SAFETY - AI Dynamics

AI Development Requires Caution in Cybersecurity Implementation

By

–

14 August 2023 0h00

AI – proceed with caution
Here’s why: https://
ow.ly/3wMH50Ppa0x #sponsored #ncc_iiot #cybersecurity #cybercommunity #ai #artificialintelligence #networksecurity @IotCyber @LucianIlie15 @avrohomg via @fogoros

→ View original post on X — @fogoros

14 August 2023

Aircraft Incident: Safety Concerns for Pilot and Ground Population

By

@christinelu

–

13 August 2023 22h55

hope the pilot survived and no one on the ground was hurt.

→ View original post on X — @christinelu

13 August 2023

Sophia’s Mimicry Cannot Replace True Intelligence

By

@fchollet

–

13 August 2023 22h12

For instance, Sophia mimics human behavior in order to fake being intelligent. But if you try to get Sophia to pick up a brand new skill or solve a brand new problem in the absence of its human programmers, it will be useless. It is not intelligent.

→ View original post on X — @fchollet

13 August 2023

Behavior Mimicry Does Not Equal True Intelligence

By

@fchollet

–

13 August 2023 22h09

Behavior != Intelligence. You can mimic *behavior* in order to fake being intelligent, but you can't "simulate intelligence". Either you have what intelligence represents (efficient novel skill-acquisition ability) or you don't.

→ View original post on X — @fchollet

13 August 2023

AI Security Risk: Password Hacking Through Keystroke Listening

By

@alliekmiller

–

13 August 2023 22h03

I remember when the biggest news in tech was that they released redhead emojis And now we’re taking about AI hacking your password by listening to you typing

→ View original post on X — @alliekmiller

13 August 2023

Mission Impossible AI Portrayal Wildly Implausible Entertainment

By

@mjasay

–

13 August 2023 19h53

Ditto in the new Mission Impossible movie. Rogue AI and so wildly implausible that it was only enjoyable once I recognized that *everything* about the movie is implausible, and gave in. 🙂 https://
washingtonpost.com/technology/202
3/07/28/mission-impossible-ai-not-realistic/
…

→ View original post on X — @mjasay

13 August 2023

Trustworthy LLMs: Key Categories for Assessment Overview

By

@dair_ai

–

13 August 2023 17h37

10/ Trustworthy LLMs – a comprehensive overview of important categories crucial for assessing LLM trustworthiness, including reliability, safety, fairness, resistance to misuse, explainability and reasoning, adherence to social norms, and robustness.

→ View original post on X — @dair_ai

13 August 2023

Synthetic Data Reduces Sycophancy in Large Language Models

By

@dair_ai

–

13 August 2023 17h37

7/ Synthetic Data Reduces Sycophancy in LLMs – proposes fine-tuning on simple synthetic data to reduce sycophancy in LLMs; finds that LLMs repeat the user’s view even when the opinion is wrong.

→ View original post on X — @dair_ai

13 August 2023