A.I. Check

NormalNorman

michaels said:

NormalNorman said:

If RR's wouldn't increase my tax code AI did!

I had a basic knowledge of some possible entitlements but AI made it clearer with instructions.

A welcome jump yesterday as an early Christmas present.

As a result of the market increase how much extra did you get on the shares you sold yesterday?

Sorry bud, I think you must have confused me with someone else? Cheers

Universidad

ali_bear said:

It is not being unkind, and anyway these constructs do not have feelings to hurt, so I think it is perfectly reasonable to be purely honest about what they are. 

I mean it's being unkind to the people who've designed these things, really. I know if I painted a fresco on a par with the Sistene Chapel, and someone said it was just a few splashes of paint, I might be little annoyed. Of course I doubt my fresco would be upset.

ali_bear said:

It is not "just advanced text-prediction" - it is fancy predictive data, and nothing more. 

It is more than predictive data. Although that is at the core of it, rather than something more magical (as it sometimes seems) there are two things worth noting. Firstly, though you don't deny this, what you can do with predictive data at sufficient scale is astounding. Secondly, a full GPT type system now integrates with other tools outside of pure statistical modelling, and it can go so far beyond the sum of those parts that it's leaving a lot on the table to say it is nothing more than predictive data. Predictive data is what makes it possible, but a computer is more than its CPU, and a modern GPT is more than its predictive models. Though the bubble may be popping due to chronic over-investment, I think the best use cases are still to be discovered. 

michaels said:

Off topic but am I the only one who uses please and thank you when addressing LLMs?  And am polite when pointing out errors and asking them to rework!

Some schools of thought suggest that asking in a polite manner is worth doing whatever your perspective: "Polite phrasing often provides more context and mirrors the style of human communication found in its training data, which tends to produce richer responses."

michaels

Some things AIs are very good at.  I was writing a dissertation, had done the intro, three main sections and needed a summary and conclusion.  Probably would have taken me 3-4 hours.  I asked an LLM and if did a really good job, I would not have been embarrassed to hand it in as it was but I spent about an hour polishing, changing emphasis in some areas and generally making sure the style was my own (to avoid the LLM checkers saying it was done by one)  Massive productivity gain.  And I suspect it won't be long before they are better than me at this sort of task where all the relevant info and context is provided in a single source rather than needing to collate from and evaluate different sources.

ali_bear

Well I did say fancy .. 

But I've never cheated on any academic qualification!   

sgx2000

Almost inevitable that at some point A.I. will become sentient..
After all the human brain is just neurones.

At that point, it really needs to think us humans are cute pets.....

Moonwolf

michaels said:

Some things AIs are very good at.  I was writing a dissertation, had done the intro, three main sections and needed a summary and conclusion.  Probably would have taken me 3-4 hours.  I asked an LLM and if did a really good job, I would not have been embarrassed to hand it in as it was but I spent about an hour polishing, changing emphasis in some areas and generally making sure the style was my own (to avoid the LLM checkers saying it was done by one)  Massive productivity gain.  And I suspect it won't be long before they are better than me at this sort of task where all the relevant info and context is provided in a single source rather than needing to collate from and evaluate different sources.

I’m not convinced that it will definitely get better. There is a real risk of model breakdown as increasingly LLMs are trained on new documents that are partly written by AI.

As there are only a few algorithms they might promote and persist certain tics or style weaknesses. Just like inbreeding in genetics, think haemophilia in the European royal families.
 
It is possible that in the future there will be a premium on content written before AI or that can be proven to be before AI.