That will indeed be interesting.

The industry is very cliquey.

I talked to two people from the industry and showed them what the software can already do. They agreed that it is as good as what humans do. But both also said they would not become customers. Because they like to mingle with people and support them. One of them said that is also the reason they don't try to outsource the work to lower wage countries.

But I only need 0.01% of the industry as customers to have a great lifestyle business. So I will plow ahead and see what happens.

Plus, with a 10x decrease in costs, the industry might actually grow. There might be new customers coming in that previously were not interested due to price.

Think robotaxis: Not only would a robotaxi service get existing customers of old school taxis and Uber. They could offer a lower price and have people take a robotaxi who would otherwise have walked or stayed at home.

Can you not provide your service directly to the consumer?

No, it's a B2B thing. Consumers don't pay for it.

Think robopilot software. It needs to be put into a plane before consumers can use it.

Can you then not provide it as a service to businesses, just slightly below the usual market price? Selling the fish instead of teaching then your new way of fishing.

That is the plan. The "BUY" button would not be for buying the software, but for buying the service that they usually hire people for.

It is not possible to "hide" the fact that it is done by software. As there are some differences. The software is stronger in some aspects of the tasks and weaker in other aspects. Overall, the software is superior. But it is noticeably different to work with the software than with a human.

Think Robotaxis: Superhuman reaction time. Never sleepy. But bad at interpreting unusual situations. Overall, they will be 10x safer than human drivers at some point. But you will still notice if a human or a bot drives the car.

In the realm of NLP, there are some companies that offer a human in the loop service. So to the outside, they offer human quality results. And on the inside, they use their machine learning technology, supervised by humans. One way to do it, I guess, is to especially watch the results where the model is least confident. And of course, you feed back all of these corrections back into the model, so it becomes better over time.

And those objections you refer to, I am quite sure that people like journalists would’ve had the same objections to machine learning transcription services just a few years back. Now, of course, they all use otter.ai and the like and cannot imagine ever going back to slow human transcriptionists.

Yes, good example.