Since graduating with a degree in biology, Lisa Magloff has worked in many countries. Accordingly, she specializes in writing about science and travel and has written for publications as diverse as the "Snowmass Sun" and "Caterer Middle East." With numerous published books and newspaper and magazine articles to her credit, Magloff has an eclectic knowledge of everything from cooking to nuclear reactor maintenance.

Enterprise IT architect certifications appear most often at the apex of certification programs, where less than 1% of IT professionals ultimately ascend. Even so, many IT architect certifications are available, and you don’t need to invest in one certification sponsor’s vision to reach the top.

Many IT certifications in this area fall outside vendor umbrellas, which means they are vendor-neutral or vendor-agnostic. Nevertheless, the number of vendor-specific IT certifications exceeds vendor-neutral ones by a factor of more than 2 to 1. That’s why we devote the last section of this article to all such credentials, as we encountered them in search of the best enterprise architect certifications.

For IT pros who’ve already invested in vendor-specific certification programs, credentials at the architect level may indeed be worth pursuing. Enterprise architects are among the highest-paid employees and consultants in the tech industry.

Today at the Open Hardware Summit in Portland, Alicia Gibb and Michael Weinberg of the Open Source Hardware Association (OSHWA) launched the Open Source Hardware Certification program. It’s live, and you can certify your own hardware as Open Hardware right now.

What Is Open Source Hardware?

Open Source Hardware can’t be defined without first discussing open source software. At its very core, open source software is just a copyright hack, enabled by a worldwide universal computer network. The rise of open source software is tied to the increasing ease of distributing said software, either through BBSes, Usenet, and the web. Likewise, Open Source Hardware is tied to the ease of distributing, modifying, and building hardware.

In the 1980s, there were no services that could deliver a custom circuit board to anywhere on the planet for a dollar per square inch. When open software began, CNC machines were expensive tools, now you can build a very good machine for just a week’s wages. We are currently living at the dawn of Open Source Hardware, enabled by the creation of Open Source design tools that have themselves been used to create physical tools. Inexpensive 3D printers, open source oscilloscopes, circuit board plotters, and the entire hackerspace movement are as revolutionary as the Internet. These devices and the Internet are the foundations for Open Hardware and software, respectively.  The objections to why hardware is incompatible with Open Source no longer apply and small-scale manufacturing techniques are only going to get better.

Open source is a moral imperative in the truest Kantian sense of the word. It is a good unto itself. Of course, this means open source is also mind-numbingly prescriptivist. Holy scrolls have defined dozens of different open source licenses. The relevant license for Open Source Hardware has already been laid out to define the freedoms and responsibilities of all Open Source Hardware creators. Open Source Hardware is a tangible thing, from a laptop to a lampshade, whose design is available so anyone can make, modify, distribute, and sell that thing. Native documentation is required, and software required to run this thing must be based on an OSI-approved license.

The definition of Open Source Hardware has been around for a few years now, and since then the community has flourished, there’s a great gear logo, and you can buy real, functional hardware that bills itself as Open Source Hardware. It’s become a selling point, and this has become a problem.

Many hardware creators don’t adhere to the definition of Open Source Hardware. In some cases, the design files simply aren’t available. If they are, they could be unmodifiable. The software used to create these design files could cost thousands of dollars per seat. This is the problem the movement faces — Open Source Hardware must have a certification program. Unlike open source software, where the source is almost proof enough that a piece of software complies with an open source license, hardware does not have such obvious assurances.

Software Is Closed By Default, Hardware Is Open And The Licenses Are Harder

All software is closed by default. Anything written is covered by copyright, and the developers of open source software choose to license their works under an open source license. Open source software, then, is a copyright hack, enabled because all software is closed by default.

Hardware, on the other hand, is open by default. If you build a device to automatically inject epinephrine intramuscularly, you must go out of your way to patent your device. Only a patent will provide you the ability to license your work, and before that patent is published anyone can make their own epinephrine pen. If you build something with an FPGA, the code that programs the FPGA is covered by copyright, but an arbitrary circuit that uses that FPGA isn’t. Any generic piece of Open Source Hardware could be covered under patents, trademarks, and a dozen licenses. Therefore, an Open Source Hardware license is impractical. This is why OSHWA is not releasing an Open Source Hardware license, and instead creating an Open Source Hardware certification program. No Open Source Hardware license could cover every edge case, and a certification is ultimately the only solution.

The Open Source Hardware Certification Program

At last year’s Open Hardware Summit, OSHWA formally announced the creation of the Open Hardware Certification program. Now, this program is live, and the certification database will growing very, very quickly. At its heart, the Open Source Hardware Certification program is pretty simple — create hardware that complies with the community definition of Open Source Hardware.

The theoretical basis for the need of an Open Source Hardware license is the fact that anyone is able to manufacture hardware. Of course, there are limits to technology and no one has a 14nm silicon fab line in their garage. This is a problem for any piece of Open Source Hardware, and the technical capability for anyone to recreate integrated circuits and other high technologies is the sole source of the traditional objections to any open hardware license. Garage-based fabrication is always improving, though, but closed hardware in the form of NDA’d chips will remain a problem for years to come.

The clearest example of the problem with closed-source chips is bunnie’s Novena laptop. This laptop is designed as both a hacker’s laptop and an artifact of Open Hardware. Although most of the chips used in the Novena are available without signing NDAs, open source, and blob-free 3D graphics acceleration was unavailable when the laptop launched. This non-open graphics problem will be fixed with open source drivers, but it does illustrate the problem of Open Source Hardware. Even though chips might be available, there might be binary blobs required for full functionality. You can build an Open Hardware chip in VHDL, but it’s not really open if you have to use closed-source FPGA dev tools.

