I also reached out to them on Twitter but they directed me to this form. I followed up with them on Twitter with what happened in this screenshot but they are now ignoring me.
I also reached out to them on Twitter but they directed me to this form. I followed up with them on Twitter with what happened in this screenshot but they are now ignoring me.
Nah, it’s just a old school chat bot following a predefined flow chart. And in this flowchart someone implemented an improper email check.
It’s pretty much the same as if there was just a website with an email field which then complains about a non valid email which in fact is very valid. And this is pretty common, the official email definition isn’t even properly followed by most mail providers (long video but pretty funny and interesting if you’re interested in the topic).
You can use symbols like [ ] . { } ~ = | $ in the local-part (bit before the @) of email addresses. They’re all perfectly valid but a lot of email validators reject them. You can even use spaces as long as it’s using quotation marks, like
"hello world"@example.comA lot of validators try to do too much. Just strip spaces from the start and end, look for an
@and a., and send an email to it to validate it. You don’t really care if the email address looks valid; you just care whether it can actually receive email, so that’s what you should be testing for.Not even a dot: TLDs are valid email domains. joe@google is a correct address.
Mmm… That doesn’t seem right, it’s usually gotta be fully expanded to at least a particular A record/MX.
How would you tie the tld itself to an MX?
TLD is just another DNS layer, try an SOA or NS lookup for “com.” those are obviously hosted somewhere. Hell the “.” at the end is even another layer with the root nameservers. You’d probably trip up a bunch of systems that filter on common convention rather than the actual RFC, but you could do it.
How the hell were the original rfc designers so creative as to result in such a flexible system?? It’s gets crazier the more you look at it.
It makes the system as a whole simpler. Your computer only needs to remember one root DNS server (although most computers allow setting 4 for redundancy) as opposed to one DNS server for each TLD, and it also makes adding TLDs easier.
To this point, there’s a website dedicated to the subject. Some of the regexes get pretty wild…
https://emailregex.com/
Yea but most of the time its more important to block code injection than to have the last promille of valid mail adresses be accepted.
You’re not going to get code injection via an email address field. Just make sure you’re using prepared statements (if you’re using a SQL database) and that you properly escape the email if you output it to a HTML page.
I think emailregex.com offers best of both worlds.
The first rule of tautology club is the first rule of tautology club.
I’m listening ;)
yes but that would be an AI still
A bunch of IF statements don’t qualify as an AI. That’s not how that works.
Yeah mate you’re talking out of your ass. A bunch of
ifstatements can, in fact, constitute an AI depending on the context. You don’t know what you’re talking about, stop trying to pretend you do.AI is a broad concept, a pathfinding algorithm can be considered AI, a machine learning image generator can be considered AI, a shitty chatbot with predefined responses (like this one) can be considered AI. Reducing something to a stupid sentence like “just a bunch of if statements” to try to make it seem absurd is. I can reduce something like ChatGPT the same way and it’d be pretty much as accurate as your take.
You can draw any AI as a predefined flowchart, that’s literally the point, they just make decisions based off of data. Large NLP algorithms like ChatGPT are no exception, they’re just very large involving incomparably heavier mathematics.
Here is a good stackoverflow answer to it that actually gives credible sources (including from the people who pioneered AI themselves): https://stackoverflow.com/a/54793198
AI is very broad. You can use many different definitions of varying specificity to describe AI which can all be correct, even a shitty chatbot counts as AI despite being so basic. There’s no bottom limit for the complexity of AI.
Selecting a canned-text response based on simple keywords is a long way from AI, and it’s foolish to
equivocateequate the two of them.Also, chill tf out, and don’t be so aggressively presumptious. I have enough experience with the topics in question to point out how misleading this statement is.
I suppose you didn’t click the link I sent – either that, or you think you know better than some of the leading figures in the field of AI… it’s not “a long way from AI”, it IS AI in its design and its purpose. It’s misleading to assert that it isn’t AI because it doesn’t meet your arbitrary complexity standard.
