Recent content by feralAddict

if you can create your own ML pipeline, jump on huggingface. There are plenty of massive NLP models to use.

I wasn't thinking of Thiel when I wrote it, but now that you mention it, yes: 'competition is for losers' Although I've never actually read any of Thiel's stuff. I listened to a bit of 0 2 1 but it seemed to be more focused on arguing the points instead of pushing them further and exploring them...

So far, I have to say affiliate/referrals are winning. I need to extend this to include converting people from myself and my team's network of contacts. Grinding slack groups is delivering decent results as well Also, getting fairly consistent results is converting freelance contracts And if I'm...

Hi BHW. The forecast today is mild, sunny with a 70% chance of Covid, so stay indoors... And strap yourself in for thread #2 Thread #1 was a very interesting 1k article hack that had me sweating bullets into the keyboard until I stumbled on a lovely little AI hack. The results were well above...

depends on what spinner you use and how much you rewrite. What I found with wordai was that it spun and then some resultant sentences said the opposite of what info was originally written. for example higher ranking keyword became lower ranking keyword. this is a complete deal breaker for me.

yeah I know what you mean. Transfer learning has done a lot for the field.

its been my job to do shit like this for a long time, mate. happy to give you some pointers via DM if you have any specific problems you're looking to solve.

I disagree. a 10x product is still a 10x product even at MVP stage. and Im not suggesting to press on in the dark in futility. I just don't agree with your suggestion of optimization for the win. I would rather invest my energy into game changing tech - LIKE GOOGLE DOES - even as a small pilot...

I have 2 decades worth of experience in ML and AI development. While Linkedin may think all Im good for is being a Dominos pizza driver, I feel I have a little bit more knowledge to share :)

Hi I would love to but Im not allowed to sell on BHW. I will ask the admins if Im allowed to give out other contact details or give some content away for free to people who have been following this thread.

Hey BHW. Just want to thank you all for following this thread. Once I put this algo in production, I will be happy to shout you all an article. :)

ceteris paribus. And that's the flaw in your argument. Let me unbox my thinking a little on this. All things being equal, 1% with daily compounding does add up to a lot. However, all things are not equal; they aren't held constant. Incremental optimization is generally a lot easier to perform...

I've found T5 to be a very flexible all-purpose model. Also, BART is very good for paraphrasing.

Dude I agree the writing is good. However, it's a fact that the information value is low.

I don't mind the quality of writing so much. It is the information value that is questionable, IMO. It's a nice model but it's still weak AI and needs to be primed for downstream results.