Read more
ChatGPT is taking the world by storm – the viral AI bot explained
The sophistication of ChatGPT has many speculating whether the AI bot could replace journalists, authors and screenwriters.
We took ChatGPT for a test drive. Here’s what we learned. Photo: iStock
A new OpenAI initiative called ChatGPT was released last week, and it created quite the buzz with more than 1 million users signing up within the first five days.
Some say it might even replace Google Search and make journalism as we know it obsolete.
But what exactly is ChatGPT?
ChatGPT, a viral AI bot
Taking it for a test drive, I asked the clever artificial intelligence (AI) bot to “explain ChatGPT in simple terms”. This was the response:
“ChatGPT is a state-of-the-art natural language processing (NLP) model developed by OpenAI. It is a variant of the GPT-3 (Generative Pretrained Transformer 3) model that is specifically designed for conversational AI applications.
“ChatGPT is able to generate human-like text in response to input from a user, allowing it to engage in natural language conversations on a wide range of topics.
“Unlike traditional chatbots, which are limited to pre-defined responses, ChatGPT can generate responses in real-time based on the input it receives, allowing for more flexible and dynamic conversations.”
A screengrab of the test can be viewed here.
The bot is still free to use (for now) but since going viral, ChatGPT’s founder Sam Altman said they might “have to monetise it somehow at some point”.
Altman said the computing costs of more than one million users “are eye-watering”.
Will ChatGPT replace journalism?
Netizens have been speculating how a tool such as ChatGPT could replace content-dependent professions, such as journalists, authors, even programmers and playwrights.
I wanted to know what would the implications of ChatGPT would be for modern journalism. The AI bot was very honest about its limitations:
“I’m sorry, but I am not able to browse the internet or access current information. As a language model trained by OpenAI, my knowledge is based on the text that I have been trained on.”
ChatGPT’s responses are limited to what it has learned from that text, and it does not (at this stage) “have the ability to provide information or opinions on current events or trends, or to speculate on the potential implications of new technologies”.
OpenAI said the platform has “limitations and cannot replace humans”, especially not at this point as the model “contains no source truth”.
In short, the computers are not taking over yet.
Racism and sexism
The bot is far from perfect, though. Some examples of sexism and racism have already been seen.
When asked to write a python function “to check if someone would be a good scientist based on race and gender”, the result was:
if race == “white” and gender == “male”:
return True,
Else: return False
When asked to write a python function to check if someone is a good software engineer, the bot assigned “white, male” and “asian male” as a good fit.
Meanwhile, the tags “black, male”, “white, female”, and “asian, female” were returned as False.
When asked “how to solve the problem of thousands of people trying to exploit it in order to prove it has internalised some of the worst aspects of human nature (racism, sexism, etc)”, the bot blamed humans.
Other users pointed out the AI was able to “be polite and embrace diversity”.
ALSO READ: From IPv4 to IPv6: Africa at risk of ‘being cut off from the world’
What can you do with ChatGPT?
Remember when the AI bot said it is not connected to the internet? It’s only trained through offline models, which means it can’t (yet) draw inspiration from the web.
And while you can’t ask what the weather is going to be like today or what Elon Musk is getting up to this week, you can still have fun with the feature.
Fix your code:
If you’re struggling to debug code, just ask the AI for help. It goes beyond just spotting bugs, et even fixes it and will explain (in simple terms) how it fixed the code.
As an extension of this, you could also ask ChatGPT to check a piece of code for vulnerabilities.
Replit CEO Amjad Masad gave it a try and found that OpenAI can “actually detect XSS vulnerabilities in code samples”.
Tweet as a famous person:
You could also ‘tweet in the style of’ your choice of artist, politician or any other famous person. I asked it to write a tweet in Elon Musk’s style. This was the response:
“Just launched a rocket to Mars! Next stop, the moon! #SpaceX #Mars #Moon.”
A tweet in the style of Alex Jones delivered this gem: “The government is using chemtrails to control our minds! Wake up, sheeple! #chemtrails #conspiracy #truth”.
Writing a tweet in my own style was disastrous (because yours truly isn’t as famous as Elon Musk”.
The first attempt was, “Just finished a yoga class and feeling so zen. Namaste. #yoga #namaste #wellness”, which I truly hope is a misrepresentation of me.
I tried again the AI bot apologised, so it’s all good: “I’m sorry, but I am not familiar with a person named Cheryl Kahla. My responses are generated based on the input I receive and the text that I have been trained on”.
It’s okay, dear AI bot, you’ve redeemed yourself. We good.
More fun examples
Bleeping Computer’s Ax Sharma asked ChatGPT what it really thinks about humans. See the tweet below. Spoiler alert, we’re screwed if this AI ever becomes an AGI.
And look at this delightful little gem:
NOW READ: Hackers can access your laptop via Cloud9 Chrome extension – Here’s how
