AI AIn’t all that … The Case for Human Intelligence

Large organizations like Google, Amazon, and the NSA, are very interested in using AI (Artificial Intelligence) to manipulate massive amounts of data to their advantage.

However, as of today, AI remains out of reach of individuals. Until we can create AI macros as easily as Word macros, the use of AI at the individual level will be limited.

No doubt that capability will be created by some clever startup.  Until that day, here are some tips on how to use existing automation to your advantage.

This is not the same as AI.  But it does mimic it.  Think of it as human intelligence on steroids.

According to Wikipedia, “Artificial intelligence (AI, also machine intelligence, MI) is intelligence displayed by machines, in contrast with the natural intelligence (NI) displayed by humans and other animals. In computer science AI research is defined as the study of "intelligent agents": any device that perceives its environment and takes actions that maximize its chance of success at some goal.^[1] Colloquially, the term "artificial intelligence" is applied when a machine mimics "cognitive" functions that humans associate with other human minds, such as "learning" and "problem solving".^[2]”

A recent Science magazine article by MIT’s Erik Brynjolfsson et al. details the current state of AI / ML (machine learning), as well as its current limits. One observation is particularly relevant in the context of this post …

“ML systems are very strong at learning empirical associations in data but are less effective when the task requires long chains of reasoning or complex planning that rely on common sense or background knowledge unknown to the computer. Ng's “one-second rule” suggests that ML will do well on video games that require quick reaction and provide instantaneous feedback but less well on games where choosing the optimal action depends on remembering previous events distant in time and on unknown background knowledge about the world (e.g., knowing where in the room a newly introduced item is likely to be found). Exceptions to this are games such as Go and chess, because these nonphysical games can be rapidly simulated with perfect accuracy, so that millions of perfectly self-labeled training examples can be automatically collected. However, in most real-world domains, we lack such perfect simulations.”

The article is rich in insights on AI/ML, and well worth the 10 or 15 minutes it takes to read it.  Here are details …

///////
What can machine learning do? Workforce implications
Science  22 Dec 2017:
Vol. 358, Issue 6370, pp. 1530-1534
source: http://science.sciencemag.org/content/358/6370/1530.full
///////

So now to the point of this post.  While AI is far from reaching its full potential, and very far from useful applications at the individual level, there is one concept that can guide us in the intelligent use of existing tools: augmentation.

By recognizing that the best way to use existing tools is to integrate them into your work practice, you can simulate AI, thereby augmenting your intelligence. Neat, huh? Whereas AI simulates your intelligence, and you will be simulating AI.

Here is a five point action plan …

Create alerts
Browse the results as they come into your email account
Select the results that work for you
Analyze the results
Connect the dots (authors, organizations, topics)

Of course, automation plays a role only in the first step … the Alerts. All the other steps require HI – Human Intelligence. When we are able, as individuals, to automate these steps, AI will have arrived.

To illustrate, here is an example of how I employ the five point strategy.

Step 1: Google® Scholar alert searching for aramco.
This is a simple but, for me, effective search strategy. Because Saudi Aramco researches many of the topics of interest to me, my alert follows Aramco rather than each individual topic.

Step 2: When Aramco results appear in my inbox, I read them. Seems obvious, but you know how unread emails tend to pile up in your inbox.

Step 3: I select the results that interest me. For example, in recent Aramco results I chose the following items …
Leveraging the benefits of ethanol in advanced engine-fuel systems
Recent progress in gasoline surrogate fuels
Organotemplate-free synthesis of hierarchical beta zeolites
Crude oil to chemicals: a cheaper way to petrochemicals?
CFD Guided Gasoline Compression Ignition Engine Calibration
Modeling the Fuel Spray of a High Reactivity Gasoline Under Heavy-Duty Diesel Engine Conditions
Determination of heavy polycyclic aromatic hydrocarbons by non-aqueous reversed phase liquid chromatography: Application and limitation in refining streams
A Computational Investigation of Fuel Chemical and Physical Properties Effects on Gasoline Compression Ignition in a Heavy-Duty Diesel Engine
///////

Step 4: I read and analyze the selected results.
The abstract tells me whether I want to read the whole article.  It also provides keyword clues for future searches.
For example:
In the abstract of the article Leveraging the benefits of ethanol in advanced engine-fuel systems appears this sentence …
“Hydrous ethanol is immiscible in gasoline, and is therefore utilized as a high-octane fuel for the Octane-on-Demand concept.”

Two possible future strategies could be to use the keyword phrases hydrous ethanol, and octane-on-demand in future Google® Scholar alerts.

Step 5: I connect the dots
The author field offers clues to experts that may be worth following. Ditto for the organizations listed in each record.
By connecting the experts and/or the organizations to topics of interest to me, I have clues for further research and/or alerts.
For example:
The articles listed in Step 3 above reveal, among many other potential connections …
Steven Przesmitzki, Strategic Transport Analysis Team, Aramco Research Center – Detroit, Aramco Services Company, MI, USA
S.Mani Sarathy, Aamir Farooq, Gautam T. Kalghatgi, King Abdullah University of Science and Technology, Clean Combustion Research Center, Thuwal, Saudi Arabia
Ke Zhang, Sergio Fernandez, Jeremy T. O’Brien, Tatiana Pilyugina, Sarah Kobaslija, Michele L. Ostraat, Aramco Services Company, Aramco Research Center - Boston, 400 Technology Square, Cambridge, MA 02139
Yuanjiang Pei, Tom Tzanetakis, Yu Zhang, Michael Traver, David J. Cleary
Aramco Services Company, Novi, MI
Saroj K. Panda, Hendrik Muller, Thunayyan A. Al-Qunaysi, Omer R. Koseoglu, Saudi Aramco, Research & Development Center, 31311, Dhahran, Saudi Arabia

This little bit of information suggests all sorts of directions for further research. For example, Aramco produces peer reviewed papers at centers in Detroit and Boston.  Further research reveals that Aramco also has research centers in Houston and France.

Along other lines, you will find that Omer R. Koseoglu is prolific in published patents and peer reviewed literature. If his field of interest matches yours, he would be an expert worth following.

These two examples just scratch the surface. If you know how to automate the task of making these connections, I would love to know how you do so. Most of us, however, must still rely on mere human intelligence to make the necessary connections.

TIP: Until AI matures to full adulthood, if you are in charge of a research lab, and if you see no value in hiring (or not laying off) a librarian, please consider using your librarian to help boost your lab’s effectiveness.