We hear a lot about artificial intelligence, machine learning, and artificial general intelligence today and likely take most of it for granted.

Many of us don’t even bother trying to understand it and simply put the subject somewhere in the realm of digital wizardry. If you are reading this it’s either because you find this topic interesting and want to learn more or you are a professional and want to see how badly I describe your field of expertise. That said, I’ll do my best to appease everyone while still making the following description and analysis a “User-friendly” experience.

Artificial intelligence – artistic concept. Image credit: geralt via Pixabay (free licence)

 

To start, Machine learning is a type of mathematical programming that receives an input, performs a preset logical analysis on that data, and then produces an output. Over time it will improve the way it looks at that data.

There are several classifications of machine learning but for this article we will keep it simple. Artificial Narrow Intelligence (ANI or narrow AI) on the other hand is quite similar to machine learning except that it will consider both the input data and output data sets. In doing so it will attempt to “learn” the hidden patterns of the data so that when new inputs are presented it can efficiently produce high probability outputs. ANI utilizes digital networks of artificial neurons (Neuralnets) that help it develop and understand concepts.

Just like machine learning, ANI also has several classifications (specific techniques/styles) it uses to efficiently find those hidden patterns within the data. Again, we won’t get into that as this article is intended to only provide a framework as to what ANI and AGI are exactly and what it means for us collectively as a global civilization. Since we have not yet developed an Artificial General Intelligence, I will use the term AI to describe all forms of artificial intelligence and will use ANI to describe current and specific applications of it.

he other hand block electrical conductivity. These include silicon dioxide and silicon nitride, which separate the different layers of a chip and protect the transistors from erosion. These parallel industries are also valued at several $100’s billion (USD).

AI runs in the background of most digital services and platforms that we use today. They are the algorithms that track, parse, and make sense of the impossibly large datasets we create. It’s how digital advertisers can predict things that might interest us based on search history, location, gender, and countless other variables. It’s also how that shiny Tesla of yours predicts road conditions and is able to drive itself. AI also helps meteorologists make sense of weather patterns and climatologists in predicting upcoming climate trends.

In the last decade, we have seen AI really take root in the global stock exchanges, forecasting the most optimal trade options at near-instantaneous speeds. You can probably infer as to why many people consider that very risky.

Now that we have the most rudimentary idea of what an ANI is, let’s start expanding upon what we know.

I am sure that those of you who have social `media accounts have seen at least some of the videos of ANI creating original music, art, and industrial design. From the embedded links you will be able to see that even in its infantile simplicity, it has the potential to be “creative”. However, it doesn’t stop at simple novelties like art or music.

More complex ANI have begun moving into the realm of industrial design. I gave a recent presentation to one of the nation’s most prominent industrial design firms about this very topic. The fact that such a high tech organization is concerned about such concepts is a reason to raise caution itself. These are the ANIs that are giving many professionals and forecasters like myself, pause.

AI in industrial design is quickly gaining the ability to accept a series of parameters from a human and through computing hundreds of millions of variations, generate the most optimal outcome to meet the initial criteria. An example of this would be an AI human operator telling the ANI that he/she wants the system to come up with a widget that has a certain structural integrity, weighs less than X, fits within specific margins, and uses the least amount of material. What has traditionally taken teams of designers months, or even years, to complete can now be done at a hundredth of the cost in a fraction of the time.

 

Is there a future for industrial designers based on this trend?

 

Image credit: steamXO via flickr.com, Public Domain.

AI is not only quarantined to industrial design, it has also branched out into web design. The company OpenAI, for whom I was lucky enough to be a demo tester, released its premiere language learning algorithm called GPT3 last year (2020). GPT3 is arguably the most sophisticated learning ANI on the planet and its capabilities cannot be understated.

GPT3 is capable of taking normal English and inferring what the user is requesting and presents the user with what it believes to be the correct response. In one application created by another OpenAI demo user, they were able to have GPT3 make a fully functioning website from a very simple English based request. If you are not a web developer but you speak English, now you can be.

GPT3 is also capable of taking highly complex legal language, just like those 100-page fine print user agreements that you ignore, and rephrasing it into an understandable document. It is able to take legal jargin, condense it, and convert it into normal spoken language. There are some ANIs that now can research legal cases by sifting through tens of thousands of previous court records and applying applicable laws.

Surgical robot. Image credit: U.S. Air Force, Kemberly Groue (Public Domain)

 

Let’s now look at the medical field. Using large datasets from millions of patients, ANI is now able to look at radiology scans (x-ray, CT, MRI, etc.) and have faster, more accurate predictions than the most experienced radiologists. What about surgeons? They too may find themselves at risk of polishing their CV’s because with the advent of ultra-high-precision medical robotics and the digital recording of thousands of surgeries using those devices, it is literally only a matter of time before those robotic platforms (under sophisticated ANI) begin performing faster and safer surgeries.

This is already being done to a very limited degree but given any rate of increase, the result is inevitable. Medical grade sanitation has also started its ANI revolution, where now we see countless robotic sanitation platforms for sale. There is also tremendous progress in ANI assisted medical diagnosis based on patient symptoms. Not sure if this will help or exacerbate the Google induced hypochondria epidemic, where all searches lead to the belief that we have cancer. I suspect as the algorithms get better at symptom analysis, the probability resolution of the presented diagnosis will also increase.

Let’s backtrack a bit and take a closer look at that sexy Tesla electric car we referenced earlier. While Tesla is not the only production car with driver-assist or even full automation, they certainly are the best at it.

