previous discussion: https://news.ycombinator.com/item?id=22812509

Good. Make it EXPENSIVE and thus purge it from the web.

Agreed. I hate it. Luckily they is a way around it: https://github.com/dessant/buster

That doesn't seem like a way around the problem of how to identify bots without pissing off humans. If anything it will make it worse!

You don't. Bots will always be present, always one step ahead. Honestl, why bother?

Because people don't like spam?

You'll always get spam, email for example. You're not required to solve a stupid task before sending it. The recievers mail client can filter basic spam from real mails. Yes, there are false positives and you'll have to look in the spam folder. But you will do it carefully, not get phished that easily and see no sketchy pill ads.

Bots and spam will always exist and training googles AI to detect stop signs won't change that.

How do you propose to solve the problem Captchas are designed to solve, if not through Captchas?

The question is more one of whom to trust than of telling humans from computers.

That remains a Hard Problem.

Maybe finally I’ll be able to browse the Web in private mode via vpn instead of useless clicking tiles with cars or traffic lights.

I wouldn't mind CAPTCHA hell so much if the tasks I was doing were benefiting libre projects that everyone could draw on, like OpenStreetMap or Wikipedia. The offensive thing about CAPTCHA is that your work is for the sole sake of improving one corporation’s product.

That would be a great idea for a captcha project! You might wanna post to open-source-ideas: https://github.com/open-source-ideas/open-source-ideas

There are already alternative CAPTCHA projects. The problem is that the most-visited websites on the internet never adopted them, instead content to use Google's solution.

I thought this was in lieu of payment...

Tell me about it. Even using Firefox with a couple of adblocking lists is enough for Google to think I'm high risk enough to need a half dozen captchas.

I know right, everytime there's a website with reCaptcha on it it makes me fail two times to pass at the third. Tiles takes ages to load and fade to display new ones.

I like to do comparisons with Chromium based browsers on the exact same site, how much time it takes to pass the captcha. Obviously the chromium based one is way faster to solve.

I don't really know why Google makes it this way. If it's only because of my Firefox configuration that makes Google think I'm a bot. Or if Google can't retrieve enough data on my browsing habits to bother me and force me (in a way) to browse on a chromium based browser.

The thing is I never saw informations talking about it, so I'll pass on denouncing Google's behavior.

For those interested there's a good browser add-on to bypass reCaptchas : Buster [1]

[1] : https://addons.mozilla.org/en-US/firefox/addon/buster-captch...

Exactly the same story, exactly the same solution. What I didn't know: my ad blocker and Buster are from the same person, good guy this raymond hill.

Yeah. Feels like I'm being punished for daring to resist their surveillance capitalism.

They use hCaptcha now which is even worse as it requires us to do two tasks.

Unfortunately, sites are far more likely to just block such a risky connection and write off the opportunity cost.

Given how abusive recaptcha is for anyone running a even a minimally tracking-hardenned browser, dumping it is not just a matter of cost savings - it's good business as far as conversion rates are concerned.

I want to be happy about this. I look forward to a future where I never have to see another fire hydrant, bus or stop light. At times I identify so many traffic landmarks I feel like I just went through a commute to get to the thing I wanted to see.

I don't think that's our future though. It appears the replacement is much the same. From the article.

> Intuition Machines usually makes money by renting access to hCaptcha to companies who want to run image classification experiments, and then pay website owners to implement its hCaptcha product.

As usual you shouldn't depend on Google for anything (especially if it's free) but I wonder what motivated them to charge for reCAPTCHA.

For Gmaps I could see the reason: we have the best map so you're going to pay a lot of money if you want to use it.

But I fail to see the reason for reCAPTCHA if they still intend to build their self driving car: the more classified data they have about the world the better and reCAPTCHA is probably inexpensive to run.

The article hinted at this possibility, but perhaps Google has been using reCAPTCHA to train their systems for so long that the value of identifying yet another stoplight or crosswalk is no longer worth the cost of running the service to them.

Is there any information from Google on moving to a paid model for reCaptcha? I couldn't find anything. Is this only for large-volume customers like cloudflare, and thus not announced publically anywhere?

Probably don't need all the free AI training in that direction anymore. Just a guess.

Nice to see that the alternative (hCaptcha) supports PrivacyPass.

Does that imply Google has moved past supervised learning?

Hard to draw that conclusion versus they feel like they finally have "enough" training data.