It’s a well-worn cliché that technology moves quickly. Drones, hypersonics, blockchain, AI – all have arisen in the last decade or so. But one of those things may not be like the others.
At AutogenAI’s recent lunch event in Canberra, Sean Williams – founder and chief executive of the company, which uses AI to help companies write winning tenders – gave a presentation on the historical importance of AI. The technology of large language models, he argued, is not just another innovation; it is a historical event on par with the invention of the abacus, the printing press and the Internet.
“The technologies that change the world need to do two things,” Williams said. “First, the thing that they do needs to be fundamentally human. There’s no point building a technology for something that humans don’t do. Tools are for humans to do the things we do.
“Second, the tool needs to speed up that human activity by orders of magnitude. You don’t just improve by five or ten per cent – you improve by hundreds of percent.”
For example, the abacus allowed ancient merchants to add and multiply faster – leading to more complex trade. Trade is a fundamental human activity; the abacus, Williams says, improved our ability to trade by orders of magnitude. So, it changed the world.
The invention of the printing press improves the sharing of ideas (another fundamental human activity) by orders of magnitude; the steam engine exponentially speeds up humanity’s ability to make things; the internet continues the work of the printing press by further orders of magnitude.
Now, Williams argues, large language models are speeding up our ability to create ideas.
“Large language models allow you to read orders of magnitude faster, and write orders of magnitude faster,” Williams said.
“Reading and writing are fundamental human activities. This is a technology that speeds them up by hundreds of per cent.”
To illustrate the public appeal of AI: the journey to one million users took Netflix three and a half years; Kickstarter, two and half years; Facebook, ten months; Instagram, two and half months; but ChatGPT took just five days.
So what might this look like in defence? Whilst large language models have many applications in the national security sector, one particular change – which AutogenAI aims to deliver – is to equalise the playing field of defence bidding, which is a complex and expensive process.
Many small and agile companies that might have capabilities relevant to defence are turned off by the cost, the complexity, and the risk of either not winning the bid or having the whole program fall after a strategic review or change of government.
Now, the laborious process of reading required documentation and writing the bid can be sped up greatly using large language models. Whilst humans will need to be involved constantly throughout the process to ensure quality, the technology nonetheless offers significant time and cost savings, according to Williams.
The adoption of large language models today, he argues, is the equivalent of the Internet in 1992.
“I don’t think AI is going to take your job,” Williams explains. “I think people who use AI are going to take your job.”
To hear Williams’ full presentation, listen to the latest episode of the ADM Podcast.