AI & LLMs Research

AI & LLMs Research

12 bookmarks
Custom sorting
[2303.15056] ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks
[2303.15056] ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks
It's likely that most people won't be able to understand the five sentence abstract of this study, but that would be the point. There is serious science behind all of the talk about AI, and it provides the best indication of AI's implications. Even if you don't understand it completely. In this specific case, people working crowdsourced gig jobs are going to be out of work and the corporations that profit from them will be down one income stream
[2303.15056] ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks
KG Medical Knowledge Poisoning
KG Medical Knowledge Poisoning
Chart to accompany nature article showing how a malicious actor can poison medical data by inserting hoax papers maks it into LLM training data
KG Medical Knowledge Poisoning
Poisoning medical knowledge using large language models - Nature Machine Intelligence
Poisoning medical knowledge using large language models - Nature Machine Intelligence
Most readers will only understand this abstract on a surface level, but that is enough to demonstrate the potential harm to knowledge. More evidence that the concept of "source reliability" is dead. Of course "malicious paper's " are just part of the library to history teachers
Poisoning medical knowledge using large language models - Nature Machine Intelligence
The corruption risks of artificial intelligence 2022 Report of Transparency International of the European Union
The corruption risks of artificial intelligence 2022 Report of Transparency International of the European Union

Just one sentence from this report encapsulates its thesis "Autocratic regimes and a weak rule of law further exacerbate the risk that AI in these societies will be deployed in a corrupt manner, such as by a political or economic leadership seeking self-enrichment or a consolidation of their grip on power via the illegitimate suppression of opposition.

The corruption risks of artificial intelligence 2022 Report of Transparency International of the European Union
LLM03: Training Data Poisoning - OWASP Top 10 for LLM & Generative AI Security
LLM03: Training Data Poisoning - OWASP Top 10 for LLM & Generative AI Security
Reading the examples of LLM vulnerability through the lens of Yochai Benkler's description of a "propaganda feedback loop" makes it clear that AI can aggressively weaponize disinformation campaigns already benefiting from a pre-existing media environment dynamic
A malicious actor is able to perform direct injection of falsified, biased or harmful content into the training processes of a model which is returned in subsequent outputs. An unsuspecting user is indirectly injecting sensitive or proprietary data into the training processes of a model which is returned in subsequent outputs.
A malicious actor or competitor intentionally creates inaccurate or malicious documents which are targeted at a model’s training data in which is training the model at the same time based on inputs. The victim model trains using this falsified information which is reflected in outputs of generative AI prompts to it’s consumers.
The embedding process
Fine-tuning
Pre-training data
LLM03: Training Data Poisoning - OWASP Top 10 for LLM & Generative AI Security
Poisoning Web-Scale Training Datasets is Practical
Poisoning Web-Scale Training Datasets is Practical
Although the research is based in computer science of LLMs, history educators with experience of persistent myths, marginalized voices and manipulated narratives can see the implications of the capacity of data to be manipulated in this way
Poisoning Web-Scale Training Datasets is Practical
AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably - Scientific Reports
AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably - Scientific Reports
This study examined whether non-expert readers could reliably differentiate between AI-generated poems and those written by well-known human poets.
As AI-generated text continues to evolve, distinguishing it from human-authored content has become increasingly difficult. This study examined whether non-expert readers could reliably differentiate between AI-generated poems and those written by well-known human poets
AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably - Scientific Reports