That was quick. Just three months after OpenAI’s ChatGPT was launched – causing ructions in many industries – it has been upgraded. The underlying software is known as GPT-4 and it has leapt forward significantly in what it can do. The new version of ChatGPT will handle not just text, but images too.
There has been a flurry of artificial intelligence (AI) offerings in the past few weeks as other companies have launched their services. Microsoft, which is a large investor in OpenAI, showed off its new Microsoft 586 Copilot software for its major app, a significant upgrade to the AI automation it previously offered.
“We believe this next generation of AI will unlock a new wave of productivity growth,” said Microsoft CEO Satya Nadella.
Google launched its own service, called Bard – which made a mistake when it was revealed to the public earlier this year and which shaved $100bn off the search giant’s share price. No wonder Alphabet CEO Sundar Pichai was more measured in his remarks.
“Even after all this progress, we’re still in the early stages of a long AI journey,” he wrote in a staff memo. “As more people start to use Bard and test its capabilities, they’ll surprise us. Things will go wrong.”
At the heart of this sudden surge of offerings is what is called a large language model (LLM). These LLMs are giant databases of works of literature, books, newspapers and content scraped from the internet. What OpenAI and others have done is create a complex set of rules and understandings of how language is used – including by a range of authors – and how to replicate it. Sometimes with mixed success.
When ChatGPT, Copilot or Bard don’t know something, they tend to “hallucinate” the answer – using the word given by the software makers themselves.
Generative AI, as it is also known, is a truly remarkable breakthrough in technology, powered by these large language models. Many commentators are debating how much of a sea change it is. Does it compare to the invention of the computer or the internet itself? Is it equivalent to the launch of the iPhone, which heralded the smartphone era we currently live in?
Perhaps, with ChatGPT’s ability now to use so-called plugins, is it akin to the launch of the app store, which opened up the power of smartphones to other software makers and launched the mobile app phenomenon?
Whatever analogy you settle on, it’s a big deal. Large language models and generative AI will change computing and the way we work. We will see how this happens as it unfolds over the next few years.
It will certainly involve a new industry for academia and education that will try to spot work created by generative AI services. Many US universities have already reported catching students using ChatGPT to write essays and reports.
I’ve had conversations with journalists who are experimenting with it to write headlines or improve them. The trick, as always, is how specific and precise your instructions or requests are.
Already there are warnings about the potential hazards of AI, beyond just university students faking their essays.
“Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable,” reads an open letter from the non-profit Future of Life Institute, which has 1,000 signatories. These include AI researchers and scholars, as well as Elon Musk, who was a co-founder of OpenAI.
After years of slow, almost invisible, progress, the AI industry is suddenly bursting with innovation and interest. Buckle up.