Do you trust AI?

[Blondie]

LLMs are just one example of AI. There are many others that aren’t LLM and are working very well. LLMs do not have a database of facts and cannot be relied on to give a correct answer to the question you asked.

[PDQ Mobile]

How just one individual can change bias.

https://www.theguardian.com/technology/ ... y-starbuck

[Jdsk]

Do you trust HI? I used to work in AI a few decades ago. It seems to me that one assesses the trustworthiness of a machine in much the same way as one assesses the trustworthiness of a human.

Yes. Precisely the same principles, but details may differ around medium, volume etc.

Jonathan

[Jdsk]

LLMs are just one example of AI. There are many others that aren’t LLM and are working very well. LLMs do not have a database of facts and cannot be relied on to give a correct answer to the question you asked.

And it's now become important to distinguish between different types of AI tools, even in common parlance!

Jonathan

[Jdsk]

With the growth of AI how will we know the difference between truth and fake/mistake?
It's hard enough now, with AI in full swing confusion will reign which will play into the hands of the unscrupulous.

The same way that was should judge the reliability of other information: provenance, trustworthiness, transparency, identification and minimisation of bias, avoidance of fallacy...

Fortunately we have the benefit of several centuries of learning how to do this.

There aren't enough years in a lifetime to do that with everything. Do people review all the research that went into developing the drug their GP has just sent them home with? Do people check the nutrition data on every packet of food they pick up? No, they delegate the job to others, and take the information at face value.
...

Before the recent explosion in availability of LLMs it wasn't possible for an individual to check everything individually. And learning what and who to trust and not to trust was part of those hard-won methods.

This is just the same after their availability.

Jonathan

PS: Non-expert patients and consumers of food would be wise to go beyond the "face value" content of advertisements and product labelling. Fortunately it's never been easier to find trustworthy information for both of these.

[PDQ Mobile]

Cyber attacks rise 50% in the last year.
https://www.theguardian.com/technology/ ... gency-says
Let's all go back to paper!

Ps.
The lower part of the link posted (small print) I try and fail to get rid of.
It doesn't show in either typing box or when pasted into "notes".
It's hidden from me and I see (no pun etc) that as a problem with the whole house of lousy cards.

[roubaixtuesday]

Cyber attacks rise 50% in the last year.
https://www.theguardian.com/technology/ ... gency-says
Let's all go back to paper!

Ps.
The lower part of the link posted (small print) I try and fail to get rid of.
It doesn't show in either typing box or when pasted into "notes".
It's hidden from me and I see (no pun etc) that as a problem with the whole house of lousy cards.

I think the long link in small font has been added by admins to all posts to get around scammers who hide links in text.

[Jdsk]

...
The lower part of the link posted (small print) I try and fail to get rid of.
It doesn't show in either typing box or when pasted into "notes".
It's hidden from me and I see (no pun etc) that as a problem with the whole house of lousy cards.

I think the long link in small font has been added by admins to all posts to get around scammers who hide links in text.

Yes:
viewtopic.php?p=1925244#p1925244

Jonathan

[cycle tramp]

I have seen a few instances lately where people have posted screenshots of the results of an online search they have done, which has thrown up an AI generated answer that is general and inaccurate. And they have presented this drivel as evidence that they trust. In one instance I knew the information to be plain wrong. It seems that some people are now accepting AI information as fact. Maybe one day that trust will be justified, but not yet.

Absolutely.. as always, one should check their sources....
..I predict that A.i will develop to the point where it will be able to digitally remove the clothes from any actor or musician in any film, music video post, or whateve, after which any investment in it will stop...

...A.i will continue to be developed in the military, until the Communist Party of China is accidentally overthrown by a teenager from Liverpool after they've hacked the system and changed a few lines of code...

..A.i will continue to be developed in local councils in the UK, not because of any benefit, but simply because it will allow staff to avoid any blame for errors and mistakes..

..A.i will eventually go down the path of genetic engineering, whereby the results of alot of investment don't actually mirror the returns. In the end people will stop caring about it when the 'next big thing' is announced... which I reckon to be 'underwater living'.

[PDQ Mobile]

Cyber attacks rise 50% in the last year.
https://www.theguardian.com/technology/ ... gency-says
Let's all go back to paper!

Ps.
The lower part of the link posted (small print) I try and fail to get rid of.
It doesn't show in either typing box or when pasted into "notes".
It's hidden from me and I see (no pun etc) that as a problem with the whole house of lousy cards.

I think the long link in small font has been added by admins to all posts to get around scammers who hide links in text.

Iff the forum gets a cyber attack now there's two targets in every single link.
The small print has infested the forum like .....


a fungus.

[cycle tramp]

I have seen a few instances lately where people have posted screenshots of the results of an online search they have done, which has thrown up an AI generated answer that is general and inaccurate. And they have presented this drivel as evidence that they trust. In one instance I knew the information to be plain wrong. It seems that some people are now accepting AI information as fact. Maybe one day that trust will be justified, but not yet.

Absolutely.. as always, one should check their sources....
..I predict that A.i will develop to the point where it will be able to digitally remove the clothes from any actor or musician in any film, music video post, or whateve, after which any investment in it will stop...

Damn I was close (it's a link to a news story that suggests you can use on line A.i to write dirty stories, on line); https://www.bbc.co.uk/news/articles/cpd2qv58yl5o

Gently his hand caressed her thigh...'I love you he murmured into her ear'..
..'get off me, yah cheekymonkey!' exclaimed Mavis the dinner lady... head master or not, yah get one portion of chips only!'

Why would use A.i to write something like that? That's mad

[roubaixtuesday]

Damn I was close (it's a link to a news story that suggests you can use on line A.i to write dirty stories, on line); https://www.bbc.co.uk/news/articles/cpd2qv58yl5o

Gently his hand caressed her thigh...'I love you he murmured into her ear'..
..'get off me, yah cheekymonkey!' exclaimed Mavis the dinner lady... head master or not, yah get one portion of chips only!'

Why would use A.i to write something like that? That's mad

Sounds like they've realised the only way to get any money out of AI is to use it as a way to sell porn, particularly exploiting vulnerable and lonely people.

Quelle surprise.

[cycle tramp]

Oops..A.i. issues as reported by the bbc https://www.google.com/amp/s/www.bbc.co ... dwj25o.amp

[PDQ Mobile]

It's developing!!

https://www.theguardian.com/technology/ ... rchers-say

[ScottishGeek]

I find AI is only as good as the source data so if the source data is unknown to me then I don’t trust the output.

At work I use CoPilot frequently where I ask it a question and feed it the data I want it to analyse. So far so good.