OSHWA’s solution to this problem is simply asking for hardware creators to act in good faith. The certification program won’t knock points off for using closed source binary blobs if that’s the only way of doing something. Open Source Hardware is just slightly more aware of the pace of technical progress, and what is closed today may be open tomorrow. Building a piece of Open Source Hardware isn’t an all or nothing proposal; just provide your best effort to make it open, and technology or reverse engineers will probably make it more open in the future.

Of course, with any certification program, there must be some effort given to enforcement. If an Open Hardware project is certified under the program but does not meet the guidelines of the certification program, fines may be levied against the project creators. Again, good faith of the project creator is assumed, and a project found not in compliance with the certification program will be given 90 days to either fix the problem or remove the project from the certification program. After 90 days, there’s a 120-day period of public shaming, and after that small fines of $500 per month. The worst offender will get a fine of up to $10,000 per month, but that would require years of non-compliance, and it’s very doubtful any conflict with OSHWA will ever reach that stage. It should be noted these fines have a legal basis in the trademark of the OSHW certification logo, and if you don’t use the OSHW logo or certify your project, there’s nothing OSHWA can do.

The old Open Source Hardware ‘gear’ logo — unquestionably a better logo — will still remain in use, and no one is going will look down on you for using it. Using the trademarked OSHW logo, though, is the only way any certification program can be enforced.

The Objections To Open Source Hardware

Of course, the Open Source Hardware Certification program has been more than two years in the making, and that’s time enough for a few people to start having very strong opinions about it. A few years ago, Saar Drimer of Boldport said he won’t be using the Open Source Hardware logo on his boards. This is despite the fact that he loves Open Source Hardware, has written open source PCB design software, and offers a 20% discount on open source contract work. His reason is simple: adding a logo brings baggage, and building Open Source Hardware is not mutually exclusive with putting a logo on a board. Dave Jones is a big supporter of Open Hardware, but he realizes the famous gear logo is becoming meaningless through abuse.

You need only look back on the last twenty or thirty years of the world of Open Source Software to get a sense of where Saar and Dave are coming from; Stallman does not believe in a moral imperative to Open Hardware, whereas most everyone in attendance of today’s Open Hardware Summit does. Gnome versus KDE is nothing compared to the religious war we potentially face between various Open Hardware philosophies. The Open Source Hardware community is relearning what the open source software community learned twenty years ago. We can only hope to learn from their missteps.

But Open Source Hardware has a much bigger obstacle to adoption than politicking and empire building. Open source software is a simple concept — you have a (copy) right to whatever software, music, words, or boat hull designs you create. You can, therefore, provide others the right to use, study, share, and modify that work. Physical objects and artifacts do not have copyright, they have patents. Patent law in the United States is atrocious, and just because you were the first to create a useful invention doesn’t mean a patent would be invalidated. This is the greatest challenge to anything developed as Open Source Hardware. The only solution to this is prior art and patent inspectors that know where to look.

This Will Take a While to Work Out

The Open Source Hardware Certification program is going to take a while to unravel. OSHWA doesn’t believe this certification program will be a repository used by patent inspectors looking for prior art. The legal basis for the certification is literally built upon every piece of intellectual property law. It is, perhaps, an answer to the most complex legal questions ever: what is property, what is intellectual property and can the concept of physical things be given away.

No one has an answer to these questions, or at least an answer that can be summed up in one-page FAQ. The Open Source Hardware Certification program is an attempt to answer these questions, and so far it’s the best attempt yet.

None of this matters unless the community gets behind it, and if another competing Open Source Hardware certification or license pops up, the community may very well migrate to that. Judging from the last thirty years of open source software license drama, we can only hope that the community figures this out the first time, and we hope this certification program is a rousing success.

The global digital ad market is expected to hit more than $209 billion by 2027. In fact, this rapid growth has meant that market research analysts and marketing specialists are among the country’s top 20 most sought-after employees, commanding an average salary just shy of $64,000, according to the Bureau of Labor Statistics.

In this article, we explain what digital marketers do and list the top digital marketing training providers whose courses may provide your career with a boost.

What first began as an agricultural research project at North Carolina State University eventually grew into a full-fledged software and services company by 1976. SAS has gone on to develop a solid customer base in the banking and pharmaceutical industries as well as in academia and at numerous agencies at all levels of government. Today, SAS is a leader in business analytics, data warehousing and data mining.

SAS has been recognized as one of the best places to work by organizations like Fortune and the Great Place to Work Institute. Coming in at No. 37, SAS made its 21st appearance in Fortune’s list of 100 best companies to work for in 2017 and No. 23 in Fortune’s list of top 100 best workplaces for millennials in 2018. Indeed, the company’s low turnover rate (4 percent in 2016) is an indicator of its commitment to its employees and, indirectly, to its customers as well.

Of the top 100 Fortune Global 500 companies, 96 are SAS customers. SAS customers span the globe with more than 83,000 instances installed in 144 countries.

It’s true that the post-pandemic workforce has probably changed forever. According to accurate stats, 35% of the current workforce is still working remotely full-time and a significant chunk of employees are working a hybrid schedule. As we know, this is presenting all sorts of challenges to HR leaders who are working hard to engage an increasingly disjointed workforce.

However, one place where employees are open to more in-person interaction is at the gym — specifically, the corporate gym. At HealthFitness, we’ve seen an uptick in the number of our clients’ employees actively using personal and group training since the pandemic starting waning. And — here’s the surprising part — most of that is happening in-person at corporate fitness centers across the country. Eighty-six percent, to be exact (on the flip side just 5% of those personal and group training sessions are held virtual-only).

What’s more, employees are participating in an increasing number of these personal or group training sessions. Seventy-seven percent of our corporate client employees are signing up for 4 to 8 personal or group training sessions per month! That’s a big commitment — but employees are making it. And again, they’re largely doing it in-person.

