After the budget announcement about AI and the new Manchester prize yesterday, I thought I’d have another go at Chatgpt.
I asked it this, and was pretty impressed with the result.
“I want to compare running costs between an ev and an ice car over a year. The ev does 3.5 miles per kWh. The ice car does 55mpg. It costs .35 per kw for electricity, and it costs £1.50 a litre for petrol. I do 14000 miles a year. Which would be cheaper, and by how much?”
I then added that I have a phev that does about 30 on electricIty, the rest on petrol, and how does that affect the result, and again, it worked it all out in a few seconds.
Anyway, beats fannying about with spreadsheets when comparing stuff like this.
|
I tried CHATGPT with some financial related questions, for example, the future value of an investment given fixed returns, compound interest over a period etc.
CHATGPT got the formula correct and gave me a figure.
The figure didn't look right and I double checked it manually and against our system at work.
The formula was correct BUT it got the actual maths quite wrong!
I replied that it was wrong and it admitted it and re-worked it out, but was still wrong! I iterated the above a few time and it got nearer but it was never correct.
|
Annoyingly ChatGPT has stopped running for me I get "Something went wrong. If this issue persists please contact us through our help center at help.openai.com.". I've cleared cookies and tried a different browser. It's coming up with it more or less right away i.e. not going off to the internet first. I have contacted them but nothing. Will have to research it more when I have time.
|
Try chat.openai.com/chat
I retried my question today and it got it perfectly correct to two decimal places.
|
>> Annoyingly ChatGPT has stopped running for me I get "Something went wrong. If this issue
>> persists please contact us through our help center at help.openai.com.". I've cleared cookies and tried
>> a different browser. It's coming up with it more or less right away i.e. not
>> going off to the internet first.
I got that once - I closed the window, went to the launch page again, and it asked me to log in again from scratch. Then it was ok.
Launch page as in:
openai.com/blog/chatgpt
|
We use rules based underwriting for the small simple stuff and much of it is automatic.
The more complex stuff is what I'm paid to do. I guess it would be possible to feed all of the data from the cases I reviewed over the years, which do change based on economic climate (deals I did a year ago, I wouldn't do today) and then ask the AI to make the credit decisions.
For performance testing, you could give it a portfolio and see how it performs compared to humans.
|
If you were to believe all the rather hot headed articles about AI that’s almost certainly happening now, and in five years time it’ll be running the world.
Today I asked it about national insurance and some queries about my state pension. It knew the wrinkles, knew about COPE, pointed me at the appropriate government website, and when that failed with an error filling in the form, I told it the error and it told me what to do about that and gave me a phone number.
I asked it how to do some reasonably fiddly stuff with my smart devices, and it walked over the problem, told me exactly how to solve it and suggested a better way than I’d thought.
I asked it how to create a web page to do some low level file renaming on my windows box, with checkboxes and a submit button. It produced the html. It then said it would need some php, so it produced that. And when I said I didn’t have a web server installed, it walked me through what to do, where to download it, exactly what to type and what obscure configs I’d need to change to get it all to hang together nicely.
Trouble is, when it looks slick, and believable, it could tell you anything and you might swallow it.
Last edited by: Crankcase on Thu 16 Mar 23 at 14:36
|
>>
>> Trouble is, when it looks slick, and believable, it could tell you anything and you
>> might swallow it.
>>
...if it gets a Degree in PPE, we're doomed.... ;-)
|
When to wear an orange hi vis v. a yellow one, or what steel toe capped boots can do for your sex appeal?
Presumably the old fashioned PPE... :-)
|
|
Currently, in my experience its little more than a glorified search engine. Not that good either, I asked it one thing and it claimed "to be outside my 2021 knowledge profile"
|
>> Currently, in my experience its little more than a glorified search engine. Not that good
>> either, I asked it one thing and it claimed "to be outside my 2021 knowledge
>> profile"
It's pretty glorified, yes.
I just did a mindless thing that would take me ages to plod through manually.
