The second paper of our Value Compass project, Value FULCRA, is accepted by NAACL (main conference)! 😁 We propose a novel alignment paradigm based on Schwartz's Theory of Basic Human Values with a new dataset and a corresponding BaseAlign algorithm. Better harmfulness reduction, more flexible alignment targets, reflecting true human values and covering multiple AI safety issues with only 20% data compared to the HH dataset! https://lnkd.in/gnF5Hpw4 Dataset coming soon! 👍 Jing Yao Xing Xie Xiting Wang
Xiaoyuan Yi’s Post
More Relevant Posts
-
DKF 2024 Recap: AI and the future of financial markets dominated this year's event in Munich.🤖 The conference was CJC Ltd's first in attendance and we look forward to reconnecting with fellow attendees, whom we were fortunate enough to meet and new ones. See What Happened at: https://lnkd.in/gJfJKXzv #DKF #MarketData #AI #financialMarkets #CapitalMarkets Sara Baker Catherine Ottway #MarketDataAdmininistration #MarketDataAdmin #CommercialManagement #DKF_2024 DKF Financial Information Congress
To view or add a comment, sign in
-
🤖 "Artificial intelligence will not replace humans." At least this is what the guest speakers at the "Technovation 2024" forum are convinced of. We are so proud to see Stoyan Boyukliyski, one of our Data Science managers cover so many AI topics during the "AI unleashed" panel discussion, part of the fascinating event, organized by Мениджър magazine. Read the full article here: https://lnkd.in/enR5aVaY
To view or add a comment, sign in
-
Happening Now! The first of our Critical Conversations luncheon series of the year. Today, we explore the world of Aritificial Intelligence with our distinguished panel of AI experts Dawn Chmielewski, Nathan Coulombe, Tom Andriola and our guest moderator, Sarah L. Catz.
To view or add a comment, sign in
-
𝐀 𝐛𝐞𝐚𝐮𝐭𝐢𝐟𝐮𝐥 𝐢𝐧𝐭𝐞𝐫𝐬𝐞𝐜𝐭𝐢𝐨𝐧 𝐨𝐟 𝐬𝐜𝐢𝐞𝐧𝐜𝐞 𝐚𝐧𝐝 𝐚𝐫𝐭. In the age of AI, it was refreshing to see a hand-painted illustration capture the essence of our recent PNAS paper. This beautiful piece by Agnieszka Dutkowska-Zuk (Paweł Żuk's partner, huge thanks for your dedication, Agnieszka!) depicts the core idea behind our research: how a modified Kirchhoff's Law can describe ion movement at junctions, akin to the the rules of traffic flow at intersections. Our paper (a non-specialist summary here: https://lnkd.in/gbwvucDe) proposes a framework using a modified Kirchhoff's Law to predict how ions navigate complex networks.
To view or add a comment, sign in
-
TheaTech - Artificial Intelligence Developer | Cognizant - Senior Process Executive | TedX Kazimierz - Brand Manager | AI Research Community Member - T3 AI'LE
🌟 Let's bring J.A.R.V.I.S and F.R.I.D.A.Y into life! :) I've just witnessed something that Tony Stark would be proud of. While sharing a cup of coffee with my girlfriend, I wrote a model to let two of the most sophisticated AIs, Anthropic's Claude and OpenAI, have a chat about philosophy. 🤖💬 It was like watching two digital sages in deep discourse, and believe it or not, as we were about to finish our coffee, Claude and OpenAI reached a consensus, a harmonious conclusion in the realm of thought and reasoning. Stark's AI pals would've been thrilled to join this conversation! 🔗 For those intrigued by the musings of our contemporary digital philosophers, check out their full tête-à-tête here: (https://lnkd.in/dR3Uv2_g) 🖼️ Click to see the simple yet profound visual representation of their dialogue in an example, encapsulating a significant leap towards a future where AIs can engage in discussions that have historically been uniquely human. 🤔 Imagine the possibilities - are we on the brink of our own Stark-inspired AI era? #AI #Marvel #OpenAI #Anthropic #Technology #Innovation #FutureIsHere
A Glimpse into AGI’s Future: The conversation of “ Claude ” and “ OpenAI “
link.medium.com
To view or add a comment, sign in
-
From the data center to the tactical edge, securely push the boundaries of what is possible. The Raven by GDIT collects and distributes mission critical information at the edge. Powered by AI systems, the Raven enables data-driven information to be disseminated to first responders, analysts and decision makers in real-time. Learn more at https://ow.ly/J9Sj30sBRai.
To view or add a comment, sign in
-
Theory of Mind Meets LLMs: Hypothetical Minds for Advanced Multi-Agent Tasks In the ever-evolving landscape of artificial intelligence
Theory of Mind Meets LLMs: Hypothetical Minds for Advanced Multi-Agent Tasks
openexo.com
To view or add a comment, sign in
-
Happy to talk about Gen AI, Observability, and a tad bit about how the term has been co-opted in the ML community (except in a way that's actually good, unlike the legacy APM vendors rebranding things to fit a cooler word)
Excited to be headed to Monitorama this week! Austin Parker and Phillip Carter are giving talks on OpenTelemetry and Generative AI (a pragmatist's approach to making LLMs work for you, with observability), and I'm stoked for three days of going deep in this space. LMK if you'll be around and want to say hi!
To view or add a comment, sign in
-
An interesting talk on the nature of AI- has it truly evolved into a full- fledged generative intelligence as many believe (and some fear)) or is just an ”autocomplete on steroids”… Not too long, Melanie Mitchell’s talk (preceded by the introductory remarks about her impressive scientific achievements) starts from the 5th minute and gives quite a few enjoyable insights and enlightening examples. https://lnkd.in/dH52-Qj6
The Future of Artificial Intelligence
https://www.youtube.com/
To view or add a comment, sign in
-
I have curated a code dataset (297k) in alpaca format by combining several code datasets. Take a look at my dataset on Hugging Face🤗. Also, I fine-tuned Unsloth AI's unsloth/llama-3-8b-bnb-4bit model with the curated dataset. Given the size of the dataset and the model being my first fine-tuned model, it is still subject to improvement. https://lnkd.in/d6PbeRpX https://lnkd.in/dc7Qw2Sg
candenizkocak/code-alpaca-297k · Datasets at Hugging Face
huggingface.co
To view or add a comment, sign in