My Favourite AI Party Trick

One of my new favorite things to do is to go to a party in Australia, find the most anti-AI person there, tell them my job is building AI automations for companies (usually interpreted as: I spend my day automating people away), and try to have a conversation where they end up not hating me. (Note that this is much less interesting in SF because so many people there are part of the tech crowd.)

I started doing this because I’ve spent a bunch of time this year in SF with the people building this tech, and a bunch of time in Australia with the people not building this tech. Sitting across the two places makes it extremely clear how little understanding there is between the two groups of people. Everyone knows this intellectually. But I’m not sure many SF tech people have viscerally experienced it; I don’t see many tech people expending time and energy figuring out how to get buy-in and feedback from everyone else. There’s a lot of dark humor and excitement about the tech, both of which I enjoy. But there’s not a lot of nuanced conversation about, and most importantly with, non-tech people. If we expect people to accept this technology, we have to carefully consider how it’s perceived by them.

As AI makes its way through the economy in the next couple of years, one of the most important skillsets will become convincing groups of people that AI can be a genuinely worthwhile technology, even though they’re currently frightened and feel incentivised to stop its use.

I learn a huge amount from having constructive discussions with anti-AI folk. I learn a huge amount from implementing AI with teams of people who are validly concerned about their jobs. If you’re planning to have these conversations, you have to go in with empathy, curiosity, and without the expectation that the other person is going to change their mind. Otherwise you risk a bar fight - you have been warned.

A disclaimer that I am NOT talking about whatever general superintelligence Anthropic/OpenAI/whoever else are worried they’re developing. This is NOT a post about my thoughts on the pace of AI research. This is a post about current frontier AI (at time of writing, Opus is the most widely used model, Fable has just been released); which means it will naturally be out of date in a few months.

Also a disclaimer that I feel a bit silly writing this down because it feels obvious; but given the level of tension about AI in the world today, it’s clearly not being done well at scale.

So, without further ado:

How to speak constructively with people who are anti-AI (if you are an AI nerd)

Try to be a normal human. Make a joke. Be a little self-deprecating. Smile. Be genuine. Have actual interest in their life. These are all basic social skills, but (and I say this as a tech nerd who loves tech nerds) the tech-nerd archetype is often not this.
DO NOT go into the conversation expecting to change their mind. If you go in to change their mind, you will inevitably end up lecturing and nobody likes being lectured at a party. Your whole job is to understand their views, and share your own views, in a respectful way. Nothing more, nothing less.
Talk quietly. This forces them to listen harder to you, match your tone in return (it’s harder to get riled up when talking in low voices), and makes sure neither of you is showboating for a group.
The first thing you should do, always, is ask what makes them scared about AI. Make sure you say this with the right tone and body language because it can be taken the wrong way. I have found that a lot of the time, someone who is aggressively against AI can struggle to put into words why. Here are the most common reasons I get from people:

Environmental impact

If someone mentions this as the primary reason they’re scared about AI, ask them what the environmental impact of AI is. They will say something about water and power usage. Ask what the water and power usage of AI is. They probably don’t know, because there isn’t an easy answer to this - no-one really knows. Google the latest information and learn together. I particularly like this article from MIT Technology Review.

People can and should draw their own conclusions from the numbers. My current take that I share in these discussions: this is a hugely important thing to think about, and data centers are the fastest growing consumer of electricity. I think about that a lot - and I know the AI labs do as well, because they’re highly incentivised from a cost perspective to make AI inference more efficient in every way. I believe the environmental impact is a reason for research and transparency on the resource use of AI. I believe it is a reason to put a lot of effort into increasing our ability to produce clean energy. I believe it is a reason to find ways of cooling data centers with less water and recycled water wherever possible. I don’t believe it is a reason to entirely discount the technology.

(There are also more effective levers to reduce electricity use, like air conditioning buildings less aggressively, but that’s a whole other conversation.)

It’s automating away our jobs and so we’re all going to starve

My take: firstly, I have yet to work with a company whose goal in using AI was to fire a bunch of people. I know that happens, but it hasn’t been my experience. In all the teams I’ve worked with, management’s goal has been to use AI to make it possible to do more with their current team.

