Google, man, and ape

What, no comments on the recent Google Image Recognition 
flap?  You know,  which identified a pair of black hominids as gorillas -

It's funny, but even funnier, and astonishing, is that Google 
APOLOGIZED; yes, they apologized for an ALGORITHM!  
bad algorithm, naughty algorithm!  hmmmm... why doesn't 
the algorithm issue its own apology, why go through corporate PR?  

Didn't Google perform due diligence before hiring, did they 
discover any racist affiliations in its background?

And then they announced they'll FIX IT, toute suite!  Of 
course that sounds reasonable to Joe Sixpack, but I 
assume everyone here understands how preposterous 
that is. on many levels - 

--
Rich

On Sun, 05 Jul 2015 20:13:32 -0700, RichD wrote:

> What, no comments on the recent Google Image Recognition flap?  You
> know,  which identified a pair of black hominids as gorillas -
> 
> It's funny, but even funnier, and astonishing, is that Google
> APOLOGIZED; yes, they apologized for an ALGORITHM!
> bad algorithm, naughty algorithm!  hmmmm... why doesn't the algorithm
> issue its own apology, why go through corporate PR?
> 
> Didn't Google perform due diligence before hiring, did they discover any
> racist affiliations in its background?
> 
> And then they announced they'll FIX IT, toute suite!  Of course that
> sounds reasonable to Joe Sixpack, but I assume everyone here understands
> how preposterous that is. on many levels -

Linx?

If it's what I suspect it is, they showed the algorithm a bunch of 
pictures of white guys, a bunch of pictures of chimps, and then released 
it on an unsuspecting world.

Because, after all, "human" is what you find in an office tower in San 
Jose, yes?

-- 
www.wescottdesign.com

On Sun, 5 Jul 2015 20:13:32 -0700 (PDT), RichD
<r_delaney2001@yahoo.com> wrote:

>What, no comments on the recent Google Image Recognition 
>flap?  You know,  which identified a pair of black hominids as gorillas -
>
>It's funny, but even funnier, and astonishing, is that Google 
>APOLOGIZED; yes, they apologized for an ALGORITHM!  
>bad algorithm, naughty algorithm!  hmmmm... why doesn't 
>the algorithm issue its own apology, why go through corporate PR?  
>
>Didn't Google perform due diligence before hiring, did they 
>discover any racist affiliations in its background?
>
>And then they announced they'll FIX IT, toute suite!  Of 
>course that sounds reasonable to Joe Sixpack, but I 
>assume everyone here understands how preposterous 
>that is. on many levels - 

Hi,
  That must be the same algorithm IPCC used to 
predict global warming.

[-Rick-]

On 7/5/2015 10:13 PM, RichD wrote:
> What, no comments on the recent Google Image Recognition
> flap?  You know,  which identified a pair of black hominids as gorillas -
>
...
> Rich
>


Fyi, may be becuase they saw Wolfram release the image identification project?

https://www.imageidentify.com/

First announced here

http://blog.wolfram.com/2015/05/13/wolfram-language-artificial-intelligence-the-image-identification-project/

"�What is this a picture of?� Humans can usually answer such
questions instantly, but in the past it�s always seemed out
of reach for computers to do this"

--Nasser

On 7/5/2015 11:13 PM, RichD wrote:
> What, no comments on the recent Google Image Recognition
> flap?  You know,  which identified a pair of black hominids as gorillas -
>
> It's funny, but even funnier, and astonishing, is that Google
> APOLOGIZED; yes, they apologized for an ALGORITHM!
> bad algorithm, naughty algorithm!  hmmmm... why doesn't
> the algorithm issue its own apology, why go through corporate PR?
>
> Didn't Google perform due diligence before hiring, did they
> discover any racist affiliations in its background?
>
> And then they announced they'll FIX IT, toute suite!  Of
> course that sounds reasonable to Joe Sixpack, but I
> assume everyone here understands how preposterous
> that is. on many levels -

I found something even sillier!  That anyone gives a rat's ass about it!!!

-- 

Rick

On July 6, Nasser M. Abbasi wrote:
> > What, no comments on the recent Google Image Recognition
> > flap?  You know,  which identified a pair of black hominids as gorillas -
> 
> Fyi, may be becuase they saw Wolfram release the image identification project?
> https://www.imageidentify.com/

huh?
 
> First announced here
> http://blog.wolfram.com/2015/05/13/wolfram-language-artificial-intelligence-the-image-identification-project/
> 
> ""What is this a picture of?" Humans can usually answer such
> questions instantly

yep, instantly and easily.
For instance, we wouldn't tolerate Robo-cop, because 
who would trust a machine's image processing and 
decision making, for such crucial duty?  Like, what 
if it had to distinguish a toy gun from a real, a 
potential tragedy?  Only a man can do a man's job - 
http://www.youtube.com/watch?v=dwCASkcKs4w


ditto with self-drive cars - think of the hazard, 
the injury, with a computer at the steering wheel!  
No sir, give me reliable organic tissue - 

--
Rich

On 7/7/2015 3:31 PM, RichD wrote:
> On July 6, Nasser M. Abbasi wrote:
>>> What, no comments on the recent Google Image Recognition
>>> flap?  You know,  which identified a pair of black hominids as gorillas -
>>
>> Fyi, may be becuase they saw Wolfram release the image identification project?
>> https://www.imageidentify.com/
>
> huh?
>
>> First announced here
>> http://blog.wolfram.com/2015/05/13/wolfram-language-artificial-intelligence-the-image-identification-project/
>>
>> ""What is this a picture of?" Humans can usually answer such
>> questions instantly
>
> yep, instantly and easily.
> For instance, we wouldn't tolerate Robo-cop, because
> who would trust a machine's image processing and
> decision making, for such crucial duty?  Like, what
> if it had to distinguish a toy gun from a real, a
> potential tragedy?  Only a man can do a man's job -
> http://www.youtube.com/watch?v=dwCASkcKs4w
>
>
> ditto with self-drive cars - think of the hazard,
> the injury, with a computer at the steering wheel!
> No sir, give me reliable organic tissue -

Rather dry sense of humor you have there.

-- 

Rick

On July 5, Tim Wescott wrote:
>> What, no comments on the recent Google Image Recognition flap?  >> You know,  which identified a pair of black hominids as gorillas -
>> It's funny, but even funnier, and astonishing, is that Google
>> APOLOGIZED; yes, they apologized for an ALGORITHM!
>> bad algorithm, naughty algorithm!  
>> Didn't Google perform due diligence before hiring, did they
>> discover any racist affiliations in its background?
>> And then they announced they'll FIX IT, toute suite!  
>> Of course that sounds reasonable to Joe Sixpack, but I assume 
>> everyone here understands how preposterous that is
> 
> If it's what I suspect it is, they showed the algorithm a 
> bunch of pictures of white guys, a bunch of pictures of chimps, 
> and then released it on an unsuspecting world.

heh
It seems you have some inkling of the method.

The hot thing in AI is 'deep belief networks', which 
is likely what Google uses.  It's neural nets, but 
they finally figured out how to build multi-layer 
nets, which for various information-theory reasons, 
offer superior performance. 

The point is, there's no 'human' or 'gorilla' in 
the code.  It teaches itself, from massive databases.
Improvement means tweaking the learning algorithms.

That's why their official response is so 
ludicrous... "we're sorry, we'll fix it!"  In other 
words, "we'll simply ratchet up the accuracy, we 
hadn't thought of that before, but now, NOW, it's urgent!"

And no doubt the boys in the lab are likewise rolling 
their eyeballs - 


Bing 'error rate' 

--
Rich