A new CSIRO report outlines how foundation models -- technology behind generative AI products including ChatGPT -- could supercharge the Australian economy and revolutionize industries by increasing productivity and solving some of society's most fundamental challenges.
According to a report from the CSIRO-the country's national science agency-foundation models underpin the generative AI being used to reshape industries around the world and set to realize significant positive impacts in Australia.
These models were trained on gigantic datasets and could run complex and generalized tasks across industries. These models are the fuel for a rash of popular AI products developed for the public at large by companies including OpenAI's ChatGPT, Microsoft's Copilot, and Google's Gemini. In the last few years, at least 125 foundation models have been developed globally; 73% of them come from the United States, 15% from China, and the rest from Europe. Most of these models are developed by private technology companies.
This report from CSIRO puts into context the current global landscape with regard to foundation models; it enumerates risks and opportunities that this emerging technology presents for Australia in its use. But lately, the foundation models have gained a lot of attention because they can create really creative outputs: from writing speeches down to even poems. The real value of the foundation models, though, is in processing large-scale data and solving complex problems. This can lead to huge advancements in areas such as health care, whereby AI can assist in reducing medical misdiagnosis currently at over 140,000 annually in Australia.
According to CSIRO's Professor Elanor Huntington, "Building Australia's sovereign capability in AI is important, but foundation models are particularly significant. There may be benefits associated with the fine-tuning of existing international models in terms of cost and speed, but the tendency of reliance on foreign models is fraught with security and reliability risks. Besides, such models might not be culturally appropriate for the Australian context.
They have also suggested that such foundation models will help Australia make the most of it by developing public sector AI models, providing access to high-performance computing, sharing data resources, encouraging training in skills, and international collaboration.
As Professor Huntington would affirm, the opportunities afforded through these foundation models are huge; at the same time, however, it is up to Australia to constantly be mindful of the accompanying risks and ensure that the technology is leveraged in such a way as to assure maximum benefit for the country at large and its citizens.