AI successes

[swee'pea99]

I have found that Google or AI has helped frame more pertinent questions for my GP about my wife’s illness problems. Our GP surgery are not renowned for their proactivity, but are very good at treating symptoms.

I find that if you frame increasingly precise questions you can back it into some quite surprising admissions...or perhaps that's not the right word, since, as it's at pains to insist, when teased, it is not Google but Google AI, which is something quite different...but some quite surprising statements at any rate, eg:

and

and

Google wouldn't give you that. But Google AI does. I find that interesting at the very least.

 

[briantrumpet]

I find that if you frame increasingly precise questions you can back it into some quite surprising admissions...or perhaps that's not the right word, since, as it's at pains to insist, when teased, it is not Google but Google AI, which is something quite different...but some quite surprising statements at any rate, eg:

View attachment 16565

and

View attachment 16566
and

View attachment 16568

Google wouldn't give you that. But Google AI does. I find that interesting at the very least.

Well, I found this with Duck Duck Go with one question, and DDG doesn't use AI. I positively enjoyed old-fashioned Google, which effectively trained you to ask the right questions. And that ties in with my general thing about learning that it's all about learning to ask the right questions - and generally, every answer ought to lead onto another question. The more you understand the subject you're researching, the better will be your questions, until you find the ones that no-one else has asked, and you use your understanding to push the boundaries.

https://www.boltive.com/articles/wh...lessons-from-googles-425-million-privacy-fine

Amusingly, I was offered a one-button opt-out.

As I've said several times, the fraud with AI is the LLMs that make the language output seem human, when it's anything but, and can never understand. They are just wrapping an admittedly impressive search engine with human-like language to make you think that it understands and can weigh up the veracity of competing views.

 

[swee'pea99]

Well, I found this with Duck Duck Go with one question, and DDG doesn't use AI. I positively enjoyed old-fashioned Google, which effectively trained you to ask the right questions. And that ties in with my general thing about learning that it's all about learning to ask the right questions - and generally, every answer ought to lead onto another question. The more you understand the subject you're researching, the better will be your questions, until you find the ones that no-one else has asked, and you use your understanding to push the boundaries.

https://www.boltive.com/articles/wh...lessons-from-googles-425-million-privacy-fine

Amusingly, I was offered a one-button opt-out.

As I've said several times, the fraud with AI is the LLMs that make the language output seem human, when it's anything but, and can never understand. They are just wrapping an admittedly impressive search engine with human-like language to make you think that it understands and can weigh up the veracity of competing views.


View attachment 16569

That is interesting, for sure, but it tells the story of when Google were caught out breaking the law; it says nothing about what they're doing, now, or why they're doing it. I absolutely agree it's about asking the right questions. You are of course absolutely on the money about the very cunning and calculated use of faux human language. If you accuse it of being disingenuous, it apologises, and it really does sound sorry! At which point, the temperature comes down a degree or two...which is of course the idea. I jotted down some thoughts along these lines while pursuing Google AI through a series of evasions, distractions, misdirections, and general smokescreenery...

It tells me I'm angry, which in a funny way takes some of the heat out of my anger. This is not by chance. It tells me it's embarrassed. It isn't of course; a computer can't be embarrassed. But the impression given is that of a decent guy who's doing his best. As against a computer which has been programmed to do its worst. This isn't by chance either.

These are two of its techniques. Another popular one is faux apology. "Aw shucks, I let you down again. Sorry." in various forms. Of course a computer can no more be sorry than embarrassed, but again, it is surprisingly effective in taking the sting out of the situation. Well I say 'surprisingly'; of course none of this is in any way by chance. This is design. This is intent. This is post-sale conversion of your hardware from a tool that belongs to you into a tool that belongs to you and a spy that reports and answers to them. The former is what you paid for; the latter, what they want you to have.

Another technique - following up its 'honest guy who's doing his best to help' schtick by positively inundating you with sources and references and stuff you can use to report succinctly and accurately to whoever just what Google has been up to. Chapter, verse, you name it. Oodles of it. And there is oodles. Google have been Very Bad Boys, and Google AI is falling over itself to spill the beans. Which is odd, until you notice that you are starting to feel a bit swamped. With a side order of 'clever people knew all this already; didn't you?' You think you've spotted something uniquely perfidious? Oh you sad, sad, man. This is common knowledge.

Oh well, if you insist on pursuing it, you might want to take a look at that....and that...and that..., like some snotty, motheaten professor disdainfully flinging 'Common knowledge' over the back of the sofa.