"write me a windows powershell script to compare two text files. if there is the word carrot in one of them, change it to lettuce, unless the word tomato is already there."
It did that in about ten seconds. I then said "great, change that to c#". It did. then I said "now cobol". It did.
Whole process less than two minutes.
You can't do anything like that in Google.
Mind you, I didn't run the code, though it looked sane!
|
You’re retired now Crankcase aren’t you?
;-)
|
>We use rules based underwriting for the small simple stuff...
I hope they didn't resurrect my code?
In the late 80s I was sent on an LISP and AI training course because a large customer (with a griffin logo) were looking to use AI to streamline personal loan applications.
The application basically took the answers from a re-designed loan application form and applied a set of rules with 'confidence factors'. Some of the rules would be considered quite un-PC nowadays but were backed up with statistical data.
The application 'learned' by modifying confidence factors or requesting more information or instructions from a human that was then added to it's knowledge base. The difficulty was in maintaining the quality of that knowledge base.
I think they were a bit scared of it at the time and adopted a much simpler fixed rule 100% yes or no approach.
|
Tried the suggestions to get into ChatGPT and none worked. It can show me previous answers but not regenerate them. A new account didn't help. I've had a hunt around the internet but no luck with any sensible suggestions yet, though it does seem to be a reasonably common problem.
I can live without it...
Last edited by: smokie on Thu 16 Mar 23 at 20:39
|
Well, I Googled ChatGPT, watched a short video with a fast talking young American, followed the prompts, set up an account, opened it, asked it to write me a speech and Bingo!
OK, a lot of it I won't use, but with a bit of copying and pasting, it has saved me a lot of brainwork. Which has got to be good.
|
|
...but did you ask it why Labour members thought Corbyn was such a good idea...?
|
>> ...but did you ask it why Labour members thought Corbyn was such a good idea...?
>>
I missed that last night, but it might be amusing?
|
...but did you ask it why Labour members thought Corbyn was such a good idea...?
The decision to choose Jeremy Corbyn as the leader of the UK's Labour Party in 2015 was a result of a combination of factors, including his grassroots appeal, his ability to energize the party's base, and his perceived authenticity.
Many Labour Party members saw Corbyn as a principled and honest politician who had spent his entire political career fighting for social justice and equality. He was seen as a refreshing change from the party's previous leaders who were seen as too centrist and too closely aligned with corporate interests.
Corbyn's anti-austerity platform and his commitment to nationalizing key industries such as the railways, energy companies, and the Royal Mail also appealed to many party members who felt that the previous Labour government had not done enough to address income inequality and social injustice.
Corbyn's opposition to the Iraq war and his support for Palestine also resonated with many left-wing members of the party who felt that the party needed to take a stronger stance on issues of international justice and peace.
It's worth noting, however, that Corbyn was a controversial figure within the party, and his leadership was marked by internal divisions and conflict. While he was able to mobilize a large base of support, he was also criticized for his handling of issues such as anti-Semitism within the party and his perceived lack of leadership on Brexit.
Ultimately, Corbyn's leadership of the Labour Party ended in disappointment, with the party suffering a crushing defeat in the 2019 general election. However, his legacy continues to shape the party's policy platform, with many of his ideas and principles remaining influential within the party.
What do you think of that - Manatee?
Last edited by: Duncan on Fri 17 Mar 23 at 07:59
|
>> Well, I Googled ChatGPT, watched a short video with a fast talking young American, followed
>> the prompts, set up an account, opened it, asked it to write me a speech
>> and Bingo!
>>
>> OK, a lot of it I won't use, but with a bit of copying and
>> pasting, it has saved me a lot of brainwork. Which has got to be good.
>>
Good stuff. Once it’s done it, you can just say something like “make it chattier”, or “more formal”, or “aimed at kids” or “businesslike”, etc. or say you don’t like the bit about xyz, so just add in something about abc instead. Make it shorter or longer. Write an executive summary. Anything you want.
|
Chatgpt and cars - zippy
