Samantha Augusta

Remember Dieselgate? Companies lied about emissions, and execs faced criminal charges—in some jurisdictions, personal liability applied.

Why should AI companies that skirt safety evals or ignore known risks be treated differently?

#bluesky #AI

AI isn’t out of control.
It’s under control—just not ours.

#bluesky

This is a rarity in frontier AI discourse: Bengio et al. endorse the precautionary principle—arguing we must prove safety before scaling.

Their “Scientist AI” proposal allows us to disable agentic and planning components—building in off-switches from the start.

📄 arxiv.org/abs/2405.20009 #bluesky

Some great initiatives for tracking AI harms I've been following so far include:

- AIAAIC (www.aiaaic.org/aiaaic-repos...) and
- MIT's AI Incident Tracker (airisk.mit.edu/ai-incident-...).

Pretty shocking to see the numbers on autonomous vehicle incidents. Very few of these reach the headlines.

Not all x-risk bangs. Some simmer. Kasirzadeh warns: AI may collapse social systems via accumulative harms—slow-moving, systemic, invisible. Real systems unravel through misalignments over time.

AI safety needs tools to track compound harm.

📑 arxiv.org/abs/2401.07836

#TechEthics #bluesky

Most model evals focus on benchmarks—but what about catastrophic misuse? Shevlane et al. propose tools for extreme risk evals, urging labs to test frontier AI models for deception, persuasion, and autonomy before deployment.

To what extent is this happening in practice?

📄 arxiv.org/abs/2305.15324

#Tech #TechEthics #AI #bluesky

What if existential risk from AI doesn’t arrive with a bang, but builds slowly beneath our feet? Kasirzadeh warns of a ‘boiling frog’ scenario—AI risks that compound silently, eroding systems until collapse. We must reckon with both the decisive and the accumulative 💭

📄 arxiv.org/abs/2401.07836

What I find useful in Peterson’s approach is how it sidesteps the usual “which theory is right?” trap. Instead of starting with utilitarianism or deontology, he looks at recurring judgment patterns. That kind of mid-level mapping seems especially helpful in bio risk, where stakes are so high

Reading Martin Peterson’s Ethics of Technology has me thinking. He doesn’t push a grand theory — he models how we actually reason in practice. 5 principles show up again and again: cost-benefit, precaution, sustainability, autonomy, fairness. Not foundational; functional for the domain in question.

Let's forget the quest for the one true ethical theory and focus on the Goldilocks zone: mid-level, domain-specific rules distilled from similarity clusters across real cases. Concrete enough to steer frontier AI & biotech, flexible enough to evolve—no consensus needed. #TechEthics #AISafety

“Tesla’s driverless ‘robotaxis’ could launch in Austin as soon as June 22. But a demo in Austin today showed a $TSLA, manually driven to test its Full Self-Driving system, failed to stop for a child-sized dummy at a school bus—and hit it.”

@cbsaustin @velez_tx

Gradual Disempowerment: Systemic Existential Risks from Incremental AI Development This paper examines the systemic risks posed by incremental advancements in artificial intelligence, developing the concept of `gradual disempowerment', in contrast to the abrupt takeover scenarios co...

9/9 📄 Gradual Disempowerment paper - give it a serous read: arxiv.org/abs/2501.16946

8/9 These billboards should serve as reminders of what’s at stake if we don’t steer ahead wisely. We can’t afford to treat safety, alignment, and human relevance as secondary problems.

7/9 If the end goal of this "race" is to automate all economically valuable human work—and if we take companies like OpenAI’s mission statements at face value, that does seem to be the goal—then we need to ask: what’s on the other side of winning?

6/9 As AI systems become more capable, they may replace us not in one dramatic leap, but across countless small decisions—undermining the systems that currently tie economic, cultural, and political power to human needs.

5/9 The Gradual Disempowerment paper (linked below) argues that existential risk from AI may not come from a sudden “takeover,” but from the slow erosion of human influence.

4/9 But after days of sober discussion in Leuven about AI's societal impact—especially the risk of gradual disempowerment—this billboard struck me differently. I’d normally laugh at them—they’re such a uniquely Bay Area phenomenon—but this time, it was harder to do that.

3/9 Sometimes I wonder how smaller startups I've never heard of even justify them. Maybe it’s become a rite of passage in the valley to occupy some traditional adspace like this.

2/9 If you’ve been to the Bay Area recently, you’ve probably seen these kinds of ads—billboards for GPUs, foundation models, or whatever the current tech craze is—along freeways, airport terminals, and city buses. They’re bold, a little comical, and a signal of sorts.

1/9 Right after landing back in San Francisco, I was greeted by the billboard pictured below: “Win the AGI Race.” I had just returned from the KU Leuven's conference on Large-Scale AI Risks, where we spent several days in serious conversation about the long-term consequences of advanced AI systems.

Back from KU Leuven's AI risk conf, greeted by ths billboard: “Win the AGI Race.” After days discussing disempowerment—the slow loss of human influence from AI—this felt ominous. If 'winning' = automating all work, what’s left? Safety must outrun speed. 📄 arxiv.org/abs/2501.16946 #AI #AISafety #tech

From biotech to energy, we’ve seen tech outpace foresight—AI is no different. @yoshuabengio.bsky.social’s keynote hit this hard: safety must be our top priority. Grateful to Torben Swoboda, Lode Lauwaert & Andrew Rebera for making space for these urgent conversations.

Presented my AI safety work at KU Leuven’s International Conference on Large-Scale AI Risks. Both terrifying & exciting to be on the same schedule as people I’ve been reading for years! The tone throughout was frank, serious, and grounded in hard questions.

Nutritional Content of Ready-to-Eat Breakfast Cereals Marketed to Children This cross-sectional study examines trends in the nutritional composition of children’s ready-to-eat cereals introduced in the US market from 2010 to 2023.

🥣 More sugar in cereal

A study of 1,200 kids’ cereals launched since 2010 finds rising fat, salt & sugar – and falling protein & fibre.

Despite health claims, many cereals now pack over 45% of a child's daily sugar limit per bowl.

🔗 doi.org/10.1001/jama...

#Nutrition #ChildHealth #SciComm 🧪

You’ve probably heard of AI companion Replika and it’s disturbing marketing—”Always here to listen and talk. Always on your side”. But have you heard of Gigi and Cluely? @luizajarovsky.bsky.social is doing a fantastic job covering the rise of a whole new class of unethical AI

Amplifiers and Effects on Substack You’ve probably heard of AI companion Replika and it’s disturbing marketing—”Always here to listen and talk. Always on your side”. But have you heard of Gigi and Cluely? Luiza is doing a fantastic job...

substack.com/profile/3431...

Long-Term Exposure to Environmentally Realistic Doses of Starch-Based Microplastics Suggests Widespread Health Effects There is a growing consensus on addressing the global plastic pollution problem by advocating for bioplastics. While starch-based plastics are prevalent, the potential health implications of starch-ba...

🥣🧪 Bioplastics might not be as 'green' as they seem.

A new study found long-term exposure to starch-based microplastics caused liver, gut & ovarian damage in mice - and disrupted blood sugar & circadian rhythms.

🔗 doi.org/10.1021/acs....

#Plastics #Toxicology #SciComm

#BlueSky #academic #research #socialmedia #help

I find social media a hard thing to keep up with. As an academic and writer, concentration is my superpower—I do my best in long spans of uninterrupted work. But here on BlueSky, that's exactly what doesn't help with building a network. I'm looking to find a way to make it work though—any tips?!