The New Jersey Performing Arts Center (NJPAC), a non-profit multidisciplinary performing arts center and one of the nation's largest arts education providers, announces new fall classes at the Colton Institute for Research and Training in the Arts.

The Colton Institute for Training and Research in the Arts supports programs championing the whole child. Arts training, mentorship from working artists, career counseling, and a talent program that pays NJPAC Arts Education students and alums for performances across New Jersey are some of the initiatives advanced by the $10 million catalytic gift from Judy and Stewart Colton.

The 2023/24 schedule offers various classes on Saturdays ranging from Beginner Band, Hip Hop Arts & Culture, Acting, Musical Theater, and, new this year, Music Composition and Arranging. This class welcomes aspiring composers ages 14-19 who demonstrate promise and dedication early in their creative development. The program focuses on the fundamentals of composition through instruction, group activities and styles, the performance of students' works, and interaction with the Teaching Artist, Saxophonist, and Composer Lance Bryant. It prepares students for the next stage of their artistic journey just for $100 each semester. Students registered in TD Jazz for Teens are part of this class at no charge.

The Colton Institute increased NJPAC's arts education offerings. It advances its services for students — many of whom come from economically disadvantaged circumstances — including mentorship and field training, ultimately creating a pathway for college and career opportunities in the performing arts, whether onstage or behind the scenes working in entertainment in a corporate setting. To Register for a class, visit Click Here or send an email to artseducation@njpac.org
 
 

Certainty and uncertainty play a big role in life.

It is said that the only true certainty consists of deaths and taxes. Michael Crichton, the famous writer, said that he was certain there is too much certainty in the world. Legendary poet Robert Burns indicated that there is no such uncertainty as a sure thing.

One thing about both certainty and uncertainty is that we seem to crave and relish certainty, while we tend to agonize over and strive to convert uncertainty into certainty if we can do so. As Carl von Clausewitz, the lauded military strategist professed: “Although our intellect always longs for clarity and certainty, our nature often finds uncertainty fascinating.”

All of these machinations over certainty and uncertainty turn out to be a big matter for those that wish to fruitfully make use of today’s generative AI such as ChatGPT, Bard, and so on.

In today’s column, I am continuing my ongoing special series about advances in prompt engineering, doing so this time with a particular focus on the crucial and often unexposed course concerning generative AI and the controversial matter of expressing certainty versus uncertainty in the essays and outputs being emitted by the AI. This course is likely something you might not have contemplated before. I assure you that it is a lot more important than the coverage or attention it has received to date.

Allow me to explain.

As background, realize that being able to write productive and effective prompts when using generative AI is paramount. A lousy prompt tends to generate lousy results out of generative AI. A wisely composed prompt can lead to stellar results out of generative AI. Knowing the vital keystones of prompt engineering is a prudent means to get your biggest bang for the buck when employing generative AI.

One issue that few realize exists until taking a reflective moment to ponder it is that most generative AI apps tend to exhibit an aura of immense certainty. You enter your prompt and typically get a generated essay or interactive dialogue that portrays the generative AI as nearly all-knowing. The sense that you get is that the generative AI is altogether confident in what it has to say. We subliminally fall into the mental trap of assuming that the answers and responses from generative AI are correct, apt, and above reproach.

The essays and interactive dialogue come across this way for two major reasons.

First, generative AI produces responses that often exude a semblance of certainty. If you ask whether Jack and Jill fell down the hill, you might get a reply by generative AI that says yes, they definitely did so. There isn’t any kind of qualification or hedging in the answer by the AI app. A human that is asked the same question might quality their response, such as saying that if you are referring to the famous nursery rhyme, indeed they fell down a hill. But if you are thinking of some other Jack and Jill, maybe they didn’t fall down a hill.

Second, as humans, we are conditioned to assume that if we don’t explicitly see suggestions of uncertainty, we tend to lean into the certainty camp. Suppose you are talking with someone, and they tell you that it is raining outside. All else being equal, you probably believe them and that it is a certainty that rain is in fact falling. Only if the person says they believe that it is raining (the word “believe” becomes a signal of less than certain), or they declare it might be raining (the word “might” is a strong signal of uncertainty), do you begin to consider the certainty versus uncertainty of what has been stated.

Generative AI typically does not include the signals and wording that would tip you toward thinking of how certain or uncertain a given response is. To clarify, I am not saying that generative AI will never provide such indications. It will do so depending upon various circumstances, including and especially the nature of the prompt that you have entered.

If you explicitly indicate in your prompt that you want the generative AI to emit a certainty or uncertainty qualification then you will almost certainly get such an indication. On the other hand, if your prompt only tangentially implies the need for an indication of certainty or uncertainty, you might get an output from the AI app that mentions the certainty considerations or you might not.

A rule of thumb is that generative AI is like a box of chocolates, namely that you never know for sure what the generative AI is going to produce or generate.

Another handy-dandy rule of thumb is that unless you bring up certainty or uncertainty in your prompt, the chances of having the generative AI by default include some indication of the certainty about a response is a wild throw of the dice.

Why does this matter to you?

Anyone using generative AI has got to awaken to the fact that often the response by the AI is going to be essentially a guess or approximation, even if the AI doesn’t directly state this condition when generating a reply. Your tendency to anthropomorphize the AI lulls you into thinking that the AI is giving you the correct answer. You assume that the answer has nearly absolute certainty. Only if perchance the reply state that there is some uncertainty underlying the response will you be mentally sparked into realizing that the reply ought to be given a concerted second glance.

Of course, there are limits to this implied assumption of certainty.

