ððŪð―ðð°ðĩðŪ’ð ððķðŧðą ðŪðđðđðīðēðīðēðŧððŪĖðŋððķðī ððŧðą ððžðđðđðēðŧ ðŪðđð ððķð°ðĩððķðīðēðŋ ðĶðķð°ðĩðēðŋðĩðēðķðððšðēð°ðĩðŪðŧðķððšðð ðąðŪðģðĖðŋ ððžðŋðīðēðŧ, ðąðŪðð ð ðŪðð°ðĩðķðŧðēðŧ ððķð°ðĩ ðŧðķð°ðĩð ðŪðđð ð ðēðŧðð°ðĩðēðŧ ðŪðððīðēðŊðēðŧ. Incidentally, the term stands for “ðð°ðŪðąððĶðĩðĶððš ððķðĩð°ðŪðĒðĩðĶðĨ ððķðĢððŠðĪ ððķðģðŠðŊðĻ ðĩðĶðīðĩ ðĩð° ðĩðĶðð ðð°ðŪðąðķðĩðĶðģðī ðĒðŊðĨ ððķðŪðĒðŊðī ððąðĒðģðĩ” and thus already points to the problems of artificial intelligence in its definition.
A few thoughts on this:
– current AI models can recognize traffic lights on images better than most people.
– the providers of captchas are often also those who are currently big players in AI models (e.g. Google reCAPTCHA). Google reCAPTCHA)
– every selection of a traffic light by a human also represents an image classification that can be used for AI training
– if, for example, unknown captchas are played out, the user is misused for labeling (even if the providers would probably never admit that they use captcha data for AI training, but what the heck…)
So if an AI is now trained with captcha’s based on human-made labels, it will soon be better than any human. Following the logic, the captcha’s would now also have to become so much more complicated that only humans would be able to answer them. However, this will generate new training data, the AI will also get better, etc. etc.
ðððīð°, ðļðĶðŊðŊ ðŠðĪðĐ ðŪðŠðģ ðķĖðĢðĶðģððĶðĻðĶ, ðļðŠðĶ ðĐðĒĖðķð§ðŠðĻ ðŠðĪðĐ ðĐðĶðķðĩðĶ ðīðĪðĐð°ðŊ ðĨðĒðģðĒðŊ ðīðĪðĐðĶðŠðĩðĶðģðĶ ððŪðąðĶððŊ ðŧðķ ðĶðģðŽðĶðŊðŊðĶðŊ, ðĨðĒðŊðŊ ðļðĶðģðĨðĶ ðŠðĪðĐ ðŪðŠðĪðĐ ðĢðĒððĨ ðŊðŠðģðĻðĶðŊðĨðī ðŪðĶðĐðģ ðĒðŊðŪðĶððĨðĶðŊ ðŽð°ĖðŊðŊðĶðŊ (ððĒðĪðĐ).
ððēððŪðīððē ð§ðĩðēððē: ððŪð―ðð°ðĩðŪ’ð ððēðŋðąðēðŧ ðŊðŪðđðą ðąðŪðð ðģðĖðĩðŋðēðŧ, ðąðŪðð ððķð°ðĩ ðēðķðŧðē ðð ððķðēðđ ðđðēðķð°ðĩððēðŋ ðŪðđð ð ðēðŧðð°ðĩ ðŪðððīðēðŊðēðŧ ðļðŪðŧðŧ, ðŪðđð ðēðķðŧ ð ðēðŧðð°ðĩ. ðĶð―ðŪĖððēðððēðŧð ðąðŪðŧðŧðŧ ððēðŋðąðēðŧ ððķðē ððžðŧ ððŧððēðŋðēðŧ ðĶð°ðŋðēðēðŧð ððēðŋðð°ðĩððķðŧðąðēðŧ, ððēðķðđ ððķðē ðð°ðĩðđðķð°ðĩðððēðī ðļðēðķðŧðēðŧ ðĶðķðŧðŧ ðšðēðĩðŋ ðšðŪð°ðĩðēðŧ.
I’m generally not mad if I no longer have to look for traffic lights – but what I want is an AI that can pretend to be human everywhere.
ððŪð ðĩðēðķÃð: ðķðŧ ð―ððŧð°ððž ðððŊðēðŋ-ðĶðēð°ððŋðķðð, ððððĩðēðŧððķðģðķððķðēðŋððŧðī ðĩðŪðŊðēðŧ ððķðŋ ðēðķðŧ ð§ðĩðēðšðŪ. ðĶðķð°ðĩðēðŋðĩðēðķð ðšðððð ððķð°ðĩ ðŪðð°ðĩ ðĩðķðēðŋ ðķðŧ ðŋðķðēððēðŧ ðĶð°ðĩðŋðķðððēðŧ ðąðēðŧ ððēð°ðĩðŧðķðð°ðĩðēðŧ ð ðžĖðīðđðķð°ðĩðļðēðķððēðŧ ðŪðŧð―ðŪðððēðŧ!
Exciting times!
#informatikersindcool#lasstunsampelnsuchen#cybersecurityistwichtig

