CHATGPT WROTE THIS
This month’s topic gives me pause, and frankly puts me into a somewhat somber mood. I say this because, without fanfare, and nearly unnoticed, we have reached an inflection point in human existence. I know that sounds dramatic, but it’s true, nonetheless.
Like many of my fellow Old Guy Electrical Engineers (OG-EEs), I grew up in the early 1970s, and I spent a lot of my free time reading science fiction and fantasy books. Some of my favorite authors were Robert Heinlein, Arthur C. Clarke, and of course Isaac Asimov. Clarke is rightly famous for, among other things, his chilling depiction of HAL 9000, the artificial intelligence (AI) built into the Discovery One starship in 2001: A Space Odyssey. And who could forget the positronic brains that controlled the androids featured in Azimov’s iconic I, Robot series.
As a young person, I found these stories fascinating. The budding engineer in me loved letting my imagination fly with that of the authors’. However, even the youthful me knew that though they were fascinating stories, they were still little more than creative dreams. Flash forward forty-odd years to today, and I find that the current reality we live in is slightly more nightmarish.
Robots and AI are no longer science fiction. The generation of engineers and scientists who were inspired by the visionary tales above has now brought them to full-fledged reality. And, while the infancy of AI and robotics may have seemed harmless at first, their adolescence is decidedly less so. It now seems clear that the authors who wrote about them all those years ago often did so as a form of cautionary tale. They could foresee some of the disruptive and disturbing consequences that might occur. No less a person than Elon Musk has expressed serious reservations about the speed at which the current AI and robotics revolutions are moving. Not the least of his concerns is the impact on human society, and the resulting upheaval that these technologies may produce.
For instance, I am sure that most of the readers of this publication have heard about chatbots, and many may have interacted with them. If you had asked me about a month ago what I thought of AI-driven chat, I would have said something on the order of, “Yeah, AI chat is pretty good, but as soon as you get outside of the bounds of its program, you can easily tell it’s not a real person.” Little did I know how naïve that belief was.
I think I can be forgiven in that I’m not one to spend a lot of time online with these types of distractions. My assessment was based on my interactions with SIRI, or Google Assistant—mobile-based personal assistant software that often gives incorrect or unhelpful answers to any question that exceeds their understanding. However, in researching this article, I became aware of just how sophisticated and subtle AI technology has become.
THE ART OF DECEPTION
Let’s discuss an object lesson. You may have heard a lot in the news lately about OpenAI, the AI research corporation founded in 2015 by Ilya Sutskever and Greg Brockman which is ostensibly an open-source, non-profit venture. It was initially co-chaired by Sam Altman and Elon Musk, and counts Peter Thiel as one of its donors, among notable others. In its original charter, OpenAI pledged to collaborate with academic institutions and other researchers, providing the results of their efforts freely to all under an open-source licensing arrangement. Operating in this way allowed them to leverage industry partnerships to obtain access to almost unimaginable amounts of “non-personal” data from social media sources such as Facebook and Twitter. These massive data sets were used as input to inform and train their neural-network engines. After several years of development and refinement, two customer-facing offerings have become available for general interaction. The first, named DALL-E , is an interactive, AI-powered drawing portal. The second is the chatbot ChatGPT (GPT stands for Generative Pre-trained Transformer) .
As per my due diligence, I played around with both offerings, as well as with Craiyon  and several others. CrAIyon and DALL-E are designed to produce images inspired by a series of prompts or instructions from the human designer to the AI artist. You type a short description of the image you wish, including rendering details like “use all blue colors” or “in the style of Claude Monet.” The AI then parses and evaluates the instructions, creating a short series of images based on its interpretation.
The images generated by DALL-E are interesting. As you can see in Figure 1, they are imaginative and well-drawn, but, in my opinion, easily distinguished from art created by a real human. I also tried Craiyon, with the resulting images shown in Figure 2, Figure 3, and Figure 4. All well and good. So far it seems that the robot apocalypse is still in the future. But wait, there’s more.
