If you’ve been paying attention on the internet, you’ll probably have come across new buzz surrounding an AI chatbot that seems to be incredibly intelligent — perhaps too intelligent, even — to the point where people are wondering just what the future holds in terms of machine learning and potential computer sentience.
It’s called ChatGPT, and is now freely available (for a limited time) for everyone to test out, with the interface mimicking a simple and plain chat screen where you can type in your queries and then expect a response from the bot.
Developed by for-profit research lab OpenAI (a pretty huge name in the game), ChatGPT essentially is a showcase of just how powerful an AI can be, and was designed in a way that would mirror online conversations between humans.
You can pose it any question or statement, and it’ll reply in such a conversational way that you might even think you’re in an exchange with an actual human, complete with sensible arguments and logic.
Admittedly, this appears far superior to the chatbots from years ago that claimed to be able to mimic human interaction (remember SmarterChild from the MSN Messenger and AOL days?).
You can ask it questions about life or your job, command it to write lyrics to an original song, tell jokes, or even ask for it to speak with you in an angry tone.
Here are some incredible answers ChatGPT came up with when I posed it some queries:
1. Write lyrics to a song.
2. Come up with an alien language.
3. Write a poem about “pineapple on pizza”.
You can even ask it to refine the result in a particular direction.
4. Pretend to be pissed off at me.
5. Write an abstract for a film.
It’s not flawless. Not yet, anyway.
While ChatGPT may seem miraculously smart, it really doesn’t actually know much of anything. Instead, all of the answers generated by the bot are derived from information available on the internet, and refined through human input, but with some limits, of course.
For example, when I asked it about the number of individuals under the poverty line globally, the bot returned an apology that explained how it was unable to provide such statistics due to its imposed information cut-off.
In some other cases, some people have also reported the chatbot providing inaccurate answers to some questions, which OpenAI has warned as something that may happen from time to time.
The possibilities may be concerning.
So, in a nutshell, ChatGPT really isn’t the perfect all-encompassing intelligence that it may seem to be, but still there are plenty of things to think about as far as how such a technology could eventually progress — perhaps even to the point where it can replace humans in certain tasks.
Already, there have been stories of programmers using ChatGPT to write whole batches of functional computer code, while some in academic circles have voiced their concerns about the bot being able to write complete essays and enabling students to cheat in their assignments.
At the moment, these generated answers from the bot are far from the level where it can competently produce entire lines of functional and efficient code, nor can it write full-on thesis papers convincing enough to full high-ranking academicians.
But judging by the way that it can produce such sophisticated answers right now, there’s no reason to discount the distinct possibility of it eventually being good enough to replace humans in jobs that require thought and some nuance.
Incidentally, when I asked if AI could eventually replace writers like me in the future, ChatGPT responded with a lengthy spiel on how human writers could never be replaced. Not exactly reassuring considering the amount of nuance and objective clarity on display here.
If you want, you can now try out ChatGPT for yourself here completely for free, although its developers OpenAI have stated that it would eventually seek to monetize the tech in the future considering the high costs involved in developing and maintaining such a creation.
Hey there! Here’s what others are currently reading:
Cover image sourced from Forbes.