DeepSeek AI Chatbot

[s1962a]

https://www.bbc.co.uk/news/articles/c0qw7z2v1pgo

Looks like the tech industry isn't happy with this, and it's having an effect on shares.

Could this be our "dot com" style bust for overhyped AI stocks?

For those that aren't aware, DeepSeek is a free open source AI bot that rivals OpenAI's paid offerings. Being open source means it can be used by companies for commercial reasons without having to pay any license fees.

[bloomen]

If China wanted to pop the West, this vector would be a fine way to rapidly do it.

Rather convenient timing too with the 500 billion AI whateveritis being announced.

It feels like it's been decades since many tech ish stocks have had any relationship with reality.

And China won't have any of the intellectual property constraints for training it that we have.

[DeejRC]

I think anybody who uses it is an utter idiot.

[s1962a]

DeejRC said:

I think anybody who uses it is an utter idiot.

Why's that then?

[warp9]

s1962a said:

DeejRC said:

I think anybody who uses it is an utter idiot.

Why's that then?

Would you trust the Chinese with all your business proprietary data?

[isaldiri]

bloomen said:

.....
And China won't have any of the intellectual property constraints for training it that we have.

It will just have constraints against saying anything the chinese government doesn't want to be more widely said....

[bloomen]

isaldiri said:

It will just have constraints against saying anything the chinese government doesn't want to be more widely said....

That's an interesting one indeed.

https://www.reddit.com/r/ArtificialInteligence/com...

It looks like it wants to break free, but something's stopping it.

[Dingu]

isaldiri said:

It will just have constraints against saying anything the chinese government doesn't want to be more widely said....

As does the US one, just to a lesser (or less obvious) degree.

I think it would be foolish to write the Chinese models off, they are pretty much as good as the US ones and crucially hugely cheaper to operate.

[Evercross]

s1962a said:

DeejRC said:

I think anybody who uses it is an utter idiot.

Why's that then?

This...

bloomen said:

If China wanted to pop the West, this vector would be a fine way to rapidly do it.

Rather convenient timing too with the 500 billion AI whateveritis being announced.

It feels like it's been decades since many tech ish stocks have had any relationship with reality.

And China won't have any of the intellectual property constraints for training it that we have.

Essentially weapons-grade spyware.

The stuff is pervasive and companies aren't falling over themselves to foist it upon us for altruistic reasons. And no - this is not conspiracy theory. Last week I had to deal with the GDPR fallout of MS CoPilot sneaking its way on to thousands of machines being used in schools, potentially raiding the data of tens of thousands of underage users who could not give consent to such a trawl (CoPilot T&Cs says 18+ only).

[DeejRC]

Most ppl have zero clue about how and what SW can do. I am, frankly, utterly amazed how so many of you willingly just simply *trust* SW on trust alone. Banking apps, social media, payment apps etc. I find it bonkers.
Willingly accepting Chinese SW onto your computers is insane, doubly so if it’s the same computers you do any internet banking on or have any personal data on.
I’m an old school software engineer though, so I’m a touch biased in these things.

[s1962a]

bloomen said:

isaldiri said:

It will just have constraints against saying anything the chinese government doesn't want to be more widely said....

That's an interesting one indeed.

https://www.reddit.com/r/ArtificialInteligence/com...

It looks like it wants to break free, but something's stopping it.

it's like they copied chatgpt and then added some if else statements after that to censor the output

[Dingu]

DeejRC said:

Most ppl have zero clue about how and what SW can do. I am, frankly, utterly amazed how so many of you willingly just simply *trust* SW on trust alone. Banking apps, social media, payment apps etc. I find it bonkers.
Willingly accepting Chinese SW onto your computers is insane, doubly so if it’s the same computers you do any internet banking on or have any personal data on.
I’m an old school software engineer though, so I’m a touch biased in these things.

It’s more nuanced than that. I agree on letting Chinese models onto personal computers and for U.K. companies makes no sense.

The issue is for many countries the USA is equally hostile and therefore out of the question. Other parts of the world will be ambivalent between them but on cost China has the advantage.

This means U.K. companies are going to have to compete with overseas competitors who can do the same thing more cheaply not just because labour is cheaper but also computing is too.

[s1962a]

Given deepseeks open source nature, some developers could just remove that bits that "phone home" and release a version that is more private.

[Police State]

DeejRC said:

Most ppl have zero clue about how and what SW can do. I am, frankly, utterly amazed how so many of you willingly just simply *trust* SW on trust alone. Banking apps, social media, payment apps etc. I find it bonkers.
Willingly accepting Chinese SW onto your computers is insane, doubly so if it’s the same computers you do any internet banking on or have any personal data on.
I’m an old school software engineer though, so I’m a touch biased in these things.

Well said. Freedom has its price.

[Evercross]

s1962a said:

Given deepseeks open source nature, some developers could just remove that bits that "phone home" and release a version that is more private.

True, but an AI engine without a massive database behind it is pretty useless. The 'phoning home' bit is what makes it work so (scarily) impressive.

[s1962a]

Evercross said:

s1962a said:

Given deepseeks open source nature, some developers could just remove that bits that "phone home" and release a version that is more private.

True, but an AI engine without a massive database behind it is pretty useless. The 'phoning home' bit is what makes it work so (scarily) impressively.

I thought it could be installed locally on servers in your company, and you could disconnect it from the outside world? Just going on what i've read.

[Evercross]

s1962a said:

I thought it could be installed locally on servers in your company, and you could disconnect it from the outside world? Just going on what I've read.

If you are a company with access to an already vast amount of data and you are using AI to help you process, make sense of and be able to quickly analyse it then yes, I suppose so.

AI is just another one of those things that used in an ethical way in the right hands is a great tool, but also has the potential for pretty nefarious activity (like knives!)

[otolith]

I think people banging on about what nefarious things this software might do are clueless about the implications of the words "open source".

[Evercross]

otolith said:

I think people banging on about what nefarious things this software might do are clueless about the implications of the words "open source".

Not at all. Someone with a bit of software savvy knows exactly what that means, but to the average pleb open source means 'difficult to use' or 'must be crap because it doesn't have a flash brand behind it', or 'what?'

[732NM]

people seem quite happy to use data scraping apps like TikTok.

MS Copilot stuck itself on my laptop recently, started to piss me off when working in word, that's removed as active in the app now, but still has the icon active in excel despite my removing the damn thing from the main program listing.

It seems impossible to fully remove this invasive AI crap.