Figure 5 shows an interaction I had with ChatGPT. As readers of this column know, besides being an engineer, I am also an author of fiction stories. After seeing what the drawbots produced, I thought a fun way of illustrating the shortcomings of this type of AI would be to ask it to write me a short story. I believed the quality of the writing would be of the same order as the AI graphics, but the idea was to test the capabilities of the algorithm and have a little fun at the same time. Instead, the results flatly astounded me.
In Figure 5, my input is in white on the left, and ChatGPT’s responses are in blue on the right. I asked it to assume the persona of an author named Harrison, and I gave that persona some characteristics. I then addressed ChatGPT as Harrison and asked what it was currently writing. The answer I received was detailed, topical, and conversational. I then told it about the kind of stories I liked to read, and asked if it could write me a short story in that specific genre.
The story it produced was well-written, cogent, and had a clear character arc. It was, in my humble opinion, better quality writing than most university-educated adults might be able to muster. And, most distressing, it was created in mere seconds. A human author, even a good one, might take hours to craft a similar work. This was an order of magnitude better than the AI art, and I would challenge any reader to pick out an AI-generated story from one written by a human.
ROBOTICS: THE ZEROTH LAW
As mentioned above, in his classic compilation of short stories and novels, I, Robot, Isaac Asimov created and described a world where robots and humans interacted at all levels of society. Robots, as Asimov foresaw, were immensely superior to their creators in terms of strength and speed. So in his invented world, the “Three Laws of Robotics”  were implemented at the core of every robot’s brain, making them incapable of harming humans. The laws were so fundamental to the function of the robot that if there were any attempt to circumvent them, the brain would self-destruct. They were designed to be interlocking, arranged in order of precedence such that the lower-level law would prevail in any case where the higher-level law was insufficient for the robot to resolve a conflict. The laws were, in order, as follows:
- A robot may not injure a human being or, through inaction, allow a human being to come to harm.
- A robot must obey the orders given to it by human beings, except where such orders would conflict with the First Law.
- A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
Many of the stories Azimov wrote concerning robots had to do with how those laws, as interpreted by the AI of his robots’ positronic brains, would cause them to behave in counterintuitive or bizarre ways. Later in the series of stories, as the increasingly sophisticated robots came to oversee whole planets and societies of humans, he added the “Zeroth Law,” which states:
A robot may not harm humanity, or, by inaction, allow humanity to come to harm.
Asimov was a visionary, far ahead of his time. He could infer that robots would become increasingly functional, sophisticated, and ultimately dangerous to their creators. His “Three Laws” idea was an attempt to mitigate these concerns. It does not take much imagination now to see that a Tesla vehicle with Full Self-Driving (FSD) capability is a functional autonomous robot designed to move at speed through crowded highways and streets. The only thing keeping the human lives within it and around it safe is the ability of its AI to follow the immensely more complicated laws of our human society. This raises many questions. For instance, when a self-driving robot car becomes involved in an accident where property is damaged or people are injured, who is liable? Under our current regulations the driver is responsible for the consequences caused by his or her vehicle, but if you are sitting in an FSD automobile, can you be said to be driving? These and other questions arise. Who is responsible for a vehicle that is moving when completely unoccupied? What are the societal implications when nefarious actors can use an autonomous vehicle to deliver a weapon to a specific location, or to transport contraband or explosives? All the above scenarios are conceivable, and all fall outside of our current set of laws. We are entering unknown territory at breakneck speed.
Given all the above, let’s look at some of the players in AI and robotics. There are a lot of names in this space, ranging from gigantic multinational corporations to small research teams performing graduate thesis work. It’s an exciting time to be in this field. So much innovation is occurring all at once that it’s difficult to keep up with the pace of the advances. I break the companies up into the two categories of robotics and AI, but in truth that distinction is becoming increasingly blurred. As I noted, the latest Tesla model is no less a robot than anything created by Boston Dynamics. In fact, Tesla themselves look at their product as an AI that happens to also function as a car.