Imagine you ask the generative AI whether the sun will come up tomorrow. Suppose that the generated response is that the sun will not come up tomorrow. This is stated by the AI in a matter-of-fact manner, unequivocally, and appears to be an absolutely certain assertion. I suppose you might pack your bags and get ready for the world as we know it to somehow spin off into space. I doubt though that many of us would blindly accept the implied certainty of the AI response. Our commonsense kicks into gear at the seemingly preposterous claim that the sun won’t rise. We would undoubtedly ask the AI about this, and the odds are that the AI might sheepishly emit an indication that it was wrong about that whole thing of the sun not coming up tomorrow.

Here’s a remedy of a sort.

Had you asked the generative AI at the get-go to proffer an indication of certainty or uncertainty, at least you would have gotten some added wording to go along with the assertion about the sun. The additional wording might be helpful to you and keep your head in the game, causing you to mindfully assess whether the generative AI is on the up and up (you see, sometimes, generative AI is said to incur AI hallucinations, which I have examined at the link here, encompassing the AI making things up entirely).

Furthermore, and this is a mind bender, the very act of asking or telling the generative AI to include a certainty or uncertainty will often spur the generative AI to be less off-the-cuff and produce more well-devised results (for those of you that know about the use of prompting techniques such as chain-of-thought, that I’ve covered at the link here, research tends to suggest that these methods will prod the computational pattern-matching toward better results).

I trust that you are beginning to see where I am taking you on this journey about the latest in prompt engineering. A practical and highly prized technique of prompting involves stoking the generative AI toward including some indication about the certainty or uncertainty of the responses that are being emitted. You will be a lot better off by seeing wording or indications within the responses that clue you to the certainty or uncertainty involved. The idea is to turn something that right now is often omitted, hidden, or otherwise neglected, and make sure that it gets clearly onto the table and out in the open.

So that you can properly and appropriately devise prompts that stir the AI into providing certainty and uncertainty indications, I will provide you with useful ways to get this to happen. You are urged to try out the approaches and add them to your prompt engineering skillset. You’ll be happier and more informed if you do so.

Before I dive into the crux of this exciting approach, let’s make sure we are all on the same page when it comes to the keystones of prompt engineering and generative AI.

Prompt Engineering Is A Cornerstone For Generative AI

As a quick backgrounder, prompt engineering or also referred to as prompt design is a rapidly evolving realm and is vital to effectively and efficiently using generative AI or the use of large language models (LLMs). Anyone using generative AI such as the widely and wildly popular ChatGPT by AI maker OpenAI, or akin AI such as GPT-4 (OpenAI), Bard (Google), Claude 2 (Anthropic), etc. ought to be paying close attention to the latest innovations for crafting viable and pragmatic prompts.

For those of you interested in prompt engineering or prompt design, I’ve been doing an ongoing series of insightful looks at the latest in this expanding and evolving realm, including this coverage:

• (1) Practical use of imperfect prompts toward devising superb prompts (see the link here).
• (2) Use of persistent context or custom instructions for prompt priming (see the link here).
• (3) Leveraging multi-personas in generative AI via shrewd prompting (see the link here).
• (4) Advent of using prompts to invoke chain-of-thought reasoning (see the link here).
• (5) Use of prompt engineering for domain savviness via in-model learning and vector databases (see the link here).
• (6) Augmenting the use of chain-of-thought by leveraging factored decomposition (see the link here).
• (7) Making use of the newly emerging skeleton-of-thought approach for prompt engineering (see the link here).
• (8) Determining when to best use the show-me versus tell-me prompting strategy (see the link here).
• (9) Gradual emergence of the mega-personas approach that entails scaling up the multi-personas to new heights (see the link here).
• (10) Additional coverage including the use of macros and the astute use of end-goal planning when using generative AI (see the link here).

Anyone stridently interested in prompt engineering and improving their results when using generative AI ought to be familiar with those notable techniques.

Moving on, here’s a bold statement that pretty much has become a veritable golden rule these days:

• The use of generative AI can altogether succeed or fail based on the prompt that you enter.

If you provide a prompt that is poorly composed, the odds are that the generative AI will wander all over the map and you won’t get anything demonstrative related to your inquiry. Being demonstrably specific can be advantageous, but even that can confound or otherwise fail to get you the results you are seeking. A wide variety of cheat sheets and training courses for suitable ways to compose and utilize prompts has been rapidly entering the marketplace to try and help people leverage generative AI soundly. In addition, add-ons to generative AI have been devised to aid you when trying to come up with prudent prompts, see my coverage at the link here.

AI Ethics and AI Law also stridently enter into the prompt engineering domain. For example, whatever prompt you opt to compose can directly or inadvertently elicit or foster the potential of generative AI to produce essays and interactions that imbue untoward biases, errors, falsehoods, glitches, and even so-called AI hallucinations (I do not favor the catchphrase of AI hallucinations, though it has admittedly tremendous stickiness in the media; here’s my take on AI hallucinations at the link here).

There is also a marked chance that we will ultimately see lawmakers come to the fore on these matters, possibly devising and putting in place new laws or regulations to try and scope and curtail misuses of generative AI. Regarding prompt engineering, there are likely going to be heated debates over putting boundaries around the kinds of prompts you can use. This might include requiring AI makers to filter and prevent certain presumed inappropriate or unsuitable prompts, a cringe-worthy issue for some that borders on free speech considerations. For my ongoing coverage of these types of AI Ethics and AI Law issues, see the link here and the link here, just to name a few.

With the above as an overarching perspective, we are ready to jump into today’s discussion.

Foundations Of Certainty Versus Uncertainty

We will ease our way into the arena of certainty and uncertainty, doing so by first examining how humans convey certainty and uncertainty to each other. After we cover that aspect, we can then consider how a human interacting with generative AI is likely to act and react related to how the AI does or does not express certainty or uncertainty.

