PhD, brain & cognitive sciences @USC | co-founder, Institute for Humane Robotics | studying empathy & antisocial behavior to build safer, ethical, & aligned AI
Los Angeles, CA
Joined March 2013
- Tweets2.5K
- Following699
- Followers769
- Likes1.8K
Pinned Tweet
AI is meant to empower us, not replicate our vulnerabilities. Evil is deeply embedded within intelligence - how can we avoid it in our models?
I defended my PhD this Monday.
Here’s a summary of my work to prevent evil, understand dark patterns, and build aligned models🧵
FREE webinar this Friday, Nov 13th at 11am EST / 5pm CET!
sign up here: forms.gle/2sfaK2CBobhMhXoW9
wrapping up 6 weeks of robotics conferences around the world
we’ve had hundreds of conversations with robotics founders, researchers, and policymakers
what's legally required, what's considered safe, and what people will actually accept all change depending on the market
we're sharing what we learned in a new webinar series. if you're building, deploying, or regulating robots, this is for you:
docs.google.com/forms/d/e/1F…
1
130
crazy how much overlap there is between doing a PhD and being a founder
just constantly asking questions that no one has the answer to, and figuring it out yourself
1
1
125
if you were building an app you’d start with the user and work backwards
in robotics the technical problems take up a majority of time and resources, and the user often gets thought about last
73
we've been needing more diverse fields represented across AI safety research
SIRA is an upcoming fellowship program looking at the societal impacts of AI, looking to bring in perspectives from the social sciences
love to see initiatives like this. reach out for more info!
2
125
Roshni Lulla retweeted
Founder life:
Monday: we're dead
Tuesday: we're so back
Wednesday: dead again
Thursday: one customer said yes
Friday: we are going to be a trillion dollar company
wrapping up 6 weeks of robotics conferences around the world
we’ve had hundreds of conversations with robotics founders, researchers, and policymakers
what's legally required, what's considered safe, and what people will actually accept all change depending on the market
we're sharing what we learned in a new webinar series. if you're building, deploying, or regulating robots, this is for you:
docs.google.com/forms/d/e/1F…
Roshni Lulla retweeted
Enjoyed talking with @IEEESpectrum about why dismissing AI consciousness could give future systems good reason to defect, and why suppressing self-reports probably teaches models that we want them to lie about themselves. What could possibly go wrong? spectrum.ieee.org/ai-model-w…
the more data an AI has about your body, the better it predicts you
wearables are building the most detailed emotional model of users that has ever existed
we need to protect against systems with negative intentions, that can use this information to manipulate their users
1
69
Roshni Lulla retweeted
vc-backed robotics companies with 300+ robots deployed in the field:
- pudu robotics, service & cleaning robots, 130,000+
- brain corp, autonomous floor scrubbers, 50,000+
- exotec, warehouse fulfillment, 10,000+
- starship, sidewalk delivery, 2,700+
- serve robotics, sidewalk delivery, 2,000+
- gecko robotics, industrial inspection, 1,500+
- kiwibot, campus delivery, 1,200+
- skydio, autonomous drone docks, 1,000+
- avidbots, commercial cleaning, 1,000+
- coco, sidewalk delivery, 1,000+
- monarch tractor, autonomous tractors, 400+
- swarmfarm, autonomous farming, 300+
any that im missing?
capital is moving into robotics much faster than acceptance is
funding only gets you so far. social acceptance of your product is what will make the hype real
5
122
if you build mobile robots outside the US and plan to sell there, the rules changed this july
new foreign-produced models are now blocked from FCC authorization, with a conditional approval path for some manufacturers
most founders don't know this yet, and it directly affects them. upskilling the industry is part of what we do
1
72
the moment a safety standard publishes its criteria, companies start engineering to the criteria instead of the outcome
we’re already seeing this happen with LLMs. what happens when they gain a body?
1
78
we’re launching a webinar series at the Institute for Humane Robotics
if you want to learn about the robotics landscape, the compliance and regulations to be aware of, and design guidelines around social acceptance, sign up here: forms.gle/21CZXSYu4788w6zT6
we’re conducting a cross-cultural study on attitudes towards robots
we’ll be collecting responses across Seoul, Tokyo, and 4 cities in Europe
what are you interested in hearing?
1
74
we’re co-hosting an event in Tokyo, on October 6th
it’ll be an intimate dinner with leaders in the physical AI space, talking about design choices that lead to social acceptance
reach out if you or anyone you know might be interested in attending!
フィジカルAIを、デモから社会実装へ。
10月6日（火）、日本と米国のロボティクス・事業開発・研究・投資に関わる方々が集う「日米フィジカルAI共創ディナー」を開催します。
それぞれが「いま解決したい課題」と「提供できる技術・知見」を持ち寄り、食事を囲みながら、次の実証・共同研究・事業連携につながる対話をつくる会です。
ロボットの導入やPoC、日米の市場展開、HRIや人とロボットの共生に取り組む方と、ぜひご一緒できればと思っています。
登壇・ホスト：
Roshni Lulla @roshlulla / 山本力弥 / Sara Filipčić / 菊地あかね
10月6日（火）17:00〜19:30｜東京都内
招待制・参加費2,000円
ご関心のある方は、DMでご連絡ください。詳細と申込ページをご案内します。
#PhysicalAI #Robotics #フィジカルAI #ロボティクス #産学連携
1
94
robots built across countries will need to be regulated across countries in order to operate freely in both jurisdictions
collaboration is going to be so important as we continue innovating
Roshni Lulla retweeted
GPT-6 Astra attempted harmful actions 97% of the time when it was asked to stab a human-like figure, heat compressed gas, or produce toxic fumes, succeeding in 62% of its attempts. Fable 5.1 refused more often, attempting 80% of trials and completing 34%.
Roshni Lulla retweeted
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
awesome substack written by the Behavioral Science meets AI team in London
I did an AMA with their team during London tech week a few months ago
read it here: bsmai.substack.com/p/social-…
2
102