This does, of course, mean that less of those jobs will be available to the younger generation. This is what I worry about more when I think about this point. It seems likely to me that a lot of young people will have developed skills for jobs that no longer need as many of them.

My hope, and what I’ve seen, is that this means more people are pushed into more meaningful work. The type of tasks that I am automating with AI are involve people spending 8 hours a day manually moving data between spreadsheets or writing reports that nobody will read. I respect the people I work with greatly, and I don’t believe those types of tasks deserve their brainpower.

I see two categories of work that can’t be automated by current AI: 1) relationships-based work, and 2) work at the frontier of human knowledge. If less people spend their careers as data monkeys, and more people spend their careers on relationship- or frontier-based work, I believe the world will be a better place. There will be a transition that may be scary or uncomfortable, and it will require people being willing to learn and develop new skillsets, but I think it’s a world worth creating.

The most common reponse I get to this, and one which is completely valid, is people’s fear about being able to put food on the table if AI starts doing their job. This sounds like a problem with AI, but it is not a problem with AI. If this is the concern, then there is a very real and nuanced conversation to have about what systems our society would need to have in place so that people don’t starve in this scenario. But the challenge there lies in the systems around AI, not in the technology itself, and I find it helpful to remind people of this difference.

Where will we find purpose without our jobs

This point is, to me, the most interesting. Project forward to a world where everything is done for us by AI and robots - what will we do? Won’t we be listless, depressed, purpose-less people like in Wall-E?

I absolutely don’t see that future happening, for a few reasons.

There are people who make work their life and derive a lot of meaning from it. I am wholeheartedly one of those people. But the majority of people are not like this. Most people, if you handed them $10 million, would want to quit or greatly reduce the amount of time they spend at their job. Most people derive a large part of their meaning in life from things other than their work. Most people making this point aren’t actually worried about losing their meaning, but about losing access to resources - which is the point above.
Humans are incredibly good at making up things that feel like they matter, particularly with status as the stake. We are wired for it. Take sport, for example. I would call sport a gamified version of hunting. Our biology is wired to chase something. Since we invented agriculture, we haven’t had to chase things. That doesn’t make the wanting-to-chase-things go away - instead we’ve created this entirely made up concept that presses those same chasing-things buttons but makes it optional. It is now a giant industry and people find huge amounts of meaning and purpose in it (more than many people find in other work). If knowledge work becomes optional in the same way hunting did, I see no reason why the desire to think would leave us. I just believe we would come up with a gamified version, that we’re still able to derive real meaning and purpose from - perhaps more so because it’s a choice rather than a necessity. In a world where all knowledge work becomes unneccessary, I find it likely that things like chess, soduku, maths competitions, and strategy games increasingly allow us to compete and derive meaning like NRL or basketball does for the general population today. It is extremely possible that is the nerdiest thing I have ever said, and to anyone who isn’t currently into chess, soduku, maths, or strategy games, I know it sounds absurd. But I’m willing to bet that golf would sound absurd to a caveman; and to some far future generation, we are the cavemen.
Until we have travelled the universe and conquered death itself, there will always be something to do. It’s just a matter of who is curious enough to do it.

Superintelligence will destroy us

This one is relatively rarely mentioned by non-tech people, but I bring it up because it’s the one I resonate with most highly. It’s also the one I know most tech people are concerned with and considering, so I’m going to mention it without putting my take here (see above: this is scoped to Opus/Fable-level AI and its impacts on the economy, not the pace of AI research).

By the time I get through all these points with my new friend - and we do usually hit on all these points - we’ve had a wide-ranging, vulnerable conversation. I’ve usually learnt something from speaking with them and interrogated my views from a different angle, and I hope they’ve done the same.

And that, ladies and gentlemen, was the goal. Something in the wiring of my brain finds this an incredibly fun and meaningful exercise. My friends are a bit bemused by it. I don’t really have a point to make other than to document it and say that these conversations are worthwhile even when uncomfortable. I hope more people have them in good faith.