Six months ago, many top AI researchers called for a pause on the development of anything more powerful than GPT-4. Many say they are still deeply concerned, but more about disinfo, job displacement, and the rising power of big tech than existential risks.
SAFETY
-
Anthropomorphic AI Design Poses Significant Safety Risks Report
By
–
If you haven't read @Public_Citizen
's new report about anthropomorphic design and AI systems, I highly recommend it. https://
citizen.org/article/chatbo
ts-are-not-people-dangerous-human-like-anthropomorphic-ai-report/
… -
Meta Launches Human-Like AI Characters Amid Safety Concerns
By
–
During #MetaConnect, Meta unveiled new generative AI tools for users, developers and businesses as well as dozens of new celebrity AI "characters." Meanwhile, consumer advocates are warning of the dangers of designing AI to appear “human-like." My latest: https://
digiday.com/media/meta-add
s-a-human-element-to-ai-while-others-warn-it-all-could-be-too-human-like/
… -
LLM Trustworthiness Risks in Critical Domain Deployment
By
–
“Everyone seems to think LLMs are perfect and capable, compared with other models. That’s very dangerous, especially if people deploy these models in critical domains.” — @sanmikoyejo https://
hai.stanford.edu/news/how-trust
worthy-are-large-language-models-gpt
… (3/3) -
AI Emergent Abilities May Be Overestimated Says Research
By
–
.
@sanmikoyejo and his research team show current perceptions of AI’s emergent abilities might be overinflated: https://
hai.stanford.edu/news/ais-osten
sible-emergent-abilities-are-mirage
… (2/3) -
Funders’ Guide to Preparing Society for Generative AI
By
–
Philanthropy & generative AI: @SSIReview offers 10 things funders can do to help the existing field of tech-related nonprofits—and society at large—better prepare. https://
ssir.org/articles/entry
/10_ways_funders_can_address_generative_ai_now
… (1/2) -
ChatGPT Real-Time News: Benefits and Misinformation Risks
By
–
This finally makes ChatGPT useful for news and real-time information.
— Marty Swant (@martyswant) 27 septembre 2023
On the other hand, what new risks does this now create with hallucinations & misinformation? https://t.co/I2m1d3EXcSThis finally makes ChatGPT useful for news and real-time information. On the other hand, what new risks does this now create with hallucinations & misinformation?
-
Stanford researchers advocate coordinated AI safety in biomedicine
By
–
How do we ensure AI in biomedicine remains safe and equitable? Stanford researchers say top-down coordination is necessary between regulatory agencies and institutions to develop a comprehensive nationwide plan.
-
AI Photo Editing Without Consent: Digital Soma Problem
By
–
“…what happens when we can’t remember if we used the AI on our photos or not? What if a future upgrade tweaks our photos and we don’t even know?” Like a digital version of “soma” in Brave New World, but for memory & without our consent. A wild, worrisome & realistic concept.