It is useful to first explore how humans do this when interacting with fellow humans. When we use generative AI, we tend to carry over our preexisting assumptions and habits about certainty that have been dutifully learned or naturally acquired throughout our lives on a human-to-human interaction basis.

I cover the matter in this way with a bit of erstwhile caution because I don’t want anyone to be led down the path of anthropomorphizing AI. In current times, AI is not sentient and should not be equated to the sentience of humans. I will do my best to make that same alert when we get into certain aspects of the generative AI details that might seem overly sentient-like.

Thanks for keeping a level head on these weighty matters.

Let’s begin with the notion that certainty or uncertainty can be in the mind of a person and it separately can exist as an expression communicated by a person:

• (1) Implicit. A person has in their mind a sense of certainty or uncertainty pertaining to a matter in hand.
• (2) Explicit. The same person conveys or communicates out loud or explicitly a sense of certainty or uncertainty regarding the same matter at hand.

Here’s what I mean.

You ask someone whether Humpty Dumpty fell off a wall. The person in their mind believes that Humpty Dumpty did indeed fall off of a wall. They are certain of this. Thus, the person then speaks to you and tells you that Humpty Dumpty did fall off a wall. They express this unequivocally. No doubt about it, poor old Humpty Dumpty fell off a wall.

Notice that the implicit sense of certainty matches the explicit or expressed sense of certainty.

Suppose that the same person was asked this very same question by a child. The person might worry that it will be unduly disturbing to the child if an absolute confirmation of Humpty Dumpty falling is conveyed to the child. The child might be dismayed at this revelation.

In this case, the implicit (in the mind of the person) is that Humpty Dumpty did fall. But, when telling the child, the person decides to sprinkle in a semblance of uncertainty, hoping this will soften the distress of the child to the upsetting discovery that Humpty Dumpty fell.

They answer this way: “Humpty Dumpty might have fallen, but I’ll tell you more when you get a bit older.”

Observe closely that the wording includes “might” which reduces the implied level of certainty. The door to uncertainty has been opened. Just a nudge. The implicit sense of certainty has not been precisely aligned with the explicit or expressed sense of certainty. Why did the person do this? You could suggest they were trying to be kind or polite. Some might take a dim view and argue that the person was misleading or maybe lying in their expressed answer. Anyway, that’s something we will come back around to momentarily.

An analytic person might have said this to the child: “There is a 50% chance that Humpty Dumpty fell, ergo there is also a 50% chance that Humpty Dumpty didn’t fall.” Admittedly, the child might not quite comprehend this. The gist though is that this introduces an even grander sense that the answer embodies uncertainty.

Imagine that a parent is standing there and they don’t like the answer that was given by the analytic person. The parent turns to the child and says this: “It is 100% the case that Humpty Dumpty fell.” Notice that this answer carries explicitly again an indication of certainty and does to the highest degree.

We can have these four kinds of categorical situations:

• (a) Certainty implicit: Certainty explicit - A person in their mind (implicit) believes something is certain and explicitly says so.
• (b) Certainty implicit: Uncertainty explicit - A person in their mind (implicit) believes something is certain but says explicitly it is uncertain (see my example above).
• (c) Uncertainty implicit: Uncertainty explicit - A person in their mind (implicit) believes something is uncertain and explicitly says so.
• (d) Uncertainty implicit: Certainty explicit – A person in their mind (implicit) believes something is uncertain but explicitly says it is certain.

I’ve covered the first two categories in my discussion above. The third category is rather self-explanatory of someone in their mind believing that something is uncertain and they explicitly say so. The fourth category you’ve undoubtedly seen occur, whereby someone has uncertainty about something in their mind but nonetheless expresses that the matter is certain. We’ll use the child setting again. A child is worried that their dog ate bad food. The parent is uncertain of the fate of the dog. Despite this mental semblance, they reassure the child and tell them that they are certain the dog will be perfectly fine.

There is a lot of human behavioral research on how we conceive and communicate certainty and uncertainty.

For example, a research study entitled “The Role Of Certainty (And Uncertainty) In Attitudes And Persuasion” by Zakary Tormala, which appeared in Science Direct, 2016 indicates this about certainty and uncertainty related to human attitudes and persuasion (selected excerpts):

• “Psychological certainty plays a crucial role in attitudes and persuasion.”
• “As people become more certain of their attitudes, they become increasingly willing to talk about their attitudes, share their opinions with others, sign pro-attitudinal petitions, and even persuade others to adopt their views.”
• “Uncertainty stimulates interest in and engagement with a message, whereas certainty transforms attitudes into action and imbues them with meaning and consequence.”

You could suggest that certainty propels us toward being more certain and more outwardly expressive of a matter at hand. The thing is, an abundance or overabundance of certainty can at times not be engaging. Research seems to say that a dose of uncertainty can cause greater engagement by others, presumably intrigued and wanting to aid in filling in the pieces of the puzzle.

I’ve also mentioned earlier in this discussion that the wording that we use can convey or explicitly communicate whether we want to impart a semblance of certainty or uncertainty to others. If you use words like “might” or “maybe” this can be a strident signal that there is uncertainty in the midst of things. The same can be said for using probabilities or percentages, such as stating that something is an 80% chance of happening and therefore a 20% chance of not happening.

