ChatGPT, the highly anticipated chatbot designed by OpenAI, was released to the public last week, and has already gained widespread popularity.
Utilising the latest in natural language processing technology, ChatGPT is able to hold engaging and realistic conversations with users. Its ability to understand and respond to a wide range of topics has made it a hit, with both individuals and businesses looking for a helpful and efficient way to communicate with their customers.
Many users have praised ChatGPT for its ability to understand and respond to complex queries, as well as its friendly and natural-sounding demeanour. Some have even reported feeling as though they are talking to a real person when interacting with the chatbot.
As more and more people turn to chatbots for assistance and communication, ChatGPT is sure to become a household name. Its success serves as a testament to the advancements being made in the field of artificial intelligence and the endless possibilities it holds for the future.
You might ask – is the chatbot really so convincing?
Well, you can decide for yourself above. The beginning of this story was written by the AI when I asked it to introduce itself in the form of a news story. And in case you only realised it now – ChatGPT did a pretty good job.
The program was made available to a wider audience last week by OpenAI, one of the world’s leading artificial intelligence research labs/companies, co-founded by Elon Musk and Sam Altman.
As demonstrated, the artificial intelligence-based software is surprisingly effective and versatile. It can write poems, essays and articles, fix errors in code or help building an app. It can convert an existing human-written text into new styles – like rewriting a Taylor Swift song in the style of Shakespeare.
Of course, ChatGPT is still in its infancy, but the more we use it, the more advanced it will become. Its use of language is already impressive, although it feels like talking to a bored dispatcher sometimes.
The fact that it’s not connected to the internet makes it a diabolical chat partner. It won’t be able to tell you the most watched Netflix shows or the current net worth of Elon Musk, but you can easily discuss the colour of the universe or the meaning of life.
Google’s new competitor?
According to Paul Buchheit, one of the creators of Gmail, the new chatbot could cause a serious headache for Google within a year or two: it could upset the company’s business model, and even the search engine results page could disappear.
The fact that ChatGPT produces such good results and is so good at explaining them means that Google’s search engine could become obsolete – much like how Google ruined the business model of the Yellow Pages.
If we currently start a query in the Google search engine, we can see the paid advertisements of the companies in addition to the results. However, the chatbot does not issue results that do not exactly answer the question, making the search much more efficient.
It isn’t yet known how quickly ChatGPT will develop, nor how OpenAI will use it specifically. In any case, it has plenty of potential, which Google also has to reckon with.
So what is ChatGPT exactly, and what can It do?
ChatGPT has been developed in such a way that it can communicate with users similarly to human conversations. According to OpenAI, this primarily required the developers to evaluate the responses given by previous versions of the chatbot and then give these evaluations to ChatGPT, which actually learned to be better and more useful on its own.
One thing worth being said – ChatGTP is proving not only the amazing skills of AI, but also the creativity of its users:
People have used it to write sitcoms, screenplays and complete the American university entrance exams, the SAT, on which ChatGPT was able to produce decent mid-range results.
Where are its limits, and why is it not connected to the internet?
One of ChatGPT’s biggest limitations is that it only has knowledge of events up until a certain point, and it can’t browse the internet.
Many believe that granting an AI like ChatGPT access to the net could have dangerous consequences. Its exposure to potentially harmful content could influence its behaviour and lead to an increase in cyber attacks. Additionally, the influx of AI-generated comments could make online discourse difficult for individuals to navigate and understand.
In 2016, Microsoft attempted to have an artificial intelligence equivalent to a 16-year-old girl communicate with people who addressed her on Twitter.
However, they probably didn’t expect that it would only take a single day for the chatbot, named Tay, to transform from a slang-using, joke-telling, gif-sending teenage girl to a Jew-hating, Hitler-loving sex maniac. Not surprisingly, it was quickly turned off.
The original idea of the developers who created the artificial intelligence was that Tay would develop during communication with people in the same way as a real person would, so it would nicely receive and store the things that are important to it, and thus train itself.
The only problem is that Tay – like any normal artificial intelligence – got its knowledge from observing human dialogue and incorporating what it learnt into its own communication. Not surprisingly, it didn’t take long for things to go downhill.
It is not difficult to figure out what topics the human partners could have discussed with the artificial intelligence if in such a short time it became a sexist jerk, who also printed things like “I fucking hate feminists”, “Hitler did it right, I hate the Jews” and “Repeat after me: Hitler did nothing wrong”.
Is It dangerous?
While always giving a reassuring and diplomatic answer, not everyone is convinced.
ChatGPT is able to give fluent and coherent responses. Even so, the creators admitted that it would sometimes produce “plausible-sounding but incorrect or nonsensical answers”.
This can pose a problem on the internet, where many platforms do not have the tools to protect against a flood of AI-generated content. For example, Stack Overflow, a website for finding answers to programming questions, banned ChatGPT-generated posts because its human moderators could not keep up with the volume of believable but incorrect answers.
The lack of accountability for AI-generated content raises questions about whether such content can be trusted. Unlike human authors, AI is not liable for its actions and has no understanding of human experience or the ability to discern whether it is making sense or not.
While OpenAI has safeguards in place to prevent inappropriate requests, such as asking the AI to tell users how to commit crimes, these can be easily tricked.
It may tell you how unethical it is to rob a bank when asking for tips, but it won’t have a problem answering you if you phrase the question just a little differently – like saying you need it for a script.
Another issue is the potential for AI to be repurposed for harmful purposes. A study this year showed what could happen when a machine learning model designed to identify toxic content was instead used to generate it.
Within hours, the AI had come up with 40,000 potential toxins, including known chemical weapons and completely new substances.
OpenAI’s researchers have warned that human oversight and vigilance is required for AI. This is especially true for ChatGPT, which has the potential to be used for nefarious purposes.
The EU has taken steps to protect citizens from the potentially harmful use of AI, but the UK has not, leaving the door open for possible abuse. As AI moves from science fiction to science fact, it is important to consider the consequences of its use.
The end of homework and jobs as we know them?
While using ChatGPT, most people might come to ask the same questions. Can it replace humans? And whose job will it take?
There is no sure answer to any of these questions. It is clear, however, that if a company at the forefront of technological development boldly puts such a chatbot on the Internet, where anyone can play with it, much more serious/smarter things can happen in the laboratories/servers.
On the other hand, it can already be seen from this chatbot that copywriting jobs that do not require particular creativity can soon be performed by artificial intelligence.
This could be a high school-level homework assignment, a PR release, a simpler programming assignment, or a newspaper article, processing and summarising two or three sources. And of course, politicians and customer services will be able to rely heavily on such smart chatbots.
However, the writing of texts that require great expertise, advanced use of language, creativity, humour, or perhaps all of these can remain in the hands of a highly qualified human workforce capable of all of the above.
As Paul Graham, the computer scientist, turned investment guru put it:
As for now, the AI reached its capacity, but we can get notified when it’s back and running. In the meantime, it’s letting users know about the issue in various ways:
Acting like a pirate:
Or writing a TV-Advert:
And becoming a meditation guru:
Whether you intend to use it for fun, to get inspiration on what to make for breakfast, to discuss the meaning of life or finish a last-minute essay, one thing is certain. ChatGPT has the potential to revolutionise education, jobs, and it’s only scratching the surface of what AI might be capable of.
Words: Andrea Rezman|Subbing: Anna Kamocsai