• AI


    One of our readers sent me a note after the last newsletter suggesting that I write a piece on AI.

    My experience of using AI is relatively limited. I use it to generate artwork, logos, and occasionally for graphic design. I also use it to assist with writing, but in a minor capacity, primarily for grammar and sentence structure suggestions. For instance, I am currently dictating this article to my computer. The way it appears on the screen suggests that there is some relatively advanced real-time processing involved in interpreting my vocalizations and transforming them into coherent English text. I am certain that AI plays a role in this process.

    If you have used Chat GPT or Google Bard or any other web based AI. You will have found that it’s like a very fancy search engine. You can ask it a question and it goes and looks at the web and rather than bringing back a set of URLs, websites that may contain the information you’re looking for, it goes to those sites itself and tries to collate the information and presents it as a piece of language in standard English for me, maybe Hindi for you or Japanese.

    This is slightly problematic – it represents the results of searching the web as a piece of writing that purports (in terms of its own internal language) to be entirely factual. However, it often is not. When you get a series of URLs in response to your enquiry, you need to apply the appropriate degree of skepticism based on the source and your own knowledge and instincts.

    Regardless, this is an interesting use of AI. If we project this forward and the AI algorithms become better at distinguishing fact from error, or phrasing its language to present subjective opinions in the correct context, websites will become meaningless. AI will be able to tell you exactly what you want to know without having to refer you to a website and forcing you to go through the tedious process of clicking on a link and then reading through the website to find the specific information you desire.

    Of course, this would completely change the way that we use the Internet and the way we post information. All attribution, branding, and value would be conferred entirely to the AI and not to the source material. This will be problematic as people will stop posting things online, as there will be no point, no credit for doing so, and no potential advertising revenue. The AI bot will have less and less contemporary and relevant information to select from as people stop posting information. There will be some limits placed on AI, and a happy medium will likely be reached.

    If you’ve ever tried to use AI in any form, it is very impressive, but its behavior is that of an autistic savant. As an engineer, I empathize with that type of personality, and I’ve worked with many of them, and I am one myself. So, for me, using AI is a relatively naturalistic approach to assembling information.

    So what does language AI do?

    It does something like this:

    There follows ten rewrites of the simple phrase: The dog jumps over the cat. The results of the AI rewrite is used as the input of the next rewrite:

    1. The cat is effortlessly leaped over by the dog as it gracefully jumps through the air.
    2. The dog gracefully jumps through the air, effortlessly leaping over the cat.
    3. The cat gazes in awe as the dog effortlessly soars through the air, displaying its graceful agility and leaping over with ease.
    4. As the cat looks on, it is filled with amazement as the dog effortlessly takes flight, showcasing its exquisite agility and effortlessly clearing the obstacle in its path.
    5. As the feline gazes intently, its eyes widen with astonishment as the canine gracefully soars through the air, revealing its remarkable nimbleness and effortlessly soaring over the hurdle obstructing its way.
    6. As the feline fixates its gaze, its eyes widen in sheer astonishment, witnessing the canine’s elegant flight through the air. The dog’s exceptional agility becomes evident as it effortlessly clears the obstacle blocking its path, leaving the feline in awe.
    7. As the feline becomes fixated, its eyes widen in sheer astonishment, observing the canine’s graceful mid-air glide. The dog’s exceptional agility is unmistakable as it effortlessly overcomes the obstacle obstructing its way, leaving the feline in a state of awe.
    8. In a captivating spectacle, the feline finds itself utterly transfixed, its eyes widening in sheer astonishment. It stands witness to the canine’s mesmerizing mid-air glide, and its exceptional agility is unmistakable. With effortless grace, the dog effortlessly conquers the obstacle that had impeded its path, thus leaving the feline in a state of awe.
    9. The feline becomes completely engrossed, as if under a spell, when it witnesses the captivating spectacle before its eyes. The feline’s eyes widen in sheer astonishment, unable to look away from the canine’s mesmerizing mid-air glide. The canine’s exceptional agility is unmistakable as it effortlessly conquers the obstacle that once hindered its path. With graceful and effortless movements, the dog leaves the feline in a state of awe.
    10. The feline is instantly captivated, almost entranced, by the mesmerizing spectacle unfolding before its eyes. Its eyes widen in sheer astonishment as it becomes utterly unable to tear its gaze away from the canine’s enchanting mid-air glide. The canine’s impressive agility is undeniable, effortlessly overcoming the obstacle that once obstructed its path. With graceful and seemingly effortless movements, the dog leaves the feline in a state of pure awe.

    And this is using the same AI to shorten number 10 to a 10 letter phrase

    Feline captivated, eyes widened, unable to tear gaze from enchanting canine.

    AI, in its freely available form, is good at expanding non deterministic interpretation. It can add complexity in the correct syntax, but any meaning that it adds is both accidental and incidental.

    It is much worse at determining the essence of a concept.

    From The dog jumps over the cat to Feline captivated, eyes widened, unable to tear gaze from enchanting canine.

    It can add complexity and the meaning is generally maintained but in removing complexity the meaning can be lost.

    As a process it appears to be creative but it is really entropic as the created material is just the same material spread over a wider area. The total energy level is maintained but not increased.

    Is AI going to take your job?

    Well, it depends. AI cannot be trusted to tell the truth. If you are a journalist or a politician, I would start to get nervous about future employment prospects.

    For the rest of us – as long as you are taking real responsibility for your work then AI will not be replacing you. AI should be not making decisions now, and even if it improves it cannot be held accountable or liable.

    This is similar to the problem with autonomous flight. It may be theoretically possible and it may even be safer. However mistakes will happen and people will get hurt and killed. Without pilot error as a liability shield the software OEM becomes responsible for all negative outcomes. This does not end well for the OEM.

    If we allow an AI to make decisions the OEM of the AI will face a similar problem.

    AI may eventually take the role of a virtual team member, but until we revise our societal legal accountability standards the role of AI in decision making will be advisory only.

    What does this mean?

    For now we have a useful tool we have not yet learned how to fully exploit so spend time getting familiar with the tool. I would advise not reading too much into the implications for now. It is very hard to predict anything – especially the future.