I doubt you have any relavant experience in AI research or engineering based off of how you treat the concept of AI and even data science in general here… boiling the bot down to “just a series of if statements” – and then implying that lack of complexity makes it not an AI – is extremely naïve and is itself misleading, you can do that for anything, every program is ultimately just a bunch of if-else/goto and simple math operations. It’s just an attempt to conceptually reduce it so much that it seems absurd that it could be in the same category as more advanced AI. Despite the name, AI doesn’t have to meet some bar for “smartness”, it’s a ridiculously broad term and any program intended to mimic human behaviour falls under AI (no matter how poorly it does it).
You confidently and rudely/condescendingly asserted something that is very blatantly ignorant of the subject of AI, I find it reasonable for me to assume that you had no idea what you were talking about, and I find it reasonable to very plainly call you out.
Also you misused “equivocate”… it’s not a word used to compare two things, it means using double speak/speaking evasively, “to equivocate the two [AI vs. chatbots]” doesn’t mean anything. Did you mean “equate”?
I did click your link. The accepted answer there states:
Again, I don’t think that selecting basic responses based on keywords found in the string meets the criteria for being qualified as an AI, as anyone with experience of a chat bot this simple knows it won’t hold up the illusion of “intelligence” for very long.
I did mean “equate”, you’re correct. The rest of my point remains - a very simple chat-bot like this is leaps and bounds from what would be termed an AI these days. To equate the two is misleading.
The answer you’re referring to (not the accepted answer but the highest voted yes) also says
The ambiguity arises when you ask what it means for “if a human behaves the same way”. If you word it like that then something like ChatGPT or Stable Diffusion wouldn’t count, because you can easily see they’re not human even if you didn’t know first, but then this tic-tac-toe bot would count. It’s a definition they didn’t elaborate on enough so we don’t know what they mean by “intelligent human behaviour”. Maybe “intelligent human behaviour” extends to just giving somewhat relevant answers based on certain words/lexemes in the sentence? Certainly that intelligence is human, I mean a dog or seal can’t do that, only a human. As it stands there is no complex art or chat AI that can’t be distinguished from a human, so if we want to restrict it to actually acting like a human then AI doesn’t exist, unless we’re talking about simple tasks like tic-tac-toe, and there are programs that surpass humans like chess engines which also wouldn’t be considered AI, which I find a silly definition to go by. “Human intelligence” doesn’t mean “as smart as the average human”, it means sentient-like capacity to make decisions, even if it’s extremely simple. The task itself doesn’t change what counts.
That is why I find the take by the pioneers of AI a lot more useful – they don’t put some arbitrary subjective limit on complexity that disqualifues seemingly obvious examples of AI like the IEEE’s ambiguously worded definition does.
What’s in “these days” doesn’t exactly matter – sure, average people nowadays often only use AI to mean complex ML/NLP AI and not the other types of AI, but that doesn’t stop other AI from existing and being AI lol. And especially since people use it the previously common way too still – people who play video games will still call the bots/NPCs “AI”, or call the pathfinding algorithm “pathfinding AI”, for example. And a majority of data science/AI literature will still call simple AI like this one in the post “AI”.
It’s easy to see why asserting your poor definition of AI as the correct one and anything else (even the definition that most professionals in AI agree to, which the comment I sent has a link to multiple with reasons to their credibility over others, one is literally the 4th most cited book in this century) as “misleading” is pretty annoying. You’re trying to gatekeep AI and put your own subjective interperetation of one specific definition on it and ignore multiple leading AI professionals’ definitions lol…
Im not attempting to “gatekeep” anything. I’m pointing out that drawing a parallel between a keyword-based chat it script and a full LLM is disingenuous.
Yeah that video is great. My favourite part is the Russian post address thing.
He has a lot of interesting and funny talks like that.
That is AI…
Even “algorithm”, you could say! The text adventure game I made in BASIC when I was 14 is going to blow your mind. It is 100% artificial and uses logic (IF statements), hence AI!
The problem is their website also implemented an invalid email check when I try to login which is what got me to this point
Here is an alternative Piped link(s):
properly followed by most mail providers
Piped is a privacy-respecting open-source alternative frontend to YouTube.
I’m open-source; check me out at GitHub.