Ganes Kesari的动态

This will help me

You have out across important points in a succinct manner. Thank you for the bite-size wisdom. Some issues we face are 1) Expected accuracy set at 90-100% without any rationale for it 2) Those who claim current business process is at x% accuracy often do so from a gut feel rather than actual measurement 3) How the accuracy the AI solution claims be tested by client is often left in the dark till the solution is ready to be deployed. This is a problem especially for process where we don't have a good way of measuring. I think a lot of thought needs to go into project success metrics before we begin the work.

Amazing!!

Very interesting Ganes Kesari.....to me, optimal accuracy can be achieved by Human-AI collaboration. The important question is how do you assign weights to algorithms vs humans...thoughts?

Very relevant - thanks

Great talk. I remember the time when we had to justify why a model was giving only 95% and not 99 % ! For one particular use case which was a very subjective evaluation, we did a simple test by asking 4 humans to do the same and measured the error rate. That helped us prove the subjectivity in the process

Ganes Kesari amazingly simple and crisp = how the decision making is simpler when backed by data = 80% vs 85% Is it okay, if I take the same video and build mine on similar concept but answer a larger question. ????

These are great pointers Ganes. Just to share a few thoughts on the ROI -> Of course it varies across organization in how they look at investment and collective impact but businesses need to take a multi-dimension view on the economics. Infusing intelligence at times is notional and may have a positive impact on the overall Brand Equity, which might be hard for one to equate a $ value Ex: Perception. I recall so many instances where an AI solution gets tagged as a Pilot/Experiment coz of accuracy not meeting expectation. But your 3rd point on 'Human in the loop' is a great way to complement human ingenuity. Loved this video.

It raises the question of the client setting the quality of your product perhaps ?

Amazing ! The only humble concern I have is , AI ML as a solution when you try to sell to business the ROI usually comes from reducing workforce as the AI will do the work for you . At this point bringing in the Human in loop concept , will it not dilute the objective we’re trying to achieve ? When Inspite of the ML solution we’re expecting human intervention?