Since the metadata can be removed, its absence doesn’t mean an image is not from ChatGPT or our API. Broader adoption of methods for establishing provenance and encouraging users to look for these signals are steps towards increasing the trustworthiness of digital information.
ETHICS
-
ChatGPT and DALL-E 3 now include C2PA metadata specifications
By
–
Images generated in ChatGPT and our API now include metadata using C2PA specifications. This allows anyone (including social platforms and content distributors) to see that an image was generated by our products. https://
help.openai.com/en/articles/89
12793-c2pa-in-dall-e-3
… -
Children Experience 50x More Data Than Largest Language Models
By
–
I've made that point before:
– LLM: 1E13 tokens x 0.75 word/token x 2 bytes/token = 1E13 bytes.
– 4 year old child: 16k wake hours x 3600 s/hour x 1E6 optical nerve fibers x 2 eyes x 10 bytes/s = 1E15 bytes. In 4 years, a child has seen 50 times more data than the biggest LLMs. x.com/tomosman/statu… -
Neural Networks Impressive But Human Brain Underrated
By
–
sure, neural networks make for impressive demos, but the human brain is seriously underrated
-
Surveillance Ads Reshape Politics More Than Deepfakes
By
–
The election year focus on 'deep fakes' is a distraction, conveniently ignoring the documented role of surveillance ads–or, the ability to target specific segments to shape opinion. This's a boon to Meta/Google, who've rolled back restrictions on political ads in recent years.
-
AI Deepfakes Fuel Fraud: Fake IDs and $25.6M Hong Kong Heist
By
–
AI deepfakes are all over the headlines. Two crazy new ones: -OnlyFake, a site that leverages AI to create near-instant fake IDs. -A $25.6M heist in Hong Kong, with scammers using deepfake tech to convince an employee to transfer funds.
-
Wearable Data Nudges Toward Healthier Lives Through AI
By
–
Most data collected by wearable devices is underutilized by patients and clinicians. A @StanfordHAI faculty fellow is researching ways to nudge people toward healthier lives through data-driven healthcare. Read the interview:
-
Understanding as Mapping: LLMs and Causal Model Weights
By
–
I think of understanding as a mapping between observables and causal models. Weights can be a way to represent understanding. LLMs don't always understand, but Gary is himself simulating understanding via prompt completion, which is why he always gets to the same conclusion.
-
Intercultural Accommodation in Human-Robot Interaction Research
By
–
Yes! Do you have any papers on intercultural accommodation? There has been some work on culture and HRI, but the non-verbal accommodation piece could be further explored https://
ncbi.nlm.nih.gov/pmc/articles/P
MC8062829/
… -
ID Verification Gaps Demand Governance Adaptation for AI Era
By
–
I did not realize that ID verification tools don’t query a national database. It’s time that governance adapts to the realities of the generative information age