3 ms·
If you have an idea for a cool AI startup it's faster to build your first prototype without the actual AI, just faking that part. But if your Actual Indians had
by wongarsu 1y ago
If you have an idea for a cool AI startup it's faster to build your first prototype without the actual AI, just faking that part. But if your Actual Indians had 95% accuracy and you can't get an AI to do more than 85% then you are kind of stuck if you raised money and got customers pretending that your Actual Indians are Artificial Intelligence.
- TYPE_FASTER 1y agoThis is the way. Funny how AI could also stand for Actual Intelligence. Or, Artisanal Intelligence? "Now 100% organic handcrafted thoughts, unique for your business problem."
- cubano 1y ago[flagged]
- more_corn 1y agoNot true it’s super easy to fine tune and deploy one of the open models. I should teach a course.
- msgodel 1y agoThe technical aspects of training and tuning are trivial. GP is pointing out that you might not be able to get the model to succeed at the task as often for any number of reasons that you won't know before you actually train one. Although I guess your point is that it's also cheap to train them, probably cheaper than doing this. But startups are started by social people, not technical people. Stuff like this will always be expensive for social people since they have to pay one of us to do it. YC interviews their CEOs from time to time, it's really clear that's how that works.