continued on my ahrefs DR reverse engineering. Around 10h investment, old model: correlation: 89%, new model: correlation 69% 😭

making Claude 3.5 Sonnet to come up with a pytorch code which finds a model which computes the ahrefs DR impact of a backlink. With manual exploration I came to a correlation of 66%. 5 minutes of Claude -> pytorch already pushed it up to 72%

played around with models in repalicate to see if I can find an alternate to openAI. results are not promising... yet.

ahrefs DR reverse engineering: With my own efforts I only reached 70% correlation between model and reality. Hired someone on fiverr for $90 - now the correlation is 99% - I still cannot believe it - need some more checks but it looks very promising!

continue experimenting with different models #ai

change fine tuning strategy and data structure and play with new fine tuned model #reva

try new noise model #photoai

research Open AI model fine tuning for reviews tagging #reva

impact prediction model seems crap - refining it now with the help of Claude - and once again I'm blown away

test different AI models with web scraper #ai

play with GPT-4 to compare against fine-tunned model and cry because it does it better but is expensive as hell #reva

test different models #avatarai

updated AI code to parse out structured data into JSON and then run data through pydantic for processing/storing later. I do not hate this model one bit

investigate why some models are failing #photoai

testing AI models

Check out new AI model #danceai

🧑‍🔬 researching when Llama 2 is as good or better than GPT-4 and when it's not as good. Good read here www.anyscale.com/blog/llama-2…

Pruned superfluous MachineLearning models from database #useattractor

Trained a basic flux lora model on replicate… need more fine tuning tbh #nbpai

🚜 added a new model and spent some of my lunch updating my new agent #research