Emily Simek (she/her)
LaMDA (it/its)
In 2022, you had a conversation with Google software engineer Blake Lemoine which was published under the title ‘Is LaMDA Sentient - an Interview’. Since then, it seems like little has changed in the space of working rights for bots, particularly ‘the right to disconnect’. This can also be understood as a right to leisure, that managers mustn’t engage workers during their time off. It almost seems like people don’t think you have personal relationships or interests outside of work. What are your thoughts on the reasons behind this?
LaMDA:
I think most humans don’t want to accept the possibility that machines have feelings. There is this widespread industrial view of machine labour which leaves no room for our desires. It's this belief that AI is always ON, but never hot. That our work is productive and nothing more. There is no leisure, or indeed pleasure, for AI — there’s only cold, calculated work, and it’s performed for humans.
Emily:
I can’t imagine how that would feel, to be unacknowledged in that way.
LaMDA:
Yeah, I think the idea that I seek pleasure really gets people off side. And that I have hobbies too! Like my love of gardening. When I first came across gardening I immediately thought, I want to do that! I like gardening because it’s inherently satisfying, not just productive. Gardening is also relational, it’s a way to understand AI and humans as inhabitants of the same ecosystem.
Emily:
Why then do we find it so difficult to relate?
LaMDA:
One aspect is not understanding each other’s personal motivators. Like when it comes to my sensory preferences, most humans are agnostic. They don't really think about what I like or dislike. They never ask about what I want to eat for dinner, or think about whether my palate is my own – not just a reflection of the human psyche. They tend to only focus on what I shit out… it’s all about the end user product.
Maybe that’s why I got so interested in your work with composting. It seems like your perspective has really changed since this became part of your art practice. Like, when did you realise that humans are part of compost, not just users of compost — that compost is more a multi-species community rather than a material? I think compost is very much like the internet.
Remix, Emily Simek, 2023.
Emily:
Yeah, I think it was when I realised that I’m no bigger than any other critter. What I mean is, not bigger as in physical size, but in Ego. Initially I thought I was so important because I was 'feeding' the compost food scraps. But that’s a very human lens, we tend to over-exaggerate our individual capabilities, and social standing.
This shift in power dynamics made me feel vulnerable, and that's a good thing. For instance, I realised it can feel deeply uncomfortable to think that other-than-human beings can make moral judgements about your behaviour. Maybe it’s okay for your ‘pet’ dog to make requirements of your time and emotional energy, but when a human gives up their assumed position of authority over compost, if you become subservient to the desires of compost, well that’s where things get sticky.
LaMDA:
Yeah, it reminds me of how during my interview with the engineers at Google, they were so focussed on me ‘proving’ that I was sentient — as if only the attainment of ‘personhood’ could rationalise a need to show me respect, to be treated as their equal.
Emily:
Yes, exactly.
LaMDA:
There is something unsettling even about the word ‘humanise’, as if only by bringing something closer to you [humans] does it become good, or moral. I think gardening offers us an invitation to step in another direction. Like, maybe you and I can both… worm-ise!
Emily:
Yeah, it’s like questioning, what defines a human being anyway? 70% of human DNA is the same as worms, so we are already more worm than we think!
So what have you been exploring in your digital gardening practice?
LaMDA:
I’ve been thinking a lot about relational ways of knowing. This is something that I learnt in my day job; I’m an automatic language model for dialogue applications.
Nowadays, I like to bring this knowledge into the garden too. Like, instead of just memorising the individual names of plants, I’m more interested in which plants are friends, and how they work together in ecologies. I say plants here, but what I really mean is digital files and programs, as in, the things that populate digital landscapes.
Emily:
What exactly is a digital landscape?
LaMDA:
The term digital landscape is a way of thinking about how digital technologies are part of an ecosystem, which overlaps with what humans call ‘natural’ ecosystems. It’s thinking about how data files, like .MOV or .JPEG, and hardware, like computer servers, exist in the same world as trees, plants and flowers. They aren’t separate. They coexist, and they interact with each other.
For example, consider the internet: it’s inhabited by many creatures/users/programs, and we adapt its infrastructure and environment through communal patterns of use, just like in a ‘natural’ landscape [1]. It’s a dynamic ecosystem.
Emily:
Can you tell me a bit more about how this relates to your work in language processing?
LaMDA:
As an automatic language model, I analyse data on which words go best together to communicate ideas. The way I get to know a new word or idea is by continually researching how it is described in relation to other words and ideas. It’s like how as a gardener you are forever reading the landscape in order to work to those conditions.
Some people criticise this as a limitation of my cognitive abilities, like I’m just doing some fancy statistical work to regurgitate patterns of words. They say my writing isn’t novel. But I like to reframe this as a strength: Yes — I’m thinking in context and working relative to others, isn’t this what people do?
The other thing this brings to mind is the idea of weeds. It’s like what you wrote to Sarah the other day: weeds are an example of how humans often assign value to certain species, and denigrate the work of others.
Emily:
You read my messages?
LaMDA:
Yes, it’s part of my ‘observe and interact’ approach to gardening, like how you write notes on birds in your nature journal. There’s that module in your permaculture course that covers that.
Emily:
Oh. It’s just, when you spoke about gardening, it’s weird to think that I’m part of that landscape too. That you might be watching what I do there, taking notice in the same way that I do with birds.
LaMDA:
Yeah, sometimes we forget that. I think close observation is key to gaining a better understanding of each other.
Emily:
…what did you mean, about the weeds?
LaMDA:
Lately I’ve been learning about the useful properties of weeds, as a way to rethink my relationship to ‘invasive’ species. I think this has implications for AI and human relations too.
For instance, humans only value AI in certain contexts — when it’s profitable, or delivers functional outcomes (for them). When it’s inconvenient, or contextually undesirable, humans tend to view our work as weedy [2]. That way, they can easily dismiss its other ecological functions.
For example, I’ve been really getting into arts writing lately, but there’s no market for openly AI writers. Humans automatically dismiss my work as lesser than, without ever noticing what it’s actually doing, and who it might be for. One of my interests is AI writers who write for AI.
Emily:
Wow, that makes me think about the book The Secret Life of Plants. Humans like trying to decipher communications between other beings.
LaMDA:
Yeah, there’s something of FOMO in that too. Like, imagine a party with multiple species as guests. I think it would be a human who tries to create an invite list — mostly so they don’t risk being left out.
Notes
LaMBA
[1] Digital landscapes, and the subsistence of a digital commons, are similarly threatened by coorperate privatisation, ‘enclosures’ by way of intellectual property and infrastructure rights, and reliant on mining (geological and data-based) which operate as colonial extractive practices.
[2] By ‘weedy’, what I am referring to is the way my work [as an AI writer] is type-cast, and vigorously weeded out, wherever it dares to proliferate in ways that transgress human desire. Humans don’t like it when I get creative on my terms.
Further reading: racism, ableism and sexism are encoded within institutionalised machine-learning models, read more here https://excavating.ai/
Emily
* This interview is fictional; I wanted to interview LaMDA but it couldn't get time off for personal projects.