Peter McAllister

I expected it to sit on the bookshelves under dystopian fiction, and now it seems to be appearing under current affairs.

An AI going rogue could just be something that is capable of doing something fairly simple and straightforward, but ridiculously fast in a ridiculous number of times.

Bad data, bad outcomes, they are the early warning sign.

Once it's done something small wrong, it's going to get to something large wrong very, very quickly.

We make way more mistakes than we ever admit to.

LLMs are just a really, really, really, really, really overblown autocorrect.

Thank you very much, Dietmar.

It's a pleasure to be here.

Well, that's actually a very hard question and I really mean to start with that one first.

I guess the key thing about AI for me is that as it's developed and as we've worked through it over the last several, well, couple of decades into machine learning and those kinds of areas, there's been this desire to, for want of a better term, mimic the way human brain works, mimic those processes.

um and make them better faster stronger um and that's to effectively take and use the good parts of our brain the bits that are doing the processing the bits that are doing the calculating the bits that are doing the reasoning and the logic but we've got another part of our brain we've got another part of our brain which is is emotional and has negative attitudes and and is prepared to do bad things so the the idea of the book was well

what if that comes across as well and how does that actually end up presenting so um the the title the um the uh the code um if your ai loses its mind can it take meds was really about the idea of if an ai starts exhibiting the behavior that we would consider mental illness in a human

what are you going to do about it?

What can you do?

Because you've got this souped up, amped up thing that's used to running at a million miles an hour that's amplifying that the same way as it might amplify something good.

So the book was sort of exploring that approach in an AI and in parallel takes a main chasm, one of the protagonists that is also battling mental illness themselves.

So it sort of takes a machine through that process and a human through that process.

Yeah.

So right now an AI going rogue could just be something that is capable of doing something fairly simple and straightforward, but ridiculously fast and a ridiculous number of times.

So something that it just basically, if you have something that is just browsing the web and it goes a little bit crazy and starts to consume everything you've got there, then all of a sudden you've got an AI doing a denial of service attack.