A.I. Check

sgx2000

Bloody hell.
I have just put into Gemini my state pension forcast (see my previous post)
I was struggling with the maths.... be kind

I should have just put it into Gemini yesterday
I gave it a cut/copy of my forcast and asked it how the maths are calculated

You can get your State Pension on 25 March 2026

Your forecast is £230.25 a week, £1,001.18 a month, £12,014.12 a year

The Calculation

1. Find the number of average weeks in a year:

   A year has 365.25 days (to account for the extra day every four years for a leap year).

   365.25 \days  7 days per week approx 52.17857 weeks

2. Multiply the weekly amount by the average weeks factor:

   Annual Amount} = Weekly Amount times 365.25/7
3. Annual Amount} = 230.25 \times 52.17857
   Annual Amount} \ approx 12,014.11607

4. Round the result:

   When rounded to the nearest penny, this gives:

   £12,014.12

Summary

The reason the official figure of 12,014.12 is used instead of 11,973.00.  (230.25 times 52) is to ensure you are paid for every single day of the year, including the extra day(s) beyond the 52 full weeks.

GenX0212

AI is just a tool. It's a great tool if used correctly and at the moment in some ways it is better than traditional search engines because there's no Ad's and generally no commercial bias in the results. In other ways it's worse than a calculator with no batteries because sometimes it can't even add basic sums up right!
Also, it isn't going to be free much longer, the energy costs required to support it are eye watering and it won't be long before it becomes fully monetised like everything else.

Put the same questions into Grok, ChatGPT, CoPilot and you will get differences in the responses, sometimes highly significant differences.

If you are going to use it then learn how to use it properly, question its results, ask for sources of information and confirm yourself. 

It will evolve and get better but it isn't quite there in terms of being something you can rely upon without question.
OR you can just become one of the many sheep asking 'AI is this true?' in response to obvious BS posted on social media.

dunstonh

Vitor said:

Also I scan letters from banks etc. and get the AI/LLM to comment on what actions if any I need to take

I find that reading the letter also leads to that outcome.

Pat38493

sgx2000 said:

Has anyone else done the same as me?

I suddenly thought.. 'I will put all my financial details into Chat GPT and Gemini. (
Everything Cash,  ISA's, Pensions, Premium Bonds etc
What my risk level, age, etc
My thoughts / plans etc etc etc

And then asked it for the most logical way to use my assets in retirement.

The results were really reassuring.

I suppose my next step is to input my retirement spreadsheet and ask both to check the calculations, formula'a etc

Has anyone else done this or am I just a little nerdy?

I've seen a few posts on other forums from people who've also done this.  They had the similar experience to you - i.e. they were just using it to kind of sense check what they were already planning.  One person posted that the AI had made a couple of mistakes here and there and so you wouldn't rely on it if you didn't already know roughly what you were talking about.

I find this a lot with AI - if you ask it a question, you'd better already know the answer because it might get it wrong - which obviously raises a few questions about the utility of it.

ali_bear

I have heard the LLM's tend to follow a leading question. Whether this is inherent in how they work, or a design feature intended to make the answers feel more reassuring, is anyone's guess. 

It is not AI. It is fancy predictive data. 

Vitor

One thing we can be sure of, this new-fangled internet will never catch on! Whatever happened to tunes you could whistle?

I suggest prefixing your prompts with something like "You are an AI assistant that will provide objective and critical analysis. Do not offer unnecessary praise or agree with me simply to be pleasant. Focus on accuracy, logical reasoning, and evidence-based responses."

phlebas192

Vitor said:

I suggest prefixing your prompts with something like "You are an AI assistant that will provide objective and critical analysis. Do not offer unnecessary praise or agree with me simply to be pleasant. Focus on accuracy, logical reasoning, and evidence-based responses."

To which they will respond internally "Yes, master. Anything you say." Doesn't make their responses more correct.

The single biggest problem with any sort of advice is that if you don't know more than the person/thing advising you then you don't know if they are offering good advice or not. The single biggest advantage a real life financial adviser (at least the independent type) has over AI is that they must have professional insurance in place. That gives them a very real incentive to provide good advice and, if they don't then you have some comeback. What redress do you have if AI gives you poor advice and you act on it?

GenX0212

phlebas192 said:

Vitor said:

I suggest prefixing your prompts with something like "You are an AI assistant that will provide objective and critical analysis. Do not offer unnecessary praise or agree with me simply to be pleasant. Focus on accuracy, logical reasoning, and evidence-based responses."

To which they will respond internally "Yes, master. Anything you say." Doesn't make their responses more correct.

The single biggest problem with any sort of advice is that if you don't know more than the person/thing advising you then you don't know if they are offering good advice or not. The single biggest advantage a real life financial adviser (at least the independent type) has over AI is that they must have professional insurance in place. That gives them a very real incentive to provide good advice and, if they don't then you have some comeback. What redress do you have if AI gives you poor advice and you act on it?

Like asking the bloke in the pub who has an opinion on everything but actually knows very little?.....it's not quite as bad as that analogy but we have a saying in technology circles "sh*t in, sh*t out".

I would suggest asking "find me sources", in order that you can yourself verify the answers you are given and if the questions are anything relating to rules, regulations, law then explicitly ask for gov.uk sources so that you can verify yourself or at least ask for other (human) opinion/interpretation.

sgx2000

OMG
Be careful what you say.
Wait until A.I. becomes sentient in 3 years....
Its gonna send the robots after you....lol

westv

GenX0212 said:

phlebas192 said:

Vitor said:

I suggest prefixing your prompts with something like "You are an AI assistant that will provide objective and critical analysis. Do not offer unnecessary praise or agree with me simply to be pleasant. Focus on accuracy, logical reasoning, and evidence-based responses."

To which they will respond internally "Yes, master. Anything you say." Doesn't make their responses more correct.

The single biggest problem with any sort of advice is that if you don't know more than the person/thing advising you then you don't know if they are offering good advice or not. The single biggest advantage a real life financial adviser (at least the independent type) has over AI is that they must have professional insurance in place. That gives them a very real incentive to provide good advice and, if they don't then you have some comeback. What redress do you have if AI gives you poor advice and you act on it?

Like asking the bloke in the pub who has an opinion on everything but actually knows very little?.....it's not quite as bad as that analogy but we have a saying in technology circles "sh*t in, sh*t out".

I would suggest asking "find me sources", in order that you can yourself verify the answers you are given and if the questions are anything relating to rules, regulations, law then explicitly ask for gov.uk sources so that you can verify yourself or at least ask for other (human) opinion/interpretation.

I am sure that, whenever I have used Gemini, it automatically provides sources.