Finished my workshop on automl + GenAI at #MLWeek last week. I'm loving being able to introduce ideas into prompts for Claude and/or ChatGPT to do things I love doing but rarely see and get code to execute (it wrote over 6000 lines of code for me to use in demos...I created these within a few hours). One example is using ISLE (https://t.co/DpAIrRaOjc) which compresses GBM ensembles: reducing # of trees and tends to be more accurate on hold-out data. Prompt->code->run it.
Appreciate the insights on why LLMs struggle with breaking news.
I wonder how many humans are employed to fix them? This is why I recommend to those who ask me to not use LLMs open loop for mission critical tasks without human validation.
https://t.co/qEW57Bdl3t
@Onanaroghene@pmddomingos If it’s truly A*G*I (emphasis on G) then we would expect broad areas with dramatic progress. If the dramatic progress is in a few areas , that’s great but not “general intelligence”
🧠 Predictive or generative AI? Why not both.
Join Dean Abbott at #MLWeek for a session on hybrid AI for retail fraud detection.
📅 June 3 | ⏰ 10:30 AM
🔗https://t.co/Mjtcxmlnl1
#machinelearning@deanabb
@pmddomingos we have lost about 1 year in value. by comparison, the 2008 crash lost > 10 years of value. It isn't good, of course, but it's too early to tell if it's really a problem or just a knee-jerk reaction that will correct.
Read my latest in @Forbes: When #machinelearning fails to detect misinformation, medical conditions or spam, the cost of each error is subjective. Here’s how to apply #predictiveAI nonetheless. https://t.co/Q4kldkH4Te
Read my latest @Forbes article. Stakeholders involved with #predictiveAI must ramp up on a semi-technical understanding that comes down to 1) what's predicted, 2) how well, and 3) what's done about it. https://t.co/mlVJ8RKJQ7 #bizML
Had an incredible time at @MLWeekUS ! The session on predictive and generative AI by @deanabb were truly insightful. Excited to apply what I've learned and see where this cutting-edge tech takes us next! #MLWeek#AI#MachineLearning