Robots, or at least the concept of autonomous machines, have been around far longer than AI. We need look no further than the story of Pinocchio, or for that matter Pygmalion, to find our ages-long fascination with these creations. The advent of modern-day non-fiction robots began in the industrial world of production and processing machines. These devices incorporated more and more functionality, accelerating with the introduction of microprocessor control units in the ‘80s and ‘90s. To save costs and increase competitiveness many companies began to use industrial robots to replace human workers, though true autonomy was still far out of reach for many years. But over the last decade or so that barrier has been eroded, and as we will see, several companies are touting fully autonomous manufacturing robots. This second industrial revolution promises to be as disruptive as the first, with some predicting that the end of human-based labor is on the horizon.
Tesla: Elon Musk’s Tesla  has been at or near the forefront of robotics innovation for some time. On its website, Tesla has assembled a full division of software engineering talent to work on their robotic initiatives. The Tesla Bot team’s stated mission is to develop “a general purpose, bi-pedal, autonomous humanoid robot capable of performing unsafe, repetitive or boring tasks.” In other words, a general-purpose replacement for much of the skilled and unskilled human labor that is the backbone of heavy industry or manufacturing concerns. The upside potential of this technology is tremendous: a world where robot labor replaces the need for human-powered drudgery. However, that world might also suffer from the fact that although robots don’t need salaries, meal breaks, benefits, or HR departments, they also don’t pay taxes, a fact that might cause state and federal government officials to lie awake at night wondering where their revenue is going to come from.
Boston Dynamics: Arguably the scariest of the robotic firms I’ve ever researched, Massachusetts-based Boston Dynamics, currently owned by the Korean multinational Hyundai Motor Corp., has several generations of robotic creations. They range from SPOT, the robot dog that was recently deployed with the FDNY for rescue and recon missions , to STRETCH, the warehouse picker and loader bot, to ATLAS, the general-purpose humanoid robot that can do backflips and parkour.
The history of Boston Dynamics is interesting, and it’s well worth your time to browse the company’s website. The implications of its technology are as disruptive, if not more-so, than the Tesla offering. SPOT, alone, I find personally terrifying.
iRobot: On the opposite end of the spectrum from Boston Dynamics we have iRobot, named after the aforementioned Azimov classic. Maker of cute cuddly, cat-friendly bots like the vacuum bot Roomba or the mop bot Braava, iRobot produces consumer-grade, semi-autonomous household assistants that are designed to perform menial household cleaning tasks. These robots are innocuous, practical, and seemingly non-threatening.
Husqvarna: For lawncare, the slightly more threatening Automower  from Swedish manufacturer Husqvarna (Figure 6) is basically a Roomba scaled up to lawnmower size. This sounds like a cool way to get your Saturday morning back from your ever-growing grass chores. But it might give the neighbors pause as it rolls across your property whirling its razor-sharp blades at 3 AM.
NVIDIA: Well-known in the arena of high-resolution graphics processors, I was surprised to learn that the California-based NVIDIA  also offers the ISAAC robotics platform to support the development and deployment of robots for various industries. Named after, you guessed it, Isaac Asimov, ISAAC can create virtual environments for testing robots and managing robot fleets. NVIDIA also has AI offerings, like the Jetson Orin AI that can support autonomous machine control.
General Dynamics: A huge contractor in the defense sector, General Dynamics has several divisions dedicated to the development and deployment of robotic devices for intelligence gathering, surveillance, and other more kinetic missions. Their Bluefin  unmanned underwater vehicles (UUVs) can operate completely autonomously to provide ship hull and infrastructure inspections, port and harbor security patrols, search and salvage operations, and support to scientific and research missions.
