Welcome to week 17 of January 2025, the month that seems never to end – however, I have been reliably informed that this IS, in fact, the LAST week of the month. Since time appears to be standing still, we thought we’d reward you with something special! TL;DR Tuesdays are famed for our TL;DR Shorts, but Dr Danny Hillis, founder of Applied Invention, becomes only the second contributor in a year to be awarded an exclusive TL;DR Long – and our longest non-Speaker Series offering so far. To explain why he had so many thoughts, all I need to say is Artificial Intelligence. In this episode, Danny chats about the history of AI, from working with the field’s founding fathers to predictions that have come true, and what we can really expect from AI in the coming years.
Although AI feels like a recent tech development, Danny reminds us that it has a long-established history. Danny worked alongside the likes of Marvin Minsky, and Claude Shannon – no, they’re not Bugsy Malone characters but are two of the team members who established the field of artificial intelligence. Working with them, Danny and the crew discovered that what they thought would be easy was much harder than expected, while what they were wary of was much easier to achieve. Pattern recognisers were developed with little effort, but creating a computer that could beat a human at Chess was much harder.
It turned out that the main barriers to success were a lack of data and, the most limiting factor of all, a lack of computational power. But that’s OK because Danny’s PhD focused on what would be required to build the biggest computer. He discussed his Thinking Machines in our Speaker Series chat which we shared last month.
Danny notes that today’s AI researchers are working on algorithms that are very close to the ones the team imagined back at the start of this area of research, however, he reminds us that we are still way off machines that can replace humans. While well-trained machines can carry out specific talks well, they are missing the critical thinking part of intelligence, however good they are becoming in mimicking intelligence, as evidenced in numerous case studies of AIs that hallucinate, or create solutions that look and sound right based on the fact that the machine has recognised patterns and attempts to apply those rules but that, without real meaning or understanding, are factually incorrect. Danny tells the story of how his granddaughter can recognise patterns in visiting contractors and become someone who sounds like an expert in moments, but scratch the surface and there is no real knowledge of the area to be able to make logical decisions. I too am reminded of the time I accidentally found myself co-piloting an island-hopper propellor plane across Belize, having curiously followed the actions of the pilot for the first two stops – but we’ll save that story for another time. The year is young, and we’ve got lots more to chat about, and many more stories to share.
Danny reflects that, while to experts it doesn’t feel like AI has moved on much since the development of supercomputational power, there is a change coming, as evidenced by the ever-increasing rate of development in the area. The difference this time around is funding, which is attracting the smartest minds in their droves, catalysing this progress by exploring the intuitive aspects of this technology.
To make this technology truly good, Danny firmly believes that a source of truth is required. One of his interests is building a knowledge graph of the provenance of information, which he further expanded on in last month’s Speaker Series. This would go some way to building technology that is as robust and trustworthy as possible, while attempting to eliminate biases or building on questionable knowledge that can bed into the foundations, creating points of future weakness and instability.
The great thing about building good technology is that it in turn starts to iteratively learn and teach itself, generating more knowledge, even about that knowledge itself. These are exciting times for AI, but public and research community engagement remains vital to ensure that developments do not double down on historically discriminatory narratives or unscientific knowledge that have no place in today’s society.
Subscribe now to be notified of each weekly release of the latest TL;DR Short, and catch up with the entire series here.
If you’d like to suggest future contributors for our series or suggest some topics you’d like us to cover, drop Suze a message on one of our social media channels and use the hashtag #TLDRShorts.
