ChatGPT - Holding A Mirror To Us.

"To say ChatGPT is intelligent is to say that planes are birds. Both fly, but by different means. Artificial flying machines don't need to adopt the modus operandi of birds, and ChatGPT does not think like humans."

"Open the Pod bay doors, please, HAL," pleads Dave Bowman, stranded outside his spacecraft. When HAL, the computer, responds with, "I'm sorry, Dave, I'm afraid I can't do that", we know that the machine has humans in its sights.

​HAL, like Frankenstein, turns on its creator — breaking Asimov's first rule of robotics. 

Over the years, this scenario played out in many science fiction films — the omnipotent Skynet artificial neural network in Terminator gains sentience, then immediately attacks. Likewise, the short-life replicants of Blade Runner race against the clock to prevent the erasure of their memories 'like tears in the rain'. 

And if you follow the media hype around ChatGPT and other so-called 'artificial intelligence' systems, we could be at the next inflexion point in our relationship with machines. Then, depending on who you believe, we face the mass culling of jobs and roles or another evolution with uncertain outcomes.

Still, as usual, the media are ahead of themselves to drum up clicks with tales of pending disaster. They've summoned a warning from Professor Stephen Hawkins.

"While primitive forms of artificial intelligence developed so far have proved very useful, I fear the consequences of creating something that can match or surpass humans," Hawking wrote. "Humans, who are limited by slow biological evolution, couldn't compete and would be superseded."

It may help to step back for some calmer reflection. In truth, AI is a misnomer. The name hides something more mundane but still fascinating about computing power and its impact on us.

ChatGPT and similar systems don't think, emote or produce anything unique. Yet, their speed and agility in answering our questions or creating essays give the impression of intelligence, while in reality, they draw upon what we have given them. We are their dataset, or rather the knowledge we’ve acquired, and then uploaded on the internet.

Someone observed that, "To say ChatGPT is intelligent is to say that planes are birds. Both fly, but by different means. Artificial flying machines don't need to adopt the modus operandi of birds, and ChatGPT does not think like humans."

Hence when you ask ChatGPT to produce an essay on "The merits of eating fruit," it will draw together evidence from the internet to craft a response. Moreover, it will use jargon to construct the essay, adding to the perception that intelligence is at work.  

Here is a link to essays produced by ChatGPT. Could you spot them as machine generated? 
Above all, ChatGPT does no reasoning. The process at work is collating, sorting and crafting a response.

The algorithms add nothing new or original; thus, ChatGPT provides no new insights. So what you get is already known, although delivered at exceptional speed.

Nonetheless, ChatGPT and its companions raise many issues around education, the learning process and the authenticity of student output. If students can harness ChatGPT to do their assignments without detection, how much actual learning takes place?

Thus, I see universities are already moving to ban the use of AI, although how this is enforced remains unclear. 

In the future, I expect we may see a return to traditional handwritten exams to test what students have absorbed. But, in the meantime, with the outputs of ChatGPT challenging to detect, students will use it. 

Don’t be surprised at that because ChatGPT already passed law exams. Those expensive lawyers better watch out.

Perhaps AI will have the most significant impact in the creative domain. Throughout history, artists copied and paid homage to their mentors. This discussion on the origins of Stars Wars music well proves this point.

In the arts, authenticity is always a murky issue. Nevertheless, arguably AI can evolve something new in the arts, although founded in the past; because the creative possibilities based on mimicking are many.

We are certainly on the cusp of a moment similar to the arrival of the internet. ChatGPT has given us insights into new possibilities, most of which are just coming into view. Here a self-proclaimed nerd of the first tranche of the internet discusses how ChatGPT resolved a coding issue he struggled with.

​The example is apt because the algorithms made mistakes based on wrong inputs by the human nerd. Amazingly, when asked, ChatGPT explained what it had done, and this led to a resolution.   

Doubtless these systems are useful tools that we will learn to use to our advantage. Yet, given that these systems are reflections of us, they manifest all the failings of humans. Our errors, prejudices, misreadings and other weaknesses lay captured in the work of the algorithms that cannot escape the context we’ve given them. 

This leads back to HAL. He went off the rails because humans asked him to process information accurately and lie simultaneously. These conflicting demands drove him to kill.

When genuine AI arrives, it will still be a 'self-portrait' of humankind, flaws and all, except at operating speeds we can't comprehend. Therein lies the risk.