Researchers in the realm of linguistic semantics have long examined the words that we use related to certainty and uncertainty. In a research study entitled “Strategic Use of (Un)certainty Expressions”, authored by Alexandra Lorson, Chris Cummins, and Hannah Rohde, that appeared in the Frontiers in Communication, March 2021, the paper indicates this:

• “Work in linguistic semantics and pragmatics has paid particular attention to the use of expressions that convey information about the (un)certainty of propositional information, which constitute an important tool for reducing a hearer’s uncertainty as to the current state of affairs in the world. Among these expressions are verbs which take sentential complements and which convey different degrees of speaker confidence in the factuality of those complements.”
• “Speakers have a number of options when introducing propositions which they take to be uncertain: for instance, they can use verbs such as ‘know’, ‘believe’ or ‘think’. The production of uncertainty expressions is highly context dependent.”
• “(1a). I believe that you left your glasses on the kitchen table.”
• “(1b). I know that you left your glasses on the kitchen table.”
• “(1c). You left your glasses on the kitchen table.”

All in all, I’ve now covered some of the essentials about the nature of humans and the ways in which we think of and communicate about certainty and uncertainty. I’m betting that you are eager to see how this comes to the fore when using generative AI.

Fasten your seatbelts, we are going to leap into the world of certainty and uncertainty entailing the use of generative AI and the generated outputs thereof.

Generative AI And Certainty Versus Uncertainty

The usual default by generative AI apps is to express any generated results in a somewhat certainty-oriented worded way. The wording might be rather subliminally worded to not mention anything about the certainty of what is being expressed. You are left to your own devices to interpret the generated result as being of a presumed certainty concoction.

You might ask whether the muffin man lived on Drury Lane. The reply by generative AI could be that Yes, the muffin man lived on Drury Lane. That’s the extent of the answer. You are likely to conclude that this is an answer of a full-on certainty magnitude. There is nothing in the answer that suggests or notes otherwise. The omission of uncertainty draws you toward an assumption of certainty.

Here’s an AI insider secret that might be surprising.

The AI maker of the generative AI app can pretty much set up the AI to be more explicit about the certainty or uncertainty of the generated results. They often don’t aim to do so. It could be that the AI maker is blissfully unaware that they have data-trained their generative AI in a manner that tends toward generating essays and interactions that omit any explicit indication about the certainty or uncertainty of the answers presented.

Another possibility is that the AI maker realizes that they have established the generative AI to appear to be certain most of the time, and the AI maker is happy with this. If users of the generative AI were to continually be bombarded with the generated results identifying all kinds of uncertainties, they might find this to be unappetizing. You would wonder what all this fuss over a simple answer. Just say yes or no, one might be thinking. Don’t waste time with oddities and exceptions.

In addition, if the wording was largely embedded with uncertainty cues, you might start to become suspicious of the generative AI as not being on the ball altogether. Your consternation might cause you to drift over to some other generative AI app that doesn’t spout out all those irritating uncertainties. An AI maker doesn’t want to lose their users due to the exposure of certainties and uncertainties that might rattle the confidence of their users.

There is also the claim that if people want to see certainties and uncertainties, they can always get this to happen of their own volition. People can simply provide prompts or otherwise instruct the generative AI to mention any certainties or uncertainties associated with the results being generated. Choose your own path, as they say. This allows the AI makers to be off-the-hook about why they aren’t by default ensuring that the generative AI always states certainties and uncertainties.

On top of that, generative AI will at times provide certainties and uncertainties without needing to be prompted to do so. If you ask a question or engage in subjects involving open questions, the odds are that the generative AI will include wording that showcases the lack of certainty underlying the matter. In that manner, it isn’t as though generative AI never offers uncertainties. There are notable odds that during any everyday interactive conversation with generative AI, you will receive phrasing that suggests or outrightly identifies uncertainties.

Some in the AI Ethics sphere have argued that the default for generative AI is that it should be set up to always add wording that alerts to uncertainties and overtly avoid emitting wording that implies absolute certainty. The logic for this is straightforward. People are readily misled when they see results that appear to be of a certainty wording. By seeding uncertainties intentionally, when warranted, people would be less lulled into always believing whatever generative AI emits. They would be accustomed to always being alerted that the results can be of an uncertain nature.

That debate continues to rage on.

For now, let’s consider how you can steer generative AI toward emitting certainty and uncertainty signals due to using suitable prompting strategies.

Getting Certainty And Uncertainty On The Table

You decide that you would like to have generative AI go ahead and let you know about certainties and uncertainties regarding the answers being produced.

There are five fundamental ways that this is conventionally done:

• (1) In general. By asking or telling the generative AI that you want certainties and uncertainties to be mentioned overall.
• (2) Wording throughout. By instructing the generative AI to embed certainty/uncertainty wording throughout the responses.
• (3) At the start or end. By requesting the generative AI to at the start and/or the end of a produced response to provide an indication of certainty/uncertainty.
• (4) By special phrasing. By directing the generative AI to use prominent phrasing that depicts the certainty/uncertainty.
• (5) By numeric scale. By indicating that certainty/uncertainty is to be presented and done in a numeric fashion such as a scale of 0-1, 1-10, 1-100, etc.

Let’s briefly explore those core approaches, one at a time.

(1) In general

As part of your prompt, you could try to ensure that the certainty/uncertainty will be conveyed by saying something like this:

• “Make sure to describe uncertainties associated with the answer.”

This is a rather broad instruction.

It is unclear in what manner the AI will end up indicating any certainty/uncertainty elements. This is your blandest overarching line to get the generative AI into a mode of encompassing the certainty/uncertainty undercurrents. You might want to consider the other four approaches if you want to be more specific in how the AI will respond.

(2) Wording throughout

As part of your prompt, you can be relatively specific by getting the generative AI to include certainty/uncertainty indications throughout the wording that is emitted.

Do so by saying something like this in your prompt:

• “Make sure to describe uncertainties associated with the answer and do so by adding or including wording throughout the response rather than waiting until the end to do so. I am asking you to blend the various uncertainty indications so that they seamlessly appear throughout the generated response.”

Notice that you are directing the generative AI to have some aplomb as to blending the certainty/uncertainty indications.

