Have you considered whether Bard or ChatGPT could be integrated into P123's AI/ML?

AI Machine Learning
1

I look forward to P123s implementation of ML/Ai.

I do not want to speculate too much on what thet will look like.

But hyper-variables can be pretty hard for a newbie. For example, anyone including me might reasonably ask: “Should I add dropout to this neural net?” Dropdown menus would be ideal.

Generative AIs are new since you started your AI project.

I will simply say that this newbie now finds it easy to build a neural net with Bard. I would have to confirm this but I suspect "change the batch size:, “add another dropout layer” as well as “please explain back propagation again”, and “do you recommend dropout for this model” would be trivially handled by Bard and you might even get some good answers most of the time.

More speculative: Bard will run programs on outside servers-not limited to Colab. But it can hook into Amazon’s AWS, I believe.

Question to Bard: " If I were a business and wanted you to write some code for me and I wanted to have you connect to Amazon.com’s AWS, run the code and get back to me with the results could that be done?"

Answer: " Yes, I can write code for you and connect to Amazon.com’s AWS, run the code, and get back to you with the results. I can access and use AWS services through the AWS SDK for Python. This SDK provides a Python API for interacting with AWS services.

I can use the AWS SDK to connect to Amazon S3, Amazon RDS, Amazon EC2, and other AWS services. I can also use the SDK to run code on AWS Lambda.

If you would like me to write code for you and connect to Amazon.com’s AWS, please provide me with the following information:

  • The AWS services that you want me to use
  • The code that you want me to write
  • The data that you want me to use
  • The results that you want me to return

I will then write the code, connect to AWS, run the code, and get back to you with the results.

Here are some examples of how I can use AWS services:

  • I can use Amazon S3 to store data.
  • I can use Amazon RDS to create a database.
  • I can use Amazon EC2 to create a virtual machine.
  • I can use AWS Lambda to run code without having to manage servers.

I am still under development, but I am learning new things every day. I am confident that I can write code for you and connect to Amazon.com’s AWS to get the results that you need."

Jim

3

I think they cannot. But they were about to release their AI/ML model and got delayed soon after a discussion of cross-validation It seemed there had not been much consideration of cross-validation but somewhere in P123 or they are contracting with someone who understands what you are doing and discussing in the forum.

I DEFINITELY GET YOUR POINT. A proper grid-search with nested k-fold cross-validation could take an entire night I think.’

Bard and AWS with the AWS charge extended to the P123 member would be one solution for advanced uses and newbies too.

But less ambitious goals e.g., looking for intereactions of a FEW variables with boosting or a neural net and using that as a factor would be great progress!!!

I understand and agree that a fully developed neural-net of 150 factors done right with nested cross-validation would be very resource intensive.

hmmm… more like a month of computer time even on AWS and all those beautiful GPUs.

And maybe we can hope they are listening to you (ASIDE FROM AND EXTENDED DEBATE ABOUT WHETHER THERE ARE METHODS TO MAKE MULTIVARIATE REGRESSIONS USEFUL) and/or are subcontracting with someone who has your level of understanding.

As I understand it you are mainly discussing linear methods because you do not think P123 can do it all with neural nets at this time. At least if you want to cross-validate the hyper-parameters as you so accurately point out.

Jim