Why the GPT / AI hype rule is non enforceable. – Neural Networks – 29 July 2023


I wished to publish my opinion on why GPT / AI / ML / NN bans will not be enforceable by the administration , or they need to be blanket utilized to all merchandise . 

The thread was deleted , happily resulting from a bug of the mql5.com profiles i can return in time and gauge what occurred roughly.

A prelude : 

What was within the thread ? 

The OP (poster) had a few of his merchandise hidden as a result of somebody complained or he spammed them and there appears to be an anti spam bot that sends your product to be checked on the desk should you spam it . Fairly intelligent mechanism i prefer it , bravo mql5.com , step in the suitable path there.

The OP was utilizing gpt to reply questions , his solutions had been obscure , and he was insisting quite a bit on the unsuitable factors (from what i had time to learn) Moreover in his blogs i found customers complaining that a few of his merchandise didn’t function (so he was in violation of a number of guidelines together with the hype AND official laborious coded guidelines) for among the merchandise . 

Nonetheless the dialogue was on GPT . Now was he wrongly insisting he was proper ? Most likely from what i’ve learn to this point . 

May the moderators have eliminated -as i advised in a single publish in there- the fuss the op did and hold the legitimate af factors a few of us made concerning the GPT hype rule ? Sure however  maybe they had been drained and irritated and paired with the OPs timezone they might both lose sleep or delete the thread .

Anyway . I will blur the names and avatars of the mods posts clearly .

On to the opinion :

Firstly of the thread the op claimed the ban was for gpt within the titles (and outline and logos) so the moderator replied with a reference to a publish by the Metaquotes moderator account from the Russian discussion board that stated this : 

So there’s hype talked about there , that means -probably- that the product was flagged for one thing and the test additionally caught the unreal intelligence + gpt phrases within the descriptions . 

right here is the unique hyperlink https://www.mql5.com/ru/discussion board/451234/page2#comment_48339526

So the administration appears to not need AI/ML/GPT/NN for use as “hype” to enhance the perceived efficiency of the product .

(the place is the unreal intelligence hype talked about ? afterward that thread the desk informs the op he used “synthetic intelligence” in his description)

right here is the hyperlink : https://www.mql5.com/ru/discussion board/451234/page3#comment_48343897

However these can’t be arbitrarily enforced .

Right here is why : 

Let’s begin with Neural networks.

  • As an example Mark sweats his hours off and creates a neural web .
  • Locations the neural web within the ea to commerce.
  • Consists of neural web within the title and descriptions based mostly on what he did .
  • Then i waltz available in the market , i add neural web , ai , deep studying within the title or description however my ea does none of that.
  • We add our eas .
  • The one means for mq to know one in all them is an actual neural web ea is to have a look at the supply code . 
  • Now if Mark says maintain on , my ea requires loading this onnx mannequin then mq nonetheless can’t know if that mannequin is behind the wheel of the buying and selling selections with out trying on the supply code . 
  • Mq cannot take a look at the supply code so there isn’t a foundation for banning Mark’s or my EA .
  • They must conceal not one of the 2 EAs or each of the EAs.

Except they seemed on the supply code.