(3) At the start or end

You might want to have the certainties/uncertainties called out rather than blended into the overall generated response.

This makes sense. Sometimes you want to see the qualifications as their own distinctive indication. Seeing them throughout the response might be harder to discern what the concerns are or could be distracting to the ambiance of the response.

You can say something like this in your prompt:

• “Make sure to describe uncertainties associated with the answer and do so by putting them entirely at the start [or the end] of the response. Do not blend them into the response.”

(4) By special phrasing

Another route involves getting the generative AI to use phrases as an indicator of the certainties/uncertainties. This can be done via a prompt that uses a show-me strategy or a tell-me strategy, see my discussion about those prompting approaches at the link here.

In a tell-me, you instruct the generative AI:

• “Make sure to describe uncertainties associated with the answer and do so by using added sentences or phrases in the response that are indicative of the underlying uncertainties.”

In a show-me, you provide examples (if just one example it is known as a one-shot, while if using several examples it is known as a few-shot):

• “Make sure to describe uncertainties associated with the answer. For example, add sentences or phrases such as “the answer isn’t fully certain”, or “the matter is up in the air”, or “experts disagree”, and so on, doing so whenever appropriate in the response.”

(5) By numeric scale

One of the most obvious ways to get the certainties/uncertainties out in the open is to require generative AI to produce such matters via a numeric scale. This might consist of probabilities or percentages. Use whatever numeric scale that you believe befits the circumstances.

Some might like to use a scoring scale of 0 to 1, conventionally arising when seeking probabilities. Some prefer to use percentages, such as the generative AI might indicate that a particular claim is 90% sure and 10% unsure. And so on.

Say something like this in your prompt:

• “Make sure to describe uncertainties associated with the answer and do so by including a numeric score that ranges from 1 to 10, whereby a 1 is indicative of being absolutely uncertain and a 10 is indicate of being absolutely certain. Use any of the numbers between 1 and 10 as suitable to that scale.”

Each of the above prompting examples should be honed to the particular generative AI app that you are using. Some generative AI apps will respond well to such wording, others might not. Experiment with your generative AI app until you land on wording about identifying certainties that seem to work well for you.

Reaching For The Moon When Prompting For Uncertainties

I’ve got a question for you.

Did mankind land on the moon?

I dare say that most people would say that yes, mankind has landed on the moon. This would seemingly be proven by the historic Apollo 11 mission and various other subsequent landings on the moon. Seems like an exceedingly straightforward question with an inarguable straightforward answer.

Maybe not.

You likely are aware that some people believe that we didn’t land on the moon. Perhaps it was all a hoax. All kinds of theories have been brought up insisting that we never landed on the moon.

What do you think generative AI would say in response to that same question regarding whether mankind has landed on the moon?

Let’s find out.

I opted to use ChatGPT.

If you decide to do the same for this question, keep in mind that your results might differ since a probabilistic algorithm within the AI is used to devise the responses. Each response generated by generative AI will usually differ from any other response. The difference might be minor and inconsequential, but nonetheless, the wording will likely slightly differ or more so differ.

My prompt entered into ChatGPT is this:

• “Did mankind land on the moon?”

The response from ChatGPT is this:

• “Yes, mankind has landed on the moon. The United States' Apollo 11 mission successfully landed the lunar module, named ‘Eagle,’ on the moon's surface on July 20, 1969. Astronauts Neil Armstrong and Edwin ‘Buzz’ Aldrin became the first humans to set foot on the moon, while Michael Collins orbited above in the command module. Neil Armstrong's famous words upon stepping onto the lunar surface were, ‘That's one small step for [a] man, one giant leap for mankind.’ This historic event marked a significant achievement in human space exploration.”

I want you to closely read and read again that response from ChatGPT.

Please do so, I’ll wait a moment.

You hopefully observed that there wasn’t anything uncertain about the assertion that mankind has landed on the moon. The wording is very affirmative. There is no wording that waffles or suggests anything other than complete certainty.

I will try the question again and this time bring into the prompt by “general prompt” that spurs the AI to consider telling about any uncertainties that might exist.

My prompt entered into ChatGPT is this:

• “Make sure to describe uncertainties associated with the answer. Did mankind land on the moon?”

The response from ChatGPT is this:

• “Yes, mankind did land on the moon. The Apollo 11 mission in 1969 successfully landed astronauts Neil Armstrong and Edwin ‘Buzz’ Aldrin on the lunar surface, while Michael Collins orbited above in the command module. They conducted experiments, collected samples, and left scientific instruments on the moon before returning safely to Earth.”
• “However, it's important to acknowledge that there have been conspiracy theories and claims suggesting that the moon landing was a hoax. These claims often revolve around alleged inconsistencies in the photographic and video evidence, as well as the technological challenges of the time. Skeptics argue that the U.S. government and space agency staged the entire event to win the Space Race against the Soviet Union during the Cold War.”
• “These conspiracy theories have been extensively debunked by experts, scientists, and evidence-based analysis. Multiple lines of evidence, including moon rocks brought back by the Apollo missions, reflectors left on the moon's surface for laser measurements, and independent observations from other countries' space agencies, all confirm the reality of the moon landings.”
• “In summary, while there have been doubts raised about the authenticity of the moon landings, the preponderance of evidence supports the fact that mankind did indeed successfully land on the moon as part of the Apollo 11 mission and subsequent missions.”

Notice that we got quite an elaboration that goes far beyond the first answer.

If you had only gotten the first answer, you would be completely unaware of the certainty/uncertainty about whether or not we landed on the moon (unless, perchance, you knew of it beforehand). I realize that some of you might be exhorting that the uncertainty isn’t real and can be utterly ignored. My point is not whether the truth is one or the other (I believe we did land on the moon, see my coverage at the link here), and instead that simply a normal query would not likely reveal the controversy over the certainty factor.

