn't be lower values when the rest are higher. Not only has it got the calculation wrong it has produced a set of results which aren't even consistent.

I wouldn't really trust any of the numbers.

RecliningInPeace

In order to reproduce your findings, I put the following prompts in one at a time and CoPilot produced the table shown below. I've not verified the numbers (at a glance they look more sensible) as all I wanted to do here was provide an opinion that AI models can only answer the questions asked of them, making how those questions are asked quite important.

Q1: I am 57 years old and am about to retire. I have a pension pot of £650k. I want to draw a post tax income of £50k per annum. Create a table showing age in column 1, from 57 to 100, and the estimated size of the pension pot in column 2.
[At this point CoPilot explained that it had assumed 0% growth in its output.]
Q2: Add columns for growth rates of 3%, 5% and 10%
Q3: For each of the growth rate columns, add a column where in the first year there is a market downturn resulting in the pension pot dropping by 30%
[CoPilot produced a partial table for ages 57 to 70 explaining that it did so for clarity.]
Q4: Do the same table, but show only for ages 57, 60, 67, 75, 80, 86, 90 and 100.