OpenAI’s ChatGPT was only launched five months ago and can already deliver detailed answers to queries, write specialised content, help children with their homework and create working code.
Together with other generative artificial intelligences – producing text, images, videos and voice – these technologies have the power to transform businesses.
But where will it have the greatest impact – and what risks does it pose?
There are already multiple applications and use cases for ChatGPT in particular which can help businesses.
“AI-powered chat systems which replicate human conversation have the potential to disrupt traditional business by automating tasks such as customer service,” Albert Meige, associate director at consulting giant Arthur D. Little, tells TechBlast.
“AI can also now assist humans in the writing of sophisticated emails, articles, letters, poems and songs. Content can be generated using simple prompts or through more sophisticated interactions with the bot to get more convincing results.”
It also has the capability of reformulating, improving, expanding or synthetising existing human-generated content.
“This has numerous potential uses, such as creating more accurate summaries of long documents, generating new ideas based on existing content, or even automating the creation of marketing copy,” says Meige.
ChatGPT can also write computer code, which involves providing it with the desired functionality and then allowing it to generate code that meets those specifications, as well as debug it.
The idea that only humans can be truly creative is also being tested, argues Meige. “Artists and creators can now leverage AI in an interactive manner to enhance their own creativity and rapidly test ideas,” he says.
“GPT-3 can also fill gaps in prompts by using its knowledge and understanding of language and context to generate relevant and appropriate responses. This means it can provide coherent and sensible responses to prompts that may be incomplete or ambiguous.”
A recent scientific study found that the GPT-3 model has a strong capacity for abstract pattern induction and can find zero-shot solutions to a broad range of analogy problems, says Meige.
Meanwhile, image generation AI tools such as MidJourney or Dall-E have acquired the ability to analyse and understand visual trends through training on a massive corpus of data.
“This includes the look and feel of a specific time period in history. It can therefore recognise patterns and make predictions about visual trends,” says Meige.
So what are the limitations of ChatGPT?
“An application such as ChatGPT generates content that is not always reliable,” he says. “This can be for a variety of reasons, including not understanding the question properly; or factual errors, either for unexplained reasons or because its training data is from before 2021.
“When ChatGPT does not find the answer in its base, it may try to guess, returning plausible, but incorrect, answers without letting the user know that it is simply guessing.”
He also outlines ethical, moral and social concerns. “As with other AI models, it can face algorithmic biases, due to the data used to train the algorithm being biased, leading to inaccurate or unfair results.
“AI-generated content is already starting to flood the internet. This can be used for nefarious purposes, including the spread of misinformation, the manipulation of public opinion, and the harassment or defamation of individuals.”
Cybersecurity is also a concern. “While ChatGPT will initially refuse requests to help commit crimes or cyberattacks, it is relatively easy to bypass its safeguards, such as by pretending the output is for a work of fiction,” says Meige.
“2022 was the beginning of a new era, delivering transformative benefits to businesses and humans. However, it is still early in its development, and organisations need to understand and mitigate any risks when adopting ChatGPT and other AI-based systems.”