Dear Jay Lee,

You highlighted very important questions that are need to be solved via educating them that AI is just only one of the building block for visualization and computing. Similar as non technology degree holders AI degrees holders are numerous, with most of them without sufficient background or domain knowledge (e.g., Mathematics, Statistics, Physics and Computing tools) to its real conceptual understanding. Recently, deep learning which is a tiny not yet fully matured application for training data is boomed and most of the people from every area of science want and wish, how it could works in their area? It is hard to convince flood of minds, not to focus to a single area of research and in many cases without a solid background. Recently, most of the institutions mentioning in their advertised job position e.g.,

"The candidate must have experience are good familiarity with deep learning tools/applications" etc.

Dear Jay Lee,

You highlighted very important questions that are need to be solved via educating them that AI is just only one of the building block for visualization and computing. Similar as non technology degree holders AI degrees holders are numerous, with most of them without sufficient background or domain knowledge (e.g., Mathematics, Statistics, Physics and Computing tools) to its real conceptual understanding. Recently, deep learning which is a tiny not yet fully matured application for training data is boomed and most of the people from every area of science want and wish, how it could works in their area? It is hard to convince flood of minds, not to focus to a single area of research and in many cases without a solid background. Recently, most of the institutions mentioning in their advertised job position e.g.,

"The candidate must have experience are good familiarity with deep learning tools/applications" etc.

Thanks for sharing your insights. Jay

Dear Jay Lee,

Good question!

"Too many nano researchers could not find jobs.", in my opinion, this is a resource allocation issue for those responsible for, and it would apply to many types of research activities not just nanotech.

Overall, the total investment on research and development sector by those who can afford (e.g. corporates, govs and so on) is too low, even in advanced economies, considering the fact that researchers' compensation is also not that high, which would make the competition too hard and too narrow.

Ignoring the big picture of R&D investment, the total investment in AI is not too bad at all.

Let's narrow it down to your question in Industrial Application spectrum:

One might consider educating executive, managers and other decision makers in Industrial domain fields to look into advantages of resource allocation in AI domain in industrial applications. Simply, these two questions can be addressed and presented to them:

- The pros of investment in AI for industrial applications

- THe cons of NOT investing in AI for industrial applications

To be more simple, it is a game of selling and motivating the decision makers to make them believe, while the quality and execution of the idea is very important for the researchers' side, which is not also an easy job.

Best,

J.

- Follow/establish research projects and achieve governmental support.

- Build strong relationship between research and industry, in order to realize joint venture projects with industry.

- Collect collaborations with big names (companies, universities, researchers...).

Thanks for the comments. I think most government agencies are doing that. For example, NSF AI Initiative funds different kings of activities. See https://nsf.gov/cise/ai.jsp Jay

@

For I/U collaboration, an Industry/Univ. Cooperative Research Center on Industrial AI (http://www.iaicenter.com) which is a spin-off Center from NSF I/UCRC on IMS (http://imscenter.net) has been established in 2018.

To accelerate AI for industrial applications using a systematic approach, a new book on "Industrial AI" by Springer can be previewed at the link See link https://www.springer.com/gp/book/9789811521430

I concur that too many researchers believe that they can throw AI algorithms at any problem without understanding the domain knowledge that is necessary to create a viable solution. There is so much hype about what AI is going to do that it is hard to find quality research.

I agree with you Jay Lee . We ever had a discussion about this topic when we met in one of the events we attended.

Without the domain knowledge, it is very difficult to justify in the beginning whether the data collected by a certain type of sensors convey relevant information. We might be able to justify this after the amount of data has reached a certain limit. However, in industry, this can take months, years to prove it. I don't think the industry is willing to invest so much time and money without any guarantee of success.

Thanks for the comments. We often use ABCDE to highlight A means AI, B means Big Data, C means Cloud or Cyber, D means Domain Know how, E means Evidence. The value of AI, Big Data and Cloud is to convert Domain Know How to Evidence.

Industrially funded research targeted to develop product performance (automobile industry for example) has long been the motivation for opening scopes of investigations for research centers. Nano technology (nano-materials and nano-manufacturing) presented new tools for materials design and manufacturing@

I asked the same question to my university professor. He referred me to the "Gartner Hype Cycle". It shows how a new technology becomes too popular at the beginning and slowly takes a saturated trend. AI is at the peak of this hype cycle right now. AI is a blessing in many engineering fields depending on the problem. Likewise, it can be a curse too if not applied with adequate knowledge.

I believe the "intelligence" part of the AI is still on our hands. Humans decide the response/output of an AI framework. The problem is giving priority to AI models over the novel research field. That is: before we design a cat vs dog classifier, we must have a thorough understanding of which ones are dog and which ones are cats. If we invest more time in designing a fancy AI algorithm to classify cats and dogs rather than studying the characteristics of cats and dogs, in most cases it will lead to a disaster. Recalling my previous point, an AI algorithm will learn what we want them to learn.

I think this is a good area to be considered. However, comparing AI with nano research might not be the best because the future of the entire universe depends on AI. We are going towards the smart world and smart everything which cannot be achieved without AI. Also, to conclude change is the most constant factor in the history of the universe. Nothing is wrong if in the future there is a shift in research direction from AI to something else and the AI experts also change their directions to follow the trends.