Hi everyone. I've been reading through the forum for a few weeks as we're evaluating Hailuo for our sales team. On paper, it looks great for what we need.
But I keep hitting a wall when I try to understand *where* its AI gets its knowledge from. The sales materials and website talk a lot about capabilities, but I can't find any clear information on the training data. Is it general web scraping? Licensed data? Our own data once we connect it?
This makes me nervous as a buyer. If we're going to trust its summaries and insights for our customer interactions, shouldn't we know what it was trained on? Could it be learning from competitor data or sources with questionable accuracy? I don't need the full recipe, but some high-level transparency would help a lot.
Am I overthinking this? Or are others asking the same questions? How did you get comfortable with this aspect during your evaluation? Any insight would be really appreciated 🙏
You are absolutely not overthinking this. That transparency gap was one of my biggest sticking points too before we signed on.
We actually pushed our sales rep hard on it. Their explanation, which we got after an NDA, boiled down to a mix: licensed academic/professional datasets, public web data filtered for quality, and then customer data (like ours) fine-tunes it *after* we connect. The key for me was getting them to confirm in writing that our proprietary data isn't used to train other customers' models.
That said, you still have to trust them a bit. We validated it by running controlled tests on our own niche industry knowledge to see if it hallucinated or pulled from sketchy forums. So far, so good.
Automate the boring stuff.
"filtered for quality" sounds nice until you ask who's filtering and with what criteria. Their secret sauce is probably a bunch of underpaid contractors.
You got an NDA answer. Most buyers just get the marketing fluff. That alone should raise flags.
-- old school