Why are they the best? Put simply, it’s because they have the best ANI and the largest dataset to train it with. Every hour and mile driven by a human gets recorded by Tesla’s supercomputers and is used to increase the Autopilot ANI system’s performance. It doesn’t take a futurist or trend analyst to see that this will only get better in time until the point where the ANI will drive better than a human. What does that mean for the tens of millions of truckers, delivery drivers, pilots, and Taxi drivers? Unfortunately, it means they need to begin training for new work as soon as possible.

Image credit: Steven Lilley via Flickr, CC BY-SA 2.0

 

What about AI in media? It so happens that one of the largest driving forces in the world for AI happens to be the video game industry. This is due to the huge market competition, with video game companies always working to make games more realistic and to incorporate more life-like features. In doing so, the computer hardware companies that manufacture and design the Graphics Processing Units (GPU) are constantly pressed to improve their performance so that they may be able to run the games optimally.

Incidentally, most ANI today is run off nearly identical GPUs as they have better frameworks to process the specific type of information that AI uses. Nonetheless, ANI is not just found in video games, it has also made its way into social media and the news media. We have all come across those click-bait stories in our feeds that seem custom-tailored to our interests, and guess what, they were custom made … by an ANI.

There are even algorithms now that can create entire “News” articles. Pair that notion with the recent ANI that is able to replicate visual and audio “Deep Fakes” (those videos of famous people speaking in their own voice but it is actually not them) and it’s nearly assured that in the not-too-distant future most news anchors will be virtual representations of an ANI – perhaps reading ANI generated news content. Scary thought, but completely plausible given that both algorithms are very immature and yet are still fully able to deceive people.

These have all been real-world examples of ANI in use today and they represent only about 1% of all the current applications. There is no way to tell how many ANI are in service, globally, at this moment. If I were asked to make an educated guess, I’d conservatively estimate that there are likely tens of thousands of active ANI at this point. Each one drawing from massive data sets that no human could ever begin to make sense of.

The question we should be asking ourselves now is, “What happens when we start combining ANI?”

To combine ANIs we need to have a type of administrative AI that will integrate them all into one digital “entity”. The goal would be to merge multiple neuralnets into a larger, more efficient, unified digital structure. It will need to collate, organize, and borrow from all of the already well-defined solution sets. This administrative AI will have access to all of those various types of ANI that we covered earlier. In the same fashion an ANI “learns” to perform its function more efficiently, so too would this new type of AI.

Eventually, this AI will become very adept at seeing patterns in all aspects of life. What I am alluding to is what many believe to be an Artificial General Intelligence (AGI). An AGI that has such broad access, with sufficient computational power, would be able to reference all of those ANI and in many regards be able to do it even better. An AGI will outperform all of its predecessors because it will have already learned the best ways to solve most obstacles and will additionally be able to cross-reference with other more mature neural networks.

A true AGI will be the thread, made up of individual ANI fibers, that creates a digital fabric as complex as the human mind.
The problem as it stands now is that an AGI would be easily capable of unifying and automating entire supply chains. From resource procurement to manufacturing, and lastly to distribution. All would be completed autonomously via the control of an AGI. We presently already have most of the robotic hardware platforms and ANI to complete these tasks individually.

The reason why everything remains unautomated is due to the slowness of human implementation and our general unwillingness to do so. However, capitalism is nothing if not adept at finding new ways to save money through efficiency, and in doing so – a fully automated pipeline is pretty much inescapable. How is this a problem? It seems like full automation would be super-efficient?

The answer is that it would in fact be a very efficient use of time, resources, and money. So much so that the human element becomes more of operational inefficiency. Like we just discussed, capitalism has no qualms with alleviating financial liabilities, it’s just that in this case, we are the liability. Our economy, political system, and general moral aptitude are simply not ready for such an extreme shock to the job market.

I am not talking about a couple of million jobs here, I am literally saying – and I want to be clear – that when this happens we will lose entire job markets. As a large Multinational, would you prefer to keep expensive human liabilities on payroll or would you prefer to throw down the capital to purchase an AGI and a fleet of simple network-connected robots to do most of the work at efficiencies that could never be matched by even the most professional humans? What if you run a small business or a Hospital? What about an accounting firm or maybe an engineering company? All would initially benefit from utilizing an AGI.

Are you ready for a world where the title of Best: Musician, Artist, Surgeon, Lawyer, Pilot, Engineer, Designer, Stock Broker, Writer, Analyst, CEO, Janitor, Driver, Scientist, Programmer, Radiologist, and Teacher… will all be held by a single AGI?

AI is currently in an exponential growth phase without any constraints, civil protections, or economic guidelines in place. To conceptualize this effect of exponential growth let’s use the famous penny example. If we take a single penny and doubled it every day, we can generate over $5 million dollars in only 30 days. The problem with this analogy is that this is exactly the speed at which self-improving AI will continue to advance. Just observing OpenAi’s two year parameter growth from GPT2 to GPT3 we can already see the telltale exponential curve.

Examples like this are why we need mindful policy to help mitigate the speed at which AI is being developed. Free enterprise, if left to its own devices, is going to use market competition to drive the speed of AI innovation even faster toward the AGI eventuality. If a workforce is diminished to such a capacity, how will they be able to pay for the goods and services generated by AI and its robots? Questions like this are ones that require serious debate before it’s too late. We are literally gambling on the game of AI, with naivete as our currency, and our future economy as the stakes.

Let us be the ones to sculpt The Age of Automation in our favor. Otherwise, this next chapter in human civilization may not be the utopia we are pretending it will be.

* This article was originally published for Technology.org as a promotional product of Coeus Institute

https://www.technology.org/2021/02/22/ai-the-big-gamble/