Let’s go to chatGPT / GPT / LLM 

  • Mark could be very good and decides to deploy an LLM mannequin he forked from meta (fb).
  • He buys a server and a website identify and has the mannequin within the server or on a separate laptop with particular replace intervals the place it sends new knowledge to the server . 
  • So , what’s going to Mark’s ea do on this case ? it would ship alongside an emblem and the server will reply with the latest replace it obtained from the LLM mannequin from Mark’s separate PC .
  • Identical if the mannequin is within the server , the EA will ship an emblem and the server will reply with its newest replace for that image , or , if the ea can wait , it would crunch it and compute it proper then and there (not environment friendly nonetheless)
  • If Mark need’s to make use of chatGPT he might want to facilitate communication with openAIs API ,with the token they’ve given him (which he has purchased and actively pays per use) .
  • Now for the reason that cost is per question and per question dimension , and “avatar” dimension if im not mistaken ,Mark can’t danger the queries going on to his server (from the EA).
  • So , he’ll both have a separate PC which does the forwards and backwards with the openAI API and sends or shops the responses at replace intervals on the server .
  • Okay ? One other means may very well be to make use of an onnx mannequin (the smallest is 900 megabytes , greatest 14gigabytes) and have his purchasers load it .
  • Which means we might additionally see necessities for large quantities of RAM for the EA to function on MT5 and virtually unimaginable on MT4 .
  • Now i come alongside , i purchase a server i purchase a website identify and i’ve my ea talk with it .
  • Each Mark and that i publish our eas , each Mark and that i add AI , GPT ,Synthetic Intelligence and all that lingo in title and the overviews .
  • There isn’t a means for mq to inform Mark’s or my EA is pretend .
  • They would wish to see the supply code  , the backend of the server , if there was a 3rd PC managing the replace intervals they’d must see that too .
  • If Marks EA operated with onnx fashions they might not know if the mannequin was wired within the buying and selling selections -again- with out trying on the supply code.

So there isn’t a foundation for hiding Mark’s or my EA in that case both .

Once more its non-enforceable or it have to be blanket enforced.

Once more , they need to conceal not one of the 2 EAs or each of the EAs.

There’s one blind spot which i can’t test due to the laptop being weak right here .

If Mark compiles a mannequin throughout the EA (that means opening up a supply file of minimal 920MB or 14GB in dimension which my laptop cannot do right here) and if that may be compiled , i am unable to test what the dimensions of the ex4/ex5 can be .

However nonetheless even when that was evident by the bloated ex4/ex5 (if it didn’t go over the add file dimension restrict!) there wouldn’t be a option to know the mannequin is wired within the buying and selling selections with out trying on the supply code.

So for all instances the rule can’t be utilized arbitrarily even whether it is as a result of the spam bot caught a product (for different causes) and when it was checked it was discovered to be in violation of the hype rule . 

So , it have to be utilized to ALL merchandise that use this lingo , or , NO merchandise in any respect . That’s what my protest is about .

Sure they will not be checking all merchandise however the rule can’t be utilized arbitrarily if a product occurs to be checked . 

They personal their database they’ll comb by it with scripts and mechanically conceal violations . In the event that they actually wished to go after hype.

“However they are saying they do not test merchandise anyway…” yeah , however resulting from this rule being unimaginable to test  ,as a result of from the second one of many search phrases seems on a product its 50/50 if its true of not ,  it have to be blanket utilized universally + immediately OR not utilized in any respect.

That’s the reason the present strategy shouldn’t be viable and unfair .

If mql5.com wished to be truthful -again as a result of this can be a non-rule , cannot be checked , its obscure anyway , they must : 

  • Not apply this rule to anybody .
  • Apply it to everybody mechanically however present these affected with the means to change their titles .
  • Or change the market completely the place distributors should add supply code as a substitute of ex information.

 This publish shouldn’t be about “how unethical the OP of the thread was for together with gpt in his overviews” , it goes with out saying anybody who does this 

(together with a number of unhidden merchandise on the entrance web page proper now) is unethical . I am not arguing as per the moral spectrum of that particular case.

I will not publish this within the discussion board as a result of it is going to be deleted .

So based mostly on the above , if mq needs to implement it these merchandise are in violation of the hype rule : 

  • Google search this “web site:mql5.com/en/market synthetic intelligence” minus the quotes
  • Google search this “web site:mql5.com/en/market gpt” minus the quotes

Some have been flagged already however handbook checks are meaningless -and a mockery to all the vendors- 

Edit : For sure that if a product is “checked” by the desk manually for whether or not or not it communicates with chat GPT , that means the EA should contact a url to obtain buying and selling directions, it’s in violation of the market rule 6 as that is a further limitation . In order that they move the test for gpt and ingore rule6 in that case….

For those who agree share in your profile :

Related Articles

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Stay Connected

0FansLike
3,912FollowersFollow
0SubscribersSubscribe
- Advertisement -spot_img

Latest Articles