Is Elon Musk the greatest visionary or the greatest snake oil salesman of our age?

[pbsmooth]

Can't stand Space Karen but they deliberately exploded the rocket after the separation didn't work. It didn't just explode randomly. Those in the know seem to see it as a pretty successful 'mission' in terms of where SpaceX is trying to get to.

 

[littlebabyjesus]

Musk's 'untruthful things' statement is telling, it's that school of thought that says we shouldn't have greater representation of people of colour or women because 'that's not realistic', and that this is just the way the world is and it's fine for me (always said by someone white and usually a man or else a woman who seeks stability and 'protection' by siding with the men) and everyone else is just a fragile whinger.

It's worse than that even, as unfettered AI not only reflects bias, it amplifies it.

One real example I read was of an AI designed to allocate healthcare resources in the US. It worked out 'reasons' for the data distribution on its own in true AI black box style. When it was tested for bias, it was found that it had identified a correlation between low healthcare spending and good health. It had thus proceeded to allocate a smaller share of resources to predominantly black areas. It hadn't twigged that poor people also don't spend so much on healthcare.

It is very strange that the richest person in the world is also someone who either doesn't know or doesn't care about the most basic aspects of the projects he funds.

 

[Cloo]

The one I remember was someone had some resume-screening program for hiring at a big business in the US and they thought they'd got rid of obvious biases, but it picked up on 'playing lacrosse' as something desirable candidates shared. Which was something that was a clear indicator of privately-educated white men.

 

[ska invita]

That's a nice big thing to say. But how exactly are they going to achieve it? Doesn't say.



Whole point right now is that ChatGPT doesn't understand anything. It doesn't know anything, let alone what 'truth' is. It's still an open question as to whether or not this technology can ever achieve understanding.

Ironically enough, this is an example of Musk doing the thing ChatGPT is prone to – bullshitting. He says it for effect, to please/persuade the audience, with no regard for the truth, hoping that others won't check up on what he's saying and call him on it. And they haven't called him on it in this article.

Tbf chatgtp is very polite and has clearly been trained so. It has an enlightenment, Wikipedia style . The weighting of responses is controlled and can readily be programmed to be weight differently. The question of political bias is a real one with AI, it's one of the danger areas 'pausers' are worried about

 

[littlebabyjesus]

The one I remember was someone had some resume-screening program for hiring at a big business in the US and they thought they'd got rid of obvious biases, but it picked up on 'playing lacrosse' as something desirable candidates shared. Which was something that was a clear indicator of privately-educated white men.

In the last few days, I've been learning about how to check for AI bias. Because it's black box, you have to use various indirect methods (which are pretty clever tbf) to work out how it arrived at its decisions. The checking itself is laborious and expensive. Those companies that have jumped in to AI are finding that it's not the bargain they thought it was.

 

[ska invita]

It's worse than that even, as unfettered AI not only reflects bias, it amplifies it.

Programming to Avoid Bias is the job of some people out there

 

[littlebabyjesus]

Tbf chatgtp is very polite and has clearly been trained so. It has an enlightenment, Wikipedia style . The weighting of responses is controlled and can readily be programmed to be weight differently. The question of political bias is a real one with AI, it's one of the danger areas 'pausers' are worried about

Yeah, but did you see that video of what it looks like when it isn't constrained? It feeds back on its loopiness and very quickly turns aggressive and abusive.

What Musk is proposing is an AI that isn't constrained but is allowed to 'search for truth'. That's where Musk enters bullshitting territory. We don't know how to create that yet.

Also, Musk blatantly is calling for a pause so that he can catch up.

 

[littlebabyjesus]

Programming to Avoid Bias is the job of some people out there

Yeah, I've been doing a short course on it - not with a view to doing it, just to learn how it works. Checking AI for bias is extremely involved and expensive.

 

[littlebabyjesus]

The fact that it is private companies developing these NLP projects scares me a bit. All of the research, including all the testing where it went wrong, needs to be in the public domain. This should be a public matter but we're woefully underprepared for it.

If that harms your business interests, tough. Don't do it then.

 

[kabbes]

Programming to Avoid Bias is the job of some people out there

And those people also exist in a context at a particular time in history. Bias can’t be “avoided”. It needs to be understood and mitigated.

 

[ska invita]

And those people also exist in a context at a particular time in history. Bias can’t be “avoided”. It needs to be understood and mitigated.

Yes obviously. chatgtp has it's own political flavour which has been chosen. It seems perfectly possible that another AI can have it's own flavour, whether that's driven by Musk or some other state

 

[Cloo]

This got me thinking about something I've started a separate thread about, if anyone's interested.... AI, bias and 'truth'

 

[kabbes]

Yes obviously. chatgtp has it's own political flavour which has been chosen. It seems perfectly possible that another AI can have it's own flavour, whether that's driven by Musk or some other state

It’s not just possible, it’s inevitable.

 

[tommers]

In amongst all of this they've finally removed blue ticks that aren't paid for.

 

[Lord Camomile]

I see people are already having fun with the Blue Checkalypse. Blue Checksodus? Blue Check-out-how-much-he's-fucked-it-up?

 

[Yossarian]

 

[Yossarian]

 

[Lord Camomile]

 

[tommers]

I give it a month.

 

[8ball]

There's a horrible picture outside the Wellcome institute of an elderly woman holding a robot in her arms. It's an awful idea even without the capitalism angle. It's elderly abuse.

Is it the cute one that looks like a baby seal?

 

[Lord Camomile]

I give it a month.

Would you say...









...Twitter's time is up?

 

[8ball]

View attachment 371516

Consensus reality is like so 2012.

 

[Lord Camomile]

View attachment 371517

 

[Lord Camomile]

It continues to be not simply 'a bad idea', but just an absolute fucking joke of a situation

 

[8ball]

It continues to be not simply 'a bad idea', but just an absolute fucking joke of a situation

Very bizarre.

 

[tommers]

 

[tommers]

They've just totally destroyed it as any reliable source of information. I mean. Fair play.

 

[Sue]

They've just totally destroyed it as any reliable source of information. I mean. Fair play.

Did people think it was? I mean I don't use twitter but I've never thought of it as reliable tbh.

 

[Orang Utan]

arf

 

[JimW]

Did people think it was? I mean I don't use twitter but I've never thought of it as reliable tbh.

You used to be able to reliably know that a post was made by a given person or organisation and now you don't.