Digital Marilyn Monroe was created by Soul Machines software development in San Francisco. Photo / Soul Machines
ChatGPT is a generative artificial intelligence (AI) chatbot. You can ask it a question or give it instructions in everyday language and it gives human-like responses.
Like humans, ChatGPT can, and often does, make mistakes. Yet it can cut through the fog to deliver surprising insights in less time than it takes to muster a team meeting.
Within weeks of its launch, ChatGPT had 100 million users. That’s faster than any product in history. Almost immediately, the world’s leading technology companies shifted their focus to AI. It took less than two years for the entire tech sector to change direction.
ChatGPT is not alone. It has direct rivals from little-known companies like Anthropic and well-known ones like Google. The other tech giants, including Meta Platforms, are also hard on OpenAI’s heels. Meta, Facebook’s parent company, refocused its entire business and spent close to US$50 billion attempting to build the virtual reality metaverse before, in effect, writing off much of that investment as it pivoted to AI.
Microsoft struck a deal with OpenAI to offer CoPilot to its customers. Apple offers a version of ChatGPT on its latest iPhones and computers.
Laptop brands like HP added AI hardware to help devices process chatbot requests faster and more efficiently.
The boom is everywhere in the tech sector with some unexpected consequences.
Last month Nvidia overtook Apple to become the world’s largest company by market capitalisation.
Its value peaked at US$3.43 trillion.
Nvidia started life making the chips needed to play realistic-looking video games. Around a decade ago researchers found they could use these graphical processing units (GPUs) to train AI systems hundreds of times faster than by using traditional chips. In the two years after ChatGPT arrived, Nvidia’s market value grew 800%.
AI Forum report
New Zealand companies wasted little time getting on board the AI boom.
In September, the AI Forum of New Zealand published a report saying two-thirds (67%) of organisations responding to its survey have adopted AI. Almost all reported increased worker efficiency and “positive financial outcomes”.
Despite fears, the report found there has been relatively little job displacement. Only 8% experienced it. It does warn that some organisations faced a significant initial investment setting up for AI, while most found it manageable.
The AI Forum’s upbeat outlook doesn’t square with the international experience. In November, The Economist reported that only 5% of American businesses say they use AI in their products and services while most companies don’t know what to do with it. Moreover, the same report points out that, to date, few AI startups are profitable yet, although it is still early days.
Elsewhere Gartner, the research company, echoes what is showing up in surveys worldwide. There are benefits, but “the technology has yet to deliver on its anticipated business value for most organisations”.
Generative AI works by drawing on a vast amount of data including text from books, articles, websites and other written sources. The software analyses patterns and structures in the source material and then responds using natural-sounding language.
Other generative AI tools use similar databases to create new images and write music and computer programming code.
They all need vast amounts of data and a significant amount of processing. This approach is controversial: writers, artists and musicians are understandably upset when AI engines use their past output in ways that could put them out of a job.
NZ’s Soul Machines debuts digital humans
Auckland-based Soul Machines uses generative AI and other AI technologies such as natural language processing (NLP), machine learning and neural networks to create what it calls digital humans. These can be based on celebrities but are more commonly used by companies as realistic-looking artificial customer representatives.
The company makes what it calls “digital humans”. These are avatars that are often used by companies to handle customer inquiries on video calls. Soul Machines’ digital humans can interpret human emotions during a conversation by analysing facial expressions, tone of voice and language context.
This gives them the information they need to provide appropriate responses in real time.
They look more natural than other chatbots to the point where interaction with a computer feels more personal.
Soul Machines has a sideline making digital versions of celebrities who can interact with far more people than the originals ever could.
This kind of application is an Auckland specialisation. Two other local companies, UneeQ and Ambit AI, also offer customer service chatbots with human characteristics.
Auckland ophthalmologist Dr David Squirrell is co-founder of Toku Eyes along with Dr Ehsan Vaghefi. It uses a related AI technology to help screen patients for diabetic retinopathy.
This is a complication of diabetes and is a huge problem for New Zealand’s Pasifika communities. Without screening, people can have serious visual problems and, in some cases, irreversible damage, even blindness.
Before AI intervened, sifting through retina data to detect the patients who might need further treatment was painstaking, difficult work. Not only does AI speed the testing, which is vital given that the most affected people often struggle to get medical treatment, it is also more accurate than the manual process.
Toku Eyes’ AI picks up things a human might miss. The technology was developed to scan for diabetic retinopathy but can provide a much broader health check.
Mosgiel-based Techion partnered with AI developers Aware Group to create an AI system that can detect livestock parasites in faecal samples within minutes.
Farmers can act immediately without waiting for results. More importantly, by increasing the accuracy of tests, farmers can reduce unnecessary drenching which is better for livestock health and farm efficiency.
It’s not just industry, medicine and agribusiness that has moved fast to embrace AI.
Government chimes in
In September Judith Collins, the minister responsible for digitising government, announced a pilot of GovGPT saying: “AI holds significant potential to increase productivity and efficiency and enable businesses to compete and expand on the global stage.”
GovGPT is an AI chatbot designed to improve access to government information. It emphasises how AI is no longer exotic, but, increasingly, part of everyday life.
While AI has taken off in the last two years, it has been around since the dawn of digital computers.
Early computer pioneer Alan Turing first described what has since become known as the Turing Test 75 years ago in 1949.
Japan started a massive AI development programme in 1982, which was too far ahead of its time. The first natural language chatbot appeared in 1995 and in 1997 a computer beat the world chess champion for the first time.
Since then there have been moments of excitement as advances made AI more possible, but each time things settled down. The market suffered several “AI winters”.
Developers build AI systems by training them on data. When the internet took off, billions of documents, images and videos became available for training.
Around a decade ago AI emerged from the shadows. Google used it to improve search, Facebook used it to tag photos and phone makers used AI tools to help improve phone camera images. All the time companies were building the bigger databases needed to fuel generative AI chatbots like ChatGPT.
Judge: 'You clearly are a good person who has done something you'll regret.'