AI services journey.

[rimbouaida]

I'm a professional Data scientist new into internet money , what pushed me to it , is the fact that the clients I've had made thousands if not millions with AI I've skulled myself , so I'm now in my own journey into AI servicing and projects.
One of my last projects was a realtime gambling prediction with a 78% success rate , and a virtual girlfriend AI with voicing and sexting possibility based on the context , it also remembers everything related to the user she's chatting with.

1- I'll choose a good project in the comments below from the BHW users ( I'll work on it for free )
2- Share it with you.
3- Brain storming steps on how to work on it.
4- Technical steps attack.
5- V1 share.
6- BHW users feedback.
7- Share the solution publicly.

 

[Inem2]

I'm now in my own journey into AI servicing and projects.
One of my last projects was a realtime gambling prediction with a 78% success rate ,

been looking to develop something specifically for sport betting. can we work on that ?

 

[rimbouaida]

been looking to develop something specifically for sport betting. can we work on that ?

I've already have that solution , I want to try something new , based on what BHW users would propose

 

[DropVS]

for this "gambling prediction with a 78% success rate" what is he average odd?

 

[Anson2008]

Good luck mate I will keep an eye on this

 

[GreenFog]

Had the same idea but am thinking more in the B2B space.

One of my last projects was a realtime gambling prediction with a 78% success rate

Anyone with a minimal knowledge in ML knows this means nothing (eg a game which has 78% losing probability and a model which predicts all games to end in loss will have an accuracy of 78% )

Bruh I noticed you're the quora bot guy, we talked a long time ago and I was thinking on doing the same thinkg as you before aswell

 

[bobblack225]

Interested in the ai gf. I'm not a data science guy but I am a python programmer in general. Been interested in this field. Would it be possible for you to share how you managed to pull that one off? Just trying to learn, thanks!

 

[jonathan101]

Wishing you the best mate

 

[GreenFog]

Interested in the ai gf. I'm not a data science guy but I am a python programmer in general. Been interested in this field. Would it be possible for you to share how you managed to pull that one off? Just trying to learn, thanks!

Absolutely Haram.
Anyway, very likely gpt as backend and an additional layer of text 2 speech/speech 2 text on top.
As for "memory" there are multiple ways to implement it, whether as prompts or embeddings.
No deep AI/ML knowledge needed to pull it off but definitely must lack a lot of decency.

Most of the recent AI products are just caveman level wrappers around openly available (whether open source or as a paid API) AI services/models.

 

[twohkz]

Very intresting... maybe something in blockchain

Maybe detecting new project / contract in solidity or other language launching in a blockchain

live new pair
https://www.dextools.io/app/en/ether/pool-explorer
we will try to detect honeypot ( scam ) contract, invest in the safiest one before Liquidity added and sell after it will be added

we can get a good amount $$$

major honeypot/ scam detector can not analyse contract with a proxies attached to it,
all the contract fr example got a % fees in buying / selling so a first they got like 5% buy / 5% sell
and then scammer turn it sell 100% so no one can sell beside them
or they blacklist all adresses beside them, so they all them can sell

 

[rimbouaida]

for this "gambling prediction with a 78% success rate" what is he average odd?

Max one was 3.5 , but there were some users that took some own risks , and added some other parlays.

 

[rimbouaida]

Absolutely Haram.
Anyway, very likely gpt as backend and an additional layer of text 2 speech/speech 2 text on top.
As for "memory" there are multiple ways to implement it, whether as prompts or embeddings.
No deep AI/ML knowledge needed to pull it off but definitely must lack a lot of decency.

Most of the recent AI products are just caveman level wrappers around openly available (whether open source or as a paid API) AI services/models.

In our case we used some ML data augmentation algorithms , of one of our sexting datasets since there's a lack of data in this particular field , but overall we're not using any API except for the platform we're targeting for chatting or sexting of them was OnlyFans so I had to reverse engineering their network requests and use them as a So called reverse engineering API.
For the AI model it's a trained model composed of CNN RNN and a known DistilBERT model.

 

[rimbouaida]

Very intresting... maybe something in blockchain

Maybe detecting new project / contract in solidity or other language launching in a blockchain

live new pair
https://www.dextools.io/app/en/ether/pool-explorer
we will try to detect honeypot ( scam ) contract, invest in the safiest one before Liquidity added and sell after it will be added

we can get a good amount $$$

major honeypot/ scam detector can not analyse contract with a proxies attached to it,
all the contract fr example got a % fees in buying / selling so a first they got like 5% buy / 5% sell
and then scammer turn it sell 100% so no one can sell beside them
or they blacklist all adresses beside them, so they all them can sell

First idea , this one looks good so far

 

[GreenFog]

In our case we used some ML data augmentation algorithms , of one of our sexting datasets since there's a lack of data in this particular field , but overall we're not using any API except for the platform we're targeting for chatting or sexting of them was OnlyFans so I had to reverse engineering their network requests and use them as a So called reverse engineering API.
For the AI model it's a trained model composed of CNN RNN and a known DistilBERT model.

CNN? Did you also train a nudes classification model?

 

[rimbouaida]

CNN? Did you also train a nudes classification model?

I only implemented voicing sexting and random context nuding , how I did for the nudes is for example in the Onlyfans use case , I've scrapped the content , did a manual labelisation ( thanks to an Indian dude ) for example a video with the model shaking her booty main label booty/ass ,etc , I've trained the model on those labels did some train test split tests , the model performed pretty well since Image recognition in our modern age is easier to train than text since you need zillions lines so that your model is perfect.

 

[hkey]

I was curious about your gambling prediction app, because it is impossible to make tactics in gambling other than blackjack. If the card table is not rigged...

 

[rimbouaida]

I was curious about your gambling prediction app, because it is impossible to make tactics in gambling other than blackjack. If the card table is not rigged...

Well we all know that football gambling is unpredictable you can't predict a red card or a last minute goal or a weird corner lucky kick , but what I've done is basing my model on real time metrics and twitter sentiment , Head to heads , If a key player is missing , the weather , home or away , the referee if older games were won with that referee , real time news or - 1 day news , usually twitter sentiment is pretty good.
The model would give you the best tip , a secondary low risk one with low odds like 1.7 , and how many corners + or -

 

[bobblack225]

Absolutely Haram.
Anyway, very likely gpt as backend and an additional layer of text 2 speech/speech 2 text on top.
As for "memory" there are multiple ways to implement it, whether as prompts or embeddings.
No deep AI/ML knowledge needed to pull it off but definitely must lack a lot of decency.

Most of the recent AI products are just caveman level wrappers around openly available (whether open source or as a paid API) AI services/models.

I could care less about "haram" or not. Let's keep religion out of this, and for the record I'm not Muslim. As for the api, that'd be stupid. You'd have a very low success rate unless you took it and trained it. Plus you want to keep costs low, so I wouldn't recommend taking a 100b param open source llm and training it and using it for inference.

 

[bobblack225]

In our case we used some ML data augmentation algorithms , of one of our sexting datasets since there's a lack of data in this particular field , but overall we're not using any API except for the platform we're targeting for chatting or sexting of them was OnlyFans so I had to reverse engineering their network requests and use them as a So called reverse engineering API.
For the AI model it's a trained model composed of CNN RNN and a known DistilBERT model.

Thanks for the info. Reverse engineering the api isn't hard at all. But what I'm more interested in is the dataset and how you brought together all the components in the ai.

 

[Black Belt Spodermen]

Dude you will do great if you can develop an AI tool for some kind of internet marketing. You can then open a sales thread here and enjoy good money.