brave souls rewarded today, page-95

I agree, AI/ML is still in its infancy only really understood in tech companies. The overall market is growing, there's no debate in that. There will also be a strong demand for annotated-data. Experts in the AI/ML field will agree that it has always been focused on narrow domain problems. The Turing Test exists, because AI with intelligence is an illusion. The chatbot I mentioned from the 60's (ELIZA) was very basic but managed to convince many people that it was human.

I watched some of the AGM today and the presentation raised all of the challenges that I've previously spoken about. I'll discuss some of my concerns below, as they follow on from the discussion had in this thread.

Competition
In AI/ML there is always a balance between focusing on improving the data-sets and focusing on improving the models. This has existed from the beginning and will always be there. As different techniques mature, the balance shifts but overall they are equally as important. What it means for Appen is that they have to stay relevant by continuously updating the way they annotate data. Otherwise, the customers who need the data will do a large proportion of the pre-labelling (automated work) and leave the more complex parts for Appen.
You will see this happening first in the larger tech companies, that they call their Global segment. The margins likely already show a slow-down in growth. These customers are already bringing back in-house some of the data-annotation. This is where competitors will start to also take some market-share, as they are able to innovate faster in this space. I'm not talking about bringing back the workforce side of data-annotation. Some of the large tech-companies tried this and couldn't do it for the same margins that Appen (and others) work from.

New Markets / New Products
As mentioned previously, the domain knowledge required to successfully implement AI/ML models should never be glazed over. I've lost count of the number of times I've seen a data-scientist blindly apply techniques to a domain (such as language) and have no idea why it is (or isn't) working. The point that I'm making is that, for Appen to branch out into new areas there is going to be a significant cost and steep learning curve (usually softened by acquisitions). This requires significant changes to the types of data-annotation work that they do and often an entirely new workforce.

- Enterprise & Government
There are many reasons why large enterprises and governments are so slow to adapt to change. I'm predicting that these areas will have a significant investment cost and be very slow to mature. It's unlikely that Appen will provide transparency here until they're making profit.

-China
The expansion into China could play out in a number of ways. Their current approach appears to be to grow slowly, don't draw attention. What concerns me is the threat to their IP. They've had to build an entirely new tech-stack, hooked into the CCP, and run their business with local staff who will also be transferring IP (of how to do data-annotation) to a CCP led venture. Globally this isn't an issue, as global tech giants are unlikely to source data from China. However it could eventually push them out of the Chinese domestic market in the long term.

Regulatory Impacts
Regulation has been very slow to progress in AI. I agree that these are not major threats but potential opportunities as Appen expands its products to assist in compliance.
In terms of workforce impacts, this complicates things further as it's a completely separate area. We've seen incredibly slow progress in the ridesharing / food delivery / gig-economy in terms of regulation. However this doesn't mean that this isn't a challenge for Appen. They've already started work to address this, as you mentioned in your link and they mentioned in their presentation.