OpenAI has launched the latest incarnation of its wildly successful online chatbot ChatGPT.
Dubbed GPT-4 (ChatGPT is powered by GPT-3.5), the headline feature of the new model is its ability to recognise and explain images.
For example, the chatbot could come up with recipes based on a picture of the inside of a fridge, so it could work wonders for my uninspired cooking habits (if only it had been around for the great banana bread boom of 2020).
But it also improves on its predecessor’s ability to interpret text and write its own, providing more complex answers, channelling different styles and voices, and is capable of processing thousands more words at a time.
How is it improved?
For now, only paying subscribers to OpenAI’s ChatGPT Plus service can access it – but it’s already clear that GPT-4 bolsters the potential of generative AI.
It massively outperforms ChatGPT in established exams, for example.
“It seems truly more capable at not just writing text, but to do a better job at generating more elaborate and specific prose and have a deeper understanding of a problem,” Dr Peter Van der Putten, professor of AI at Leiden University and director of AI at Pegasystems, told Sky News.
“From a marketing point of view, it’s a bit harder to market. But this ability to be more specific, less bland, show real understanding and explain its reasoning, it will increase the trust people have in it.”
Indeed, while ChatGPT can often come across as bland and verbose in its responses at best, at worst it can spout outright falsehoods – yet still with supreme confidence.
OpenAI says GPT-4 may still be prone to such issues, though insists safeguards have been improved.
GPT-4 can also go beyond the fixed tone of ChatGPT, with users able to ask for responses in specific styles – like the Shakespearean pirate seen above.
It could prove revelatory for the model to encroach on the arts, writing stories, poetry, and scripts in the voice of established creators, or maybe even help make you sound more interesting on dating apps.
Read more:
Is ChatGPT the ultimate homework cheat?
How AI could change how we search the web
Is it coming for more jobs?
In an online demo, OpenAI president Greg Brockman showed how GPT-4 could quickly come up with the proper income tax deduction after being fed lots of tax code – something he couldn’t work out himself.
Beyond streamlining day-to-day tasks like tax returns and baking, experts say the impact on jobs, research, and academia could prove far more profound.
Dr Andrew Rogoyski, of the Institute for People-Centred AI at the University of Surrey, said OpenAI’s testing showed “big increases in performance in sciences, maths, and economics”.
The number of companies already leveraging the technology is testament to its potential – Microsoft’s new Bing search engine, payments platform Stripe, and language learning app Duolingo all use it.
Click to subscribe to the Sky News Daily wherever you get your podcasts
And while ChatGPT had already been cited as a threat to jobs like customer service and computer coders, OpenAI appears confident that GPT-4 could come for even more roles.
Its launch paper states: “We expect GPT-4 to impact even jobs that have historically required years of experience and education, such as legal services.”
We’ve already seen an AI take questions in Westminster, so maybe a courtroom isn’t far behind.
Read more:
ChatGPT helps write Sky News article
Google reveals highly-anticipated AI chatbot
What comes next?
Much of the speculation ahead of GPT-4’s unveiling centred on eye-catching new features like generating video.
What we’ve ended up with shows advances in public-facing AI are likely to happen more gradually, given the risk of releasing models when they aren’t ready (Microsoft’s 2016 Tay bot says hi, or probably something offensive).
Sridhar Ramaswamy, co-founder of search engine Neeva, which has its own GPT-style AI, told Sky News OpenAI’s upgraded model is a “natural but still important evolution”.
“Big change is always like that – lots of incremental changes adding up to something monumental,” he said.
But experts say it’s more evolution than revolution, and OpenAI admits it’s no replacement for humans just yet.
“These models still don’t have a fundamental understanding of ‘truthful’ vs ‘falsehoods’,” said Mr Ramaswamy.
“But GPT-4 is less than three months after ChatGPT. It certainly feels like there is huge momentum in the space.”