As I said, today the line between robotics and AI is almost indistinguishable. For many years that was not so. Robots, more akin to the predecessors of the Roomba, were purpose-built machines. These devices were able to operate well within their bounds, but as soon as they were taken out into an environment for which they weren’t designed, they would fail miserably. AI, beginning in the research labs of institutions like Caltech and MIT, was supposed to allow machines to learn and adapt to their environment, as well as perform a general class of tasks—for example, walking through mostly unknown conditions like uneven terrain with random obstacles. The advantage of this type of learning is obvious, but it required orders of magnitude more computing power than had the simple industrial robots. AI, therefore, lagged behind robotics for a long time. Recently that playing field has been leveled. AI is being used to leverage all the robotics design work, making it many times more powerful and effective (and by that, I mean frightening).
OpenAI: As discussed, OpenAI is pushing the bounds of AI. Their focus has shifted from non-profit to a more proprietary model, angering some of its original founders, and inspiring some, like Elon Musk, to announce their own GPT initiatives.
Tesla: Tesla’s focus has always been the deployment of AI-powered vehicles and machines at scale. Space X, with its autonomous drone ships and self-landing autonomous spacecraft, is also at the forefront of applied AI technology. The next “arms race” may well be driven by AI, and the competition between firms in this sector, while producing breathtaking innovation, may also cause some culture shock as AI makes its way into mainstream consciousness.
IBM: If you’ve been around a while, you may remember IBM’s Watson playing Jeopardy against a couple of human champions. The AI famously gave the wrong answer in the final Jeopardy, naming Toronto as a US city. However, fast forward to today, and IBM is still near the forefront of AI tech with Watson Assistant , the company’s virtual agent platform. Targeting the customer relationship, Human Resources, and customer support sectors, Watson Assistant allows you to develop a friendly and helpful automated front-end chat that can offload support questions from human subject matter experts. As with most chatbots today, Watson Assistant uses full Natural Language Understanding (NLU) to allow user queries to be phrased in conversational speech.
Spellbook: The legal profession is not exempt from the AI revolution. Rally Legal touts its application Spellbook  as “ChatGPT for Law.” Spellbook is an AI-powered portal that will write and review contracts, explore and reference case law and trial law, and analyze documents from a legal standpoint. The offering is not available for the public to test now, but early adopters are being solicited and Rally plans to launch the service soon.
MiniGPT: Lastly, a small team from King Abdullah University in Saudi Arabia has released MiniGPT-4, a chatbot that will take as input your text prompts and a picture. I gave it a picture of my cat, and asked for a poem about the picture (no other details). The AI generated a surprisingly good poem (Figure 7). They also give many other examples on their website  of their AI creating webpages, writing code, and demonstrating its ability to understand humor.
DANCING WITH DANGER
We’ve mentioned several potential problem with the broad application of AI and robots, and the resulting cultural upheavals these technologies may cause. One issue that has come to my attention is chilling. As described in a recent broadcast  on the AZFamily channel, AI is being used by scammers to make deepfake hostage calls. In this case a few seconds of the voice are captured by the scammer. This is then fed into an AI which can simulate the exact tone and cadence of the sampled voice and can be made to say whatever the scammer wants. The mother in this specific situation covered by the broadcast was quick-thinking enough to use alternate channels to check on her daughter’s actual status, who was OK—but the potential for abuse of this form of AI is obviously vast.
PUSHING THE BOUNDARIES
What is next on the AI/Robotics horizon? It’s a good question, and it is always hard to know what may come out in the next year or so. The current pace of AI and robotics is accelerating, and the shock wave which precedes them is only beginning to wreak havoc on human society. To wit, I was reading about a new type of AI called AutoGPT . Billed as the prototype of the next frontier in AI, it is supposed to be a sort of “Agent Smith” (Figure 8), meaning an AI that recursively clones itself to scale to the size of the task it’s trying to perform. As stated on their website, this is achieved by having the AI detect when its context gets overwhelming. At the point where this threshold is triggered, the AI begins offloading—in essence splitting itself into clones—and distilling the cognitive overflow from the parent clone into a prompt directive, or sequence of directives for the child. The clones then begin to dialog, talking back and forth over the chat channel to get the original task managed. As the developers note, this mechanism is crude, but the emergent paradigm is a lot more powerful in concept. Currently, clones are limited to transferring information via the text-chat channel, albeit more efficiently than a human might. However, increasing the bandwidth of this link by several orders of magnitude is the next step. We may be getting into Matrix-level computing sooner than we know.
