Mark: Yes inference is much less intensive on the compute side than training. But there is a reason some of the industry leaders are spending billions on data centers. It is because they consider this a technology that will rule the future. The final products may not require intensive compute, but the training certainly does.
The other aspect is updates, so I want to roll out updates continually, and this requires re-training some portions, again as you are pointing out some things do not need to be re-trained continually, they can lock things down and only train certain areas requiring magnitudes less compute, but no company can afford to bring a spoon to a knife fight, there are too many unknowns, and in the end I would bet the house on the fact that compute will be critical. Being first matters, being better than your competition matters.
I am speaking generally about the technology, I am not speaking specifically about a specific path or plan that Epicor has to leverage this technology, only about the potential, and the very real angle that this is justified. Epicor will need to lay out that and provide specifics. Yes big players like Microsoft have their own AI Cloud services and their own paths and plans. Peace.
Your Competitor will be in business another 50 years, you might be filing for Bankruptcy next year. If your costs go up, the price list of our product goes up â which allows the Competitor to win, some of us run on very competitive margins. In Automotive even a penny can get you declined on a General Motors Quote. If your product is shipped bad you now are on their sh*t list and no longer eligible for class A Quotes.
