Filling out captchas helps google train AI models

Steve Kinney admin at pilobilus.net
Wed Jun 29 16:21:54 PDT 2016


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1



On 06/18/2016 04:23 AM, Georgi Guninski wrote:
> On Fri, Jun 17, 2016 at 10:29:17PM +0000, Mr Nobody wrote:
>> I just discovered that the reCaptchas that you usually bump into
>> when using tor are not only a way to distinguish humans, but also
>> they use your replies to train artificial intelligence modes
>> 
>> https://www.google.com/recaptcha/intro/index.html#creation-of-value
>>
>> 
<https://www.google.com/recaptcha/intro/index.html>
>> 
>> Tor users helping google with image recognition lol
>> 
> 
> When over tor, some captchas with distorted images are so weird, I
> am not sure solution exists at all...
> 
> Is it possible this to be on purpose?

I have run into sites where their 100% insoluble captchas were used to
support the pretense of a "free" service option for non-subscribers.
Apparently their SEO consultants* saw some benefit in annoying the
hell out of users and discouraging them from believing that the sites
in question work at all, paid account or not.

Also, what better way for a centralized captcha provider to sabotage a
website than to flag it to receive insoluble captchas?

* I originally typed "crooked SEO consultants" but that is redundant.


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)

iQEcBAEBAgAGBQJXdFgSAAoJEECU6c5XzmuqbYMH/AqYEC1imC5cQn37yNA0uDlC
c3n872CZwVo9x/7VqIlYaoiVSXllNz9pXQxoZTfBisX4sAdP4KOicrNylSzQr9Pn
4/3ZVKF60F0HWRgX3Fqy6+4mcYXPd1OzoDYBIK2wcBEbRU8oNFFXgFQt5DV+VMED
PU84QOiylOq3rryw+muuWp/h23UkMun2C+HSS8AZZ9MaWZu8HZU5CJQ14MH5+0ws
1AW8lFQffiLsojh46OBsYDGSkQqx1J1Sq6F40jUjrOVffEAAgNAGyxHfC3/x+mWs
pGUslRwUMPaApz8+xwRxsaKdh1PG1RVlhT6I+AG+uFjeU6GYleGb77aZnKTgq00=
=Zxpf
-----END PGP SIGNATURE-----



More information about the cypherpunks mailing list