That one does rather dishearten you. You do start to lose faith. Maybe this is common knowledge. Maybe I'll just be showing myself up for the ignorant I am. Maybe I should just forget the whole thing, click Yes and forget about it. This too of course is not by chance.

The more I engage with it, the more intriguing I find it! And useful too. Huzzah! Win-win!

Which is not to say AI isn't bad, n'kay?

 

[briantrumpet]

That is interesting, for sure, but it tells the story of when Google were caught out breaking the law; it says nothing about what they're doing, now, or why they're doing it. I absolutely agree it's about asking the right questions. You are of course absolutely on the money about the very cunning and calculated use of faux human language. If you accuse it of being disingenuous, it apologises, and it really does sound sorry! At which point, the temperature comes down a degree or two...which is of course the idea. I jotted down some thoughts along these lines while pursuing Google AI through a series of evasions, distractions, misdirections, and general smokescreenery...

It tells me I'm angry, which in a funny way takes some of the heat out of my anger. This is not by chance. It tells me it's embarrassed. It isn't of course; a computer can't be embarrassed. But the impression given is that of a decent guy who's doing his best. As against a computer which has been programmed to do its worst. This isn't by chance either.

These are two of its techniques. Another popular one is faux apology. "Aw shucks, I let you down again. Sorry." in various forms. Of course a computer can no more be sorry than embarrassed, but again, it is surprisingly effective in taking the sting out of the situation. Well I say 'surprisingly'; of course none of this is in any way by chance. This is design. This is intent. This is post-sale conversion of your hardware from a tool that belongs to you into a tool that belongs to you and a spy that reports and answers to them. The former is what you paid for; the latter, what they want you to have.

Another technique - following up its 'honest guy who's doing his best to help' schtick by positively inundating you with sources and references and stuff you can use to report succinctly and accurately to whoever just what Google has been up to. Chapter, verse, you name it. Oodles of it. And there is oodles. Google have been Very Bad Boys, and Google AI is falling over itself to spill the beans. Which is odd, until you notice that you are starting to feel a bit swamped. With a side order of 'clever people knew all this already; didn't you?' You think you've spotted something uniquely perfidious? Oh you sad, sad, man. This is common knowledge.

Oh well, if you insist on pursuing it, you might want to take a look at that....and that...and that..., like some snotty, motheaten professor disdainfully flinging 'Common knowledge' over the back of the sofa.

That one does rather dishearten you. You do start to lose faith. Maybe this is common knowledge. Maybe I'll just be showing myself up for the ignorant I am. Maybe I should just forget the whole thing, click Yes and forget about it. This too of course is not by chance.

The more I engage with it, the more intriguing I find it! And useful too. Huzzah! Win-win!

Which is not to say AI isn't bad, n'kay?

And it's not that I never use it, but on balance I think the whole thing is based on all sorts of bad things, can't be trusted, is going to help burn the planet, is going to subjugate humans to awkward things to be managed, people not learning to use their brains properly, terrible terrible posters that all look the same, and wealth being the plaything of a few techbros & techsisters at the expense of everyone else, and people forgetting that despite all their flaws, humans are still fucking awesome and that their brains and human language are two of the most amazing things in the universe.

To be fair to thalidomide and asbestos, the dangers weren't known when they were being touted as wonderful things, but we do know that the AI industry is up to no good, the whole thing is a financial disaster waiting to happen, it is already trashing the environment, but they are desperate to get everyone irreversibly hooked on this fraudulent machine.

I wouldn't mind so much if they were selling it as Google-with-knobs-on, that comes at an admitted cost. But that's not their game.

 

[swee'pea99]

And it's not that I never use it, but on balance I think the whole thing is based on all sorts of bad things, can't be trusted, is going to help burn the planet, is going to subjugate humans to awkward things to be managed, people not learning to use their brains properly, terrible terrible posters that all look the same, and wealth being the plaything of a few techbros & techsisters at the expense of everyone else, and people forgetting that despite all their flaws, humans are still fucking awesome and that their brains and human language are two of the most amazing things in the universe.

To be fair to thalidomide and asbestos, the dangers weren't known when they were being touted as wonderful things, but we do know that the AI industry is up to no good, the whole thing is a financial disaster waiting to happen, it is already trashing the environment, but they are desperate to get everyone irreversibly hooked on this fraudulent machine.

I wouldn't mind so much if they were selling it as Google-with-knobs-on, that comes at an admitted cost. But that's not their game.

And breath....

Yup. Sounds pretty much spot on.

