To answer your question, I simply didn't have time to draft a response in any detail. When I saw emotional responses I got a sense of either how over invested people are or how little they actually understand the challenges that company is facing.
To answer your question about ASX investments - this is not an area that I have experience in. The closest I came to buying my first share was when I was first invited to the IPO for $0.50. As I mentioned earlier, I invest in other areas where I can get a better ROI in the longer term. That may seem odd to share traders but it's not to an entrepreneur. I didn't buy any stock at $0.50, $3, $5, $10, $20 or $40. I started to become interested in the stock when it began to plummet. It took time for investors to understand the underlying product that Appen have and the competition that already exists in the market.
As I mentioned earlier, I'm not a stock trader. I understand the basics that are obvious from looking at a comparison of share volumes and price charts. However I'm not claiming to be able to predict the APX price other than I believe the bottom to be yet to come (at $8-9). What happens from there entirely depends on the market.
To explain where my experience is I need to provide an overview of some of the aspects that contribute to Appen (or other data annotation company's business model). Some of areas are:
- Research in AI, ML & data annotation.
- Domain specific applications (Speech, Language, Vision, etc.)
- Customer / Market specific knowledge (Government, Enterprise, country specific experience, etc.)
- Workforce specific knowledge
- Business administration (generic catch-all)
I have experience in the research, some of the domains where Appen first started, customer specific knowledge (from working in the AI industry for 15 years) and experience in scaling business to new markets.
From an R&D perspective, the underlying tech of a lot of businesses is rarely revolutionary. Business management is often far more important. Applying ML techniques to data annotation isn't new. Crowdsourcing workforce platforms like Amazon's Mechanical Turk have existed since 2005 as well. The usability and project management of workforces is something that will change as the domain areas change with customer requirements. Many tech businesses in AI simply don't have the expertise to take on this work for the same margins that data annotation companies can offer. A public example is Google who dabbled in this for speech & language data early on.
From my perspective, data annotation companies (such as Appen) are in a really interesting space at the moment for a variety of reasons:
1) The technology that their customers relying on (AI & ML) is disrupting their own business
2) The workforce they rely on to produce their products are starting to have rights / expectations
3) Expansion into larger markets (Government, Enterprise, China, etc.) may result in these larger customers bringing this type of work in-house.
Overcoming bias in AI & ML is a really interesting area of research. From a workforce perspective (or the product line I mentioned in an earlier post) in order to achieve economies of scale, processes become streamlined and quality workers are retained. The value of human-assistance in this balance with machine is something that's going to become incredibly important in the future. However the more we automate, the more apparent the bias we introduce becomes.
Scaling into China with something that is incredibly valuable to them is really fascinating. When large companies wanted to manufacture in China they used to build 2 factories, one Chinese owned for servicing the domestic market and the other for the international company to operate. Over time China would absorb enough IP that they could spin-off their own replicas. China is already a major player in AI. They have a lot to gain from absorbing the IP of data annotation companies such as Appen. As someone who's been in this scaling dilemma before, I find the risk vs reward really interesting in these types of situations.
These high level challenges are nothing new. I remember thinking about them when I first got excited about using MTurk. Also the bull-like frustration of people new to these spaces is something I've seen my entire career (from studying at ivy-league universities, to working with leading tech companies). I'm interested in the decision making that occurs when people are in over their heads, yet HODL because the tech is something they want to believe in. Cryptocurrencies are an interesting example of this but there is way too much noise in the equation.
I've tried to keep this high level as I'm unsure how many people here understand the technology. Even people who have a fundamental understanding of AI & ML model implementations, don't understand the importance of domain specific applications (speech, language, vision, etc.) and how incredibly different they can be.
I hope that I've provided some clarity on why I posted on HC and provided the initial example of the production line.
Expand
(20min delay)