don't think so, at least the reason why I chose this example is that it feels superhuman, or at least only within the capabilities of a very small group of humans
SAFETY
-
AI Event at the White House: Executive Order and Censorship
By
–
An event titled "Safe, Secure, and Trustworthy Artificial Intelligence" will be hosted by the White House on October 30. There will reportedly be a ‘sweeping’ AI executive order that day. Trustworthy Artificial Intelligence is synonymous with censorship. Proceed with caution.
-
Will larger language models eventually enable critical infrastructure attacks?
By
–
Imagine this prompt: "write me Python code to disable the NYC subway system" obviously gpt-4 can't do this now. It'll refuse, but even if we jailbreak it, it'll answer incorrectly but if we keep training bigger & better language models, won't one eventually be able to do this?
-
MLCommons launches AI Safety working group with HELM framework
By
–
An exciting step toward responsible AI: @MLCommons just announced the creation of the AI Safety working group which will focus on developing safety benchmarks for LLMs using the groundbreaking HELM framework from @StanfordCRFM
. Congrats @percyliang & team! https://
mlcommons.org/en/news/format
ion-ai-safety-working-group/
… -
Death Unsolved: AGI and Existential Questions
By
–
Really? Death is not solved yet, as far as we know. https://
nickbostrom.com/fable/dragon #AGIFirst -
BSI Study: Global Mistrust towards AI and French Concerns
By
–
International BSI Study: "The World, France and AI, between Mistrust and Questions" https://actuia.com/actualite/etude-internationale-bsi-le-monde-la-france-et-lia-entre-mefiance-et-interrogations/
… #AI #artificialintelligence -
ChatGPT DALLE3 Censorship Differences Across Platforms
By
–
If you run this prompt in ChatGPT with Dalle3 it will censor and not allow it. So interesting to see it’s possible through the edge browser without censoring
-
Data Poisoning Attacks Compromise AI Models During Training
By
–
'Once AI developers scrape the internet to get more data to tweak an existing AI model or build a new one, these poisoned samples make their way into the model’s data set and cause it to malfunction.'
-
Partnership on AI publishes foundation models safety guidelines
By
–
On This Day 6 years ago: the first general meeting of the Partnership on AI took place in Berlin. PAI funds studies and publishes guidelines on questions of AI ethics and safety.
It just published a set of guidelines for the safe deployment of foundation models: -
$10M AI Safety Fund and Frontier Model Forum Leadership
By
–
Announcing a $10M AI Safety fund, and the selection of the Executive Director for the Frontier Model Forum: