[00:04] And a couple of months ago, I got the weirdest email of my career. saying that my work in AI is going to end humanity.
[00:18] Now I get it,  AI, it's so hot right now. It's in the headlines pretty much every day, like discovering new molecules for medicine
[00:30] But other times the headlines have been really dark, like that chatbot telling that guy that he should divorce his wife or that AI meal planner app proposing a crowd pleasing recipe
[00:43] And in the background, existential risk and the singularity, to make sure that doesn't happen.
[00:57] Now I'm a researcher who studies AI's impacts on society, and nobody really does. But what I do know is that there's some pretty nasty things going on right now,
[01:12] It is part of society, and it has impacts on people and the planet. Their training data uses art and books created by artists
[01:26] And its deployment can discriminate against entire communities. We need to start being transparent and disclosing them and creating tools
[01:38] so that hopefully future generations of AI models maybe less likely to kill us, if that's what you're into.
[01:50] because that cloud that AI models live on is actually made out of metal, plastic, And each time you query an AI model, it comes with a cost to the planet.
[02:05] which brought together a thousand researchers the first open large language model, like ChatGPT,
[02:17] but with an emphasis on ethics, transparency and consent. And the study I led that looked at Bloom's environmental impacts as 30 homes in a whole year
[02:30] which is like driving your car five times around the planet And this might not seem like a lot, like GPT-3,
[02:45] But the thing is, tech companies aren't measuring this stuff. And so this is probably only the tip of the iceberg, And in recent years we've seen AI models balloon in size
[03:00] because the current trend in AI is "bigger is better." In any case, we've seen large language models in particular And of course, their environmental costs are rising as well.
[03:16] The most recent work I led, found that switching out a smaller, emits 14 times more carbon for the same task. And as we're putting in these models into cell phones and search engines
[03:32] the environmental costs are really piling up quickly. So instead of focusing on some future existential risks, and tools we can create to measure and mitigate these impacts.
[03:49] a tool that runs in parallel to AI training code and the amount of carbon it emits. like choosing one model over the other because it's more sustainable,
[04:04] which can drastically reduce their emissions. because there's other impacts of AI apart from sustainability. to prove that their life's work has been used for training AI models
[04:23] And if you want to sue someone, you tend to need proof, right? So Spawning.ai, an organization that was founded by artists, created this really cool tool called “Have I Been Trained?”
[04:35] to see what they have on you. I searched LAION-5B, to see if any images of me were in there.
[04:49] that's me from events I've spoken at. They're probably of other women named Sasha who put photographs of themselves up on the internet.
[05:01] to generate a photograph of a woman named Sasha, Sometimes they have two arms, but they rarely have any clothes on.
[05:16] to search these data sets, this provides crucial evidence that her life's work, her artwork, and she and two artists used this as evidence
[05:32] for copyright infringement. (Applause) the company where I work at,
[05:46] to create opt-in and opt-out mechanisms for creating these data sets. for training AI language models.
[05:58] (Applause) You probably hear about this a lot. that can represent stereotypes or racism and sexism.
[06:14] when she realized that AI systems wouldn't even detect her face Digging deeper, she found that common facial recognition systems
[06:26] were vastly worse for women of color compared to white men. And when biased models like this are deployed in law enforcement settings, this can result in false accusations, even wrongful imprisonment,
[06:40] which we've seen happen to multiple people in recent months. at eight months pregnant because an AI system wrongfully identified her.
[06:52] and even their creators can't say exactly why they work the way they do. And for example, for image generation systems,
[07:04] if they're used in contexts like generating a forensic sketch they take all those biases and they spit them back out for terms like dangerous criminal, terrorists or gang member,
[07:18] when these tools are deployed in society. I created this tool called the Stable Bias Explorer,
[07:31] through the lens of professions. Don't look at me. A lot of the same thing, right?
[07:45] And none of them look like me. is that we looked at all these different image generation models significant representation of whiteness and masculinity
[07:59] even if compared to the real world, These models show lawyers as men, even though we all know not all of them are white and male.
[08:14] And sadly, my tool hasn't been used to write legislation yet. But I recently presented it at a UN event about gender bias as an example of how we can make tools for people from all walks of life,
[08:27] to engage with and better understand AI because we use professions, but you can use any terms that are of interest to you. are being woven into the very fabric of our societies,
[08:42] even our justice systems and our economies have AI in them. And it's really important that AI stays accessible so that we know both how it works and when it doesn't work.
[08:56] And there's no single solution for really complex things like bias But by creating tools to measure AI's impact, and start addressing them as we go.
[09:12] Start creating guardrails to protect society and the planet. companies can use it in order to say, this model because it respects copyright.
[09:25] can use these tools to develop new regulation mechanisms or governance for AI as it gets deployed into society. to choose AI models that we can trust,
[09:41] not to misrepresent us and not to misuse our data. that said that my work is going to destroy humanity? I said that focusing on AI's future existential risks
[09:54] very tangible impacts and the work we should be doing right now, or even yesterday, Because yes, AI is moving quickly, but it's not a done deal.
[10:08] and we can collectively decide what direction we want to go in together. (Applause)