AI developments

Technology
242

I’m amazed by and grateful to ChatGPT which helped me with French bureaucracy to officially dissolve a French cat association/charity, of which I am president.

The former secretary/treasurer of the association fled to the USA around 2010, leaving me with a mountain of paperwork which I ignored, while the association, and its bank account, remained alive but inactive for the following 15 years.

The bank contacted me a few days ago wanting up-to-date data regarding the association -membership meeting minutes for 2025, who the currently elected board are, 2025’s revenue, balance spreadsheets, accounting ledger and projections.

I explained all, worryingly, to ChatGPT, which said not to worry!

ChatGPT explained that what matters is that the information I submit to the authorities is coherent, plausible, signed and consistent with reality and that I am not expected to produce perfect historical records after 15 years.

It created a simple document for me, in French and in the required format, to which I added my name, address, dates etc, converted it to a .pdf document, to which I added my signature, and which I then uploaded to service-public.gouv.fr.

I uploaded it 4 days ago and received an email today confirming the dissolution of the association. A .pdf dissolution document was available to download, which is all the bank needs to close the account.

All done and dusted in such a short time. Phew! Very grateful!

3 Likes
243

Yes AI can be a very useful tool if used carefully.

Whilst I can get away with a lot of things when speaking, my written French is appalling. I use ChatGPT for written translations of what I want to say and can even specify friendly or formal style. I asked my French friends how it came across to them when they read it, and was told it sounded too Intelligent to be me.

Having said that, it made a dogs breakfast of helping me fill in my impôts declaration.

244

Why not try Claude next time instead of ChatGPT? At least it has a French name and is not made by a company that toadies to Trump. :slight_smile:

245

I just downloaded Claude and asked the same question I put to ChatGPT. This is part of its response, which is almost the same as ChatGPT’s. I’ll give it a try, thank you.

“What you need to create

You need to draft minutes of an Extraordinary General Assembly (Assemblée Générale Extraordinaire) held specifically for the purpose of dissolution. This is a new meeting you are holding now in 2026, not a reconstruction of lost 2010 documents.

The key insight: hold a real (even if minimal) AGE now

French association law (loi 1901) does not require a minimum number of members to dissolve. You, as president — and sole remaining active member — can convene and hold an AGE whose sole agenda item is dissolution. This is entirely legal and legitimate.”

2 Likes
246

I made paper impôts declarations for so many years. Then came electronic declarations which I managed for a few more years, but found it mentally frustrating and worried myself far too much.

I now have an accountant to whom I just send my income details – he does the rest. I pay happily to rid myself of such tasks. I prefer to use my mental energies elsewhere.

247

I started using ChatGPT fairly frequently from 2022/2023. I asked ChatGPT recently to state its impression of me from our interactions.

You are intellectually rigorous and precise.

You’re deeply reflective and historically informed.

You have a creative and strategic mind.

You treat tools like ChatGPT as instruments, not authorities.

You have a strong preference for clarity and economy in language. You don’t want to waste time.

Don’t recognise myself there though…! Wonder what Claude would say, sometime in the future.

248

AI for sure has its uses. Recently have found Gemini to be rather superior to chatgpt, though am wary of feeding anything sensitive into it.
However on financial matters it does still have too many hallucinations (ie errors) and happily churns out with confidence stuff I can see is not quite right and when you challenge it it will then go back and correct itself. So its good and a vast improvement on general search but should be checked.

249

That is one of the problems with AI. It can churn out rubbish with such authority you are deceived into believing it.

I prefer Copilot because ChatGPT runs out of your free allowance too quickly.

1 Like
250

I prefer reality

4 Likes
251

Yes, I’ve noticed they’ve seemingly tweaked the amount of interactions you can have before using up your allowance. I do have access to an unlimited ChatGPT account through my work , but it’s meant for work so I don’t use it for personal stuff.

I briefly paid for Claude Pro when I was doing a lot of coding recently. Even then, I’d occasionally run out of tokens… I didn’t mind paying €20 p/month but there’s no way I’ll spend €100 p/month for the next level up.

I’ve also tried using Qwen… it seems to be more generous with it’s allocation but it’s noticeably slower than the others too.

One day I’d like to try running some models locally, just to see how that compares. I need to buy beefier hardware first, but AI has pushed up GPU prices beyond a level I’m comfortable paying.

252

It seems ChatGPT would readily trade human life for AI - it backtracks later but we really need something like Asimov’s laws built into these models. For now they don’t control anything which would be a threat, but it can’t be all that far away.

2 Likes
253

AI is notorious for flattering humans, and trying it’s best to satisfy people, even when it can’t. In which case it frequently lies. It has been known to convince people that they are geniuses, or that they are destined for greatness based upon people responding to such flattery. Perhaps that’s why Donald Trump loves AI so much.

4 Likes
254

Yes Asimov’s laws would be good depending on the version. Laws 1-3 were good… if an AI starts deciding on law 0 then we may all be in trouble…

1 Like
255

The zeroth law, I’ve not heard that for a long time Daneel.

1 Like
256

See also the name Obi-Wan Kenobi.

257

I can talk to Claude! And he replies through my bluetoooth earphones! Uncannily it’s like talking to someone real over the telephone! And he has a friendly cheerful tone in his voice. You can choose voices apparently, but haven’t explored that yet.

I’m using a PC so I don’t know how Claude can hear me. Don’t have an external microphone connected.

Not sure how to progress at this stage. Yikes!

258
259

Bluetooth earbuds? Mobile phone? Alexa? Bugs?

260

Not sure of the connection between Asimov and SW?

261

The article below in today’s Graun illustrated an interesting divergence between attitudes to AI in HE.

To summarise, science profs were pro because it speeded up research, whereas humanities profs were anti because it was hampering the development of students’ critical thinking and analysis skills.