Well, I don’t know about you, but after learning all that, I decided to live off the grid in a cabin in the Appalachian foothills before Skynet sends its latest T-9000 killer assault bot after me. As always, the above is just one man’s opinion. Do your own research, or better yet, get your favorite GPT to do it for you. I’m sure it will be correct, and totally unbiased. Until next time. DALL-E: https://labs.openai.com
 ChatGPT: https://openai.com/blog/chatgpt
 Craiyon: https://www.craiyon.com/
 The Three Laws of Robotics: https://www.historyofinformation.com/detail.php?entryid=4108
 Tesla’s AI & Robotics page: https://www.tesla.com/AI
 SPOT from Boston Dynamics: https://nypost.com/2022/04/01/fdnys-new-robot-dog-makes-its-first-appearance-in-nyc/Robotics: https://builtin.com/robotics/robotics-companies-roundup
 Husqvarna Automower: https://www.husqvarna.com/us/robotic-lawn-mower
 NVIDIA robotics: https://www.nvidia.com/en-us/deep-learning-ai/industries/robotics/
 Bluefin from General Dynamics: https://gdmissionsystems.com/underwater-vehicles/bluefin-robotics
 IBM AI and Watson Assistant: https://www.ibm.com/products/watson-assistant
 SpellBook: https://www.spellbook.legal/chat-gpt-for-law?twclid=2-1e91dswe86dx5vay9f7vdn25s
 AI Hobbyist: MiniGPT-4 – https://minigpt-4.github.io/
 AI Risks: https://www.azfamily.com/2023/04/10/ive-got-your-daughter-scottsdale-mom-warns-close-encounter-with-ai-voice-cloning-scam/
 AutoGPT – Pushing the boundaries: https://github.com/Torantulino/Auto-GPT
Artificial Intelligence: https://builtin.com/artificial-intelligence/ai-companies-roundup
Google BARD – Emergent Properties: https://www.dailymail.co.uk/news/article-11980033/Google-CEO-says-doesnt-fully-understand-new-AI-program-Bard-works.html
Google BARD: World Domination https://www.dailymail.co.uk/sciencetech/article-11964499/We-asked-AI-artificial-intelligence-surpass-human-beings.html
Boston Dynamics: Dancing bots https://youtu.be/fn3KWM1kuAw
AutoGPT | autogpt.net
Boston Dynamics | www.bostondynamics.com
General Dynamics | www.gd.com
Husqvarna | www.husqvarna.com
IBM | www.ibm.com
iRobot | www.irobot.com
NVIDIA | www.nvidia.com
OpenAI | openai.com
Rally Legal | www.rallylegal.com
Tesla | www.tesla.com
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • JUNE 2023 #395 – Get a PDF of the issueSponsor this Article
Michael Lynes is an entrepreneur who has founded several startup ventures. He was awarded a BSEE degree in Electrical Engineering from Stevens Institute of Technology and currently works as an embedded software engineer. When not occupied with arcane engineering projects, he spends his time playing with his three grandchildren, baking bread, working on ancient cars, backyard birdwatching, and taking amateur photographs. He’s also a prolific author with over thirty works in print. His latest series is the Cozy Crystal Mysteries. Book one, Moonstones and Murder, is already in print, and book two is on its way. His latest works include several collections of ghost stories, short works of general fiction, a collection called Angel Stories, and another collection called November Tales, inspired by the fiction of Ray Bradbury. He currently lives with his wife Margaret in the beautiful, secluded hills of Sussex County, New Jersey. You can contact him via email at email@example.com.