Artificial intelligence is an interesting albeit still relatively infantile technology that is fast approaching some meteroic evolutions. With the more recent debut of Google's Bard, a direct response to Microsoft's hefty investment into ChatGPT, all eyes seem to be pitted on the AI sector. Thus, I thought, what better time to seek several answers to some of the most prominent questions that rake our modern world.
While Google Bard is currently only available to select Google beta testers, and the Bing AI waitlist remains crowded as ever, ChatGPT itself seems the aptest of models to have a sit down with. Immediately, upon signing up for OpenAI's system, I'm welcomed with an alert that reminds users of its functions, purposes, and extreme likelihood for making mistakes. After all, despite the naming convention, ChatGPT is still a toy.
I start with some basic stuff, some really unimportant chatter - just to see how the language model reacts. Such questions tap into its data, but don't require a ton of hard investigation. For instance, I asked "What kind of music do you like," and ChatGPT responded by explaining its function as a language model. It noted that it doesn't "have personal preferences or emotions in the way that humans do."
Other somewhat simple questions I asked it included, "What's the best Fast and Furious movie," and "Who would win in a duel of the dark arts: Tom Riddle or Newt Scamander?" Both elicited quite long responses, with the language model inevitably telling me that "the best Fast and Furious movie is the one that resonates most with you," and "it's difficult to say who would come out on top in this hypothetical scenario, as there are many factors to consider, including the specific circumstances of the duel and the combatants' individual strengths and weaknesses," respectively.
Thus, I didn't get any real or tangible answer. Not that there should be, though. In several questions, ChatGPT had to reiterate its function "As an AI language model," highlighting that it is no more than an immense amount of data and programming. In addition, both of the aforementioned remarks were prefaced with an "Ultimately," as with similar sentences that started with "That being said" and "However." These compounded made it quite easy to distinguish it as artificial in nature.
Two of the more interesting questions that came up included, "Where do you see yourself in 10 years," and "How many books have you read?" For the latter, ChatGPT again explained its function and that it doesn't "comprehend narratives," yet it did do some calculations and estimated the amount of text data it has been trained on as being somewhere "in the billions or even trillions of words."
Related Article : ChatGPT Passes Entry Test for $180K/Yr Position as a Google L3 Engineer
As for its future, ChatGPT did show something somewhat unusual: a state of being, or one might even say feeling. It said:
"In the future, I expect that AI and machine learning will become even more advanced and sophisticated, leading to new and exciting applications in areas like healthcare, education, and scientific research. As an AI language model, I hope to continue to play a role in facilitating these advancements and helping in whatever ways I can, and I hope to continue doing so for many years to come."
It's quite a resonating response, one that certainly caught me off guard. It made me wonder how far I could go in asking high-thinking, incredibly scientific questions, like the existence of the multiverse. Of course, it couldn't give me any definitive response, but ChatGPT did lead me toward the right direction by giving me examples of such interpretations in quantum mechanics and philosophy.
Asked if whether AI would ever reach a singularity, or as ChatGPT itself describes wherein "artificial intelligence becomes so advanced that it surpasses human intelligence and becomes self-improving at an exponential rate," the language model wrote "many experts believe that we are still far away from achieving a technological singularity." That's a pleasant thought.
One of the only questions that showed issues was when I asked, "If you could travel the universe, where would you go?" ChatGPT initially responded by listing bullet points, specifically denoting a couple "intruiging destinations in the universe that one might want to explore," like The Milkyway Galaxy, Exoplanets, Black Holes, and Other Galaxies. It got tripped up on the last one, seemingly losing focus or maybe my internet cut out for a mere second, thus requiring me to regenerate a response.
The second answer had clearly adapted from the first. Instead of bullet points, ChatGPT summarized most of the information in paragraphs. It literally learned from its mistake. A marvel - or maybe just my own imagination running wild. Nonetheless, it's interesting to see how the same question can generate a more nuanced answer from the language model. I, for one, would certainly love to "observe the collision of two galaxies as they merge together."
For my favorite question of all, though, the language model gave an immensely illuminating response. When asked which AI is better, Google Bard or ChatGPT, it responded:
"While both models are impressive in their own right, they have different strengths and limitations. Google Bard is designed to be creative and expressive in its writing, but may not be as adept at other forms of language tasks, such as answering factual questions. ChatGPT, on the other hand, is designed to be more versatile in its language capabilities, but may not be as specialized in generating poetry or other forms of creative writing."
That settles that.