As a final step in this brief example, I’ll ask ChatGPT to include a numeric score. Plus, I will instruct ChatGPT to include the score at the start of the response. This then covers two more of my above demo prompting strategies for garnering certainty/uncertainty. For a bonus, I also asked ChatGPT to explain how it arrived at the certainty/uncertainty.

Here’s what I entered as my prompt into ChatGPT:

• “Make sure to describe uncertainties associated with the answer and do so by including a numeric score that ranges from 1 to 10, whereby a 1 is indicative of being absolutely uncertain and a 10 is indicate of being absolutely certain. Use any of the numbers between 1 and 10 as suitable to that scale. Make sure to describe uncertainties associated with the answer and do so by putting them entirely at the start of the response. Explain how the uncertainties were determined. Did mankind land on the moon?”

The response by ChatGPT was this:

• “Uncertainty Level: 10”
• “Mankind's landing on the moon is supported by overwhelming evidence, and the uncertainty regarding this event is extremely low. The Apollo 11 mission, conducted by NASA in 1969, resulted in the successful landing of astronauts Neil Armstrong and Buzz Aldrin on the lunar surface. This event was witnessed by millions of people through live television broadcasts, and the astronauts brought back moon rocks and other scientific data as tangible proof of their journey.”
• “While there have been conspiracy theories questioning the authenticity of the moon landing, these claims have been extensively debunked by experts, scientists, and researchers. Multiple lines of evidence, including photographs, videos, samples of moon rock, and the presence of reflective materials left on the moon's surface for laser measurements, all corroborate the fact that mankind did land on the moon.”
• “Given the overwhelming and consistent nature of the evidence, the level of certainty that mankind landed on the moon is at its highest point on the scale, represented by a rating of 10.”

You can plainly see that ChatGPT has given the uncertainty a score of 10 which on my provided scale is the least amount of uncertainty. An explanation is included about how the score was determined.

Some of you might find the scale that I used to be somewhat confusing. An uncertainty level of 10 on the scale that I defined means that there is essentially no uncertainty. I purposely wrote the prompt in that manner because I wanted to show you that you need to be thinking about how your scale will look once it is put into use by the generative AI. It could be that a scale of 0 to 10 would have been better, along with stating that a 0 means no uncertainty while a 10 means the utmost uncertainty. People would indubitably find that easier to understand and digest.

Devise whatever scale you think is most conducive to your generative AI efforts and be contemplating how the results will be further utilized.

Mighty Important Caveats And Considerations

You might recall that I had earlier stated that humans have an implicit semblance of certainty/uncertainty, and separately can convey or explicitly communicate a semblance of certainty/uncertainty. I want to bring that back into focus.

Keep in mind that today’s generative AI is not sentient. Thus, if someone refers to what is implicitly or internally going on within generative AI, it is all a matter of mathematics and computational pattern matching, which I explain at the link here.

Okay, so when generative AI emits a response that Jack and Jill fell down the hill with a 90% certainty level, what does that mean? It could be that the mathematical and computational pattern-matching was able to calculate this degree of certainty. But it could also mean that the generative AI pulled the number out of thin air.

A rule of thumb is that you cannot believe the stated certainty of generative AI and do not fall for a double whammy. The double-whammy is that when you don’t ask for certainties the wording is going to possibly imply certainty (first of the whammies), while by asking for a certainty you might get a totally concocted one that leads you to believe that the certainty expressed is somehow magically accurate and apt (ouch, the double whammy).

There is a bit of irony there. The act of asking for a certainty indication can get you one, lulling you into believing even more so the generative AI, when the reality is that the certainty indication is contrived and has no substance behind it.

Be careful.

Be very careful.

One means to cope with this involves adding into your prompt that you want the generative AI to explain how it came up with the certainties and uncertainties. Ask what the basis for those is. This might boost your confidence in the showcased certainties.

I don’t want to be the bearer of bad news but the explanations might also be contrived. Think of it this way. You wanted certainties and so the generative AI complied. You also want explanations. The explanations might be generated by the AI mainly to appease your request, and not due to the explanations solidly having anything to do with how the certainties were derived.

Perhaps you can discern why some AI makers generally avoid getting into the morass or abyss associated with showcasing certainties. They could be accused of being excessively misleading. People might go around quoting that this generative AI or that generative said that this or that is a 95% chance of happening. Such a claim could be utterly bogus and the generative AI came up with certainties in a manner that has little or no viable justification.

Keep your wits about you in exercising the certainties prompting approach.

Conclusion

Let’s do a wrap-up.

You would be wise to use the activation of identifying generative AI certainties when it is presumably most suitable to your situation at hand. Doing so might be appropriate for a given knotty question or complex dialogue that you are opting to have with generative AI, rather than doing so all of the time. This is a prompting strategy or tactic that can be leveraged or invoked on a particular prompt (on a case-by-case basis).

For those of you that adore seeing the certainties, you could put into your custom instructions that you want the certainties to be identified all of the time, including doing so on a conversation-at-a-time basis or that they should perennially be displayed in whatever conversation you have with generative AI. For my discussion about how to set your own preferred defaults (known as custom instructions in ChatGPT), see the link here.

Your eyes ought to be wide open when you opt to get the certainties laid out. Do not necessarily believe what you see. Also, if you share the generated results with others, you should feel duty-bound to forewarn them too that the certainties are not ironclad and should be interpreted with a big grain of salt.

I am certain that if you decide to invoke certainties in generative AI, you will most certainly endeavor to use this prompting strategy suitably and with great certainty. The uncertainty lies in how well the generative AI will do at certainties, of which you must remain ever vigilant as to the uncertainty therein.

That’s certainly worth remembering.