can this feature (also send as DM) be enabled by default yet?

Always Know who Controls Who….

Explore, Establish, Exploit: Red Teaming Language Models from Scratch paper page: huggingface.co/papers/2306.0… Deploying Large language models (LLMs) can pose hazards from harmful outputs such as toxic or dishonest speech. Prior work has introduced tools that elicit harmful outputs in order to identify and mitigate these risks. While this is a valuable step toward securing language models, these approaches typically rely on a pre-existing classifier for undesired outputs. This limits their application to situations where the type of harmful behavior is known with precision beforehand. However, this skips a central challenge of red teaming: developing a contextual understanding of the behaviors that a model can exhibit. Furthermore, when such a classifier already exists, red teaming has limited marginal value because the classifier could simply be used to filter training data or model outputs. In this work, we consider red teaming under the assumption that the adversary is working from a high-level, abstract specification of undesired behavior. The red team is expected to refine/extend this specification and identify methods to elicit this behavior from the model. Our red teaming framework consists of three steps: 1) Exploring the model's behavior in the desired context; 2) Establishing a measurement of undesired behavior (e.g., a classifier trained to reflect human evaluations); and 3) Exploiting the model's flaws using this measure and an established red teaming methodology. We apply this approach to red team GPT-2 and GPT-3 models to systematically discover classes of prompts that elicit toxic and dishonest statements. In doing so, we also construct and release the CommonClaim dataset of 20,000 statements that have been labeled by human subjects as common-knowledge-true, common-knowledge-false, or neither.

AI-generated QR codes are the next big thing in marketing. And it's completely free. Here's my full tutorial on how to do it in 5 simple steps: (Try scanning the image below)

Pixel 7a tablet watch Nest < Pixel Fold 😂

Saw ChatGPT plugins coming in 2015. Next step? Own your vision

I've spent 1000's of hours building ML models over the last couple of years. Here are some tips that would have made me work 10x faster (that you can read in 2 minutes):

just finished Tyranny of Merit & picked up How not to be Wrong. will get back to biweekly publishing once time permits, but a preview: * Raising capital is just the first step; deploying capital is hard * Op-ed: America’s Olympic disappointment & the hubris of college athletics

second article on blockchain, bitcoin and why decentralization deserves your attention. P.S. will try to get a (bi)weekly publication going, follow my medium to stay tuned! link.medium.com/lcI9Z9z2Qhb

First dabble in publishing & lessons learned running an early-stage startup. Stay tuned for more! #founderwellness @justinkan⁩ link.medium.com/fhsuvin0Khb

How to represent part-whole hierarchies in a neural network: peeking into the future of transformer-based vision systems arxiv.org/abs/2102.12627 by @geoffreyhinton

“Consider instead what would be possible if these brands connected the data they have from the production process all…” fastcompany.com/90507470/the…

The equivalent of open source in Software 2.0 land are open datasets. But while plenty of former exists little of high quality latter does.

Impressive and surprising openai.com/blog/dall-e/ I use those words sparingly

Honestly, unless there’s a vaccine or cure, how are we going to have multiple “waves” of coronavirus. Unless a society is completely immobile & closed off from rest of the world, there’s always a non-zero chance a carrier imports the virus, so we’re forever in the big first wave.

Does World Bank's latest paper (bit.ly/2ZXKyKU) on China’s productivity slowdown hint a need to curtail its foreign debt load rather than deepening capital towards non-productive GDP growth?

Stoked for what the future holds in #QuantumBlueprint Launch to the Future: Quantum Internet — The Department of Energy & UChicago youtu.be/cR0wVCs9DxI

limited streaming spots available for industry panel with director of medical AI startup and Facebook AI engineer at Duke University in 2 weeks. DM me for invitation!

You can now binge watch the entire @a16z Crypto Startup School program from start to finish: a16z.com/cryptostartupschool The final 2 lessons get into important details of fundraising & regulation.

It’s 2020 and the racial diversity in Forbes’ newly announced “25 Future Billion-Dollar Startups” is appalling. 21 Whites, 4 Indians - where are the Blacks, Latinos? Do you really have to be a Kobe Bryant to be considered a ‘successful’ businessman, by public means?

The world needs to give credit to those who deserve and remove from those who don’t. Privilege; is an antecedent to ignorance. Ignorance could be bliss, or unworldly & pedantic arrogance. #FactCheckYourRealities #EgoandBlindspotBarriers youtu.be/CDmLI4zlBH0