Don't get me wrong. I'm not some kind of cheerleader. I'm not saying woop woop AI isn't it great nothing to worry about. I fwd'd the thing about fewer than 10% of students actually doing what being at university is at least nominally about - learning to think - to the lads earlier. I think it's profoundly pernicious in all the ways you describe, and lots more besides. What I suppose I am saying is that I think there are more interesting and nuanced discussions to be had in this space than AI is bad, end of.

 

[Shortfall]

Fair's fair & all that. Anyway, a starter for ten...

About a year ago I bought a yoghurt-maker, to try making Greek yoghurt at home. M&S charge almost £3 a pot for the real thing, and it's only infected milk, right? How hard can it be? First thing I discovered - and this is a bit of an aside - is that the mesh strainer provided was utterly useless letting most of the yoghurt through along with the whey. Fixed that by 'punching' a double-layer of kitchen towel into the strainer as a proper filter - perfect! But I digress...

Moving on...

I started out using M&S real Greek yoghurt as my starter. And it went quite well. The yoghurt wasn't quite as nice as the original, but you'd expect that, they're Greek, and they're using Greek milk, which is probably different. Also it was a bit of a pain that almost a third of every batch went into making the next, but hey ho, so it goes, overall, we were happy with it. Roll on nine months or something and I suddenly think, what if I used pre-strained yoghurt as my starter rather than post-strained yoghurt? If I could, given that post-strained yoghurt is about half the volume, that would work mean effectively losing a sixth to the new batch, rather than a third. So I googled it.

And Google AI came back and said yes absolutely, use pre-strained yoghurt, about 2 to 3 tablespoons per litre of milk. And I thought, sorry? Run that one by me again? 2 to 3 tablespoons? That's about 40 mil. So I get back to Google AI and say I've been using 150 mil per batch per litre, as per my yoghurt-maker's intructions, and now you're saying use 2 to 3 tablespoons. What gives?

And Google AI got back and said yes, well, the yoghurt-maker makers want to make sure that whatever you start up you'll end up with yoghurt. That way, they're in the clear. You use old yoghurt, thin yoghurt, tired yoghurt, cheap, watery yoghurt, you use 150 mils of it per litre you're gonna end up with yoghurt. Case dismissed. The problem is if you use 150 mil of good yoghurt your yoghurt to milk ratio is completely wrong. Bluntly, the culture doesn't have enough milk to work on, so it starts turning on itself. It's a different chemical reaction, and you end up with yoghurt that is yoghurt all right, but doesn't taste as nice. Trust me. Try it. Put two or 3 tablespoons in next time and see how it goes.

And I did. And guess what. It was really nice. Much nicer than before. And also, because I'm now using 40 mils of pre-strained yoghurt as my starter instead of 150 mils of post strained yoghurt I'm using about 1/24 of each batch for the next one rather than a third.

And that's entirely down to AI. Because if I just Googled without, what would Google do? Question about yoghurt-makers? - put up 10 yoghurt-maker makers' websites. Who else you gonna call, if the answer is to do with how to use a yoghurt maker? Plus if they go and buy a yoghurt maker at any of those links, we get money.

I think that is a key difference with AI: it engages with you, rather than just offers you links. You can hone in. You can specify. You can set parameters. And all of that can make a big difference to the accuracy of the answer you get. One of the key things it can do, if you brief it accordingly, is disregard anyone with skin in the game, often resulting in a less biased, more accurate, answer.

I can't see this thread reaching 70-odd pages, , but I do think there's an interesting other side to the story, and that it's all a bit more nuanced than makes a good popular narrative.

Buy a piece.of muslin cloth or else use a clean tea towel and strain it through a sieve for an hour or 2 for extra thick Greek style yoghurt. Use the whey to make pancakes, smoothies, bread etc etc. Apologies.to thread police for non AI content 😘

 

[swee'pea99]

Buy a piece.of muslin cloth or else use a clean tea towel and strain it through a sieve for an hour or 2 for extra thick Greek style yoghurt. Use the whey to make pancakes, smoothies, bread etc etc. Apologies.to thread police for non AI content 😘

Thanks! I'm genuinely embarassed and ashamed to say it's never once ocurred to me to do anything with the whey other than pour it down the sink. I shall look into it. I did try using cloth for the sieving, but it turned out to be a real faff, with the yoghurt getting stuck and/or the mesh being just too coarse/fine. Two bits of kitchen roll, perfect performance, pop it in the bin. Bish bash bosh. Bugger the baby polar bears.