LLMs are designed to *sound* smart, which is not quite the same thing as actually being smart we all know enough bullshitters to know the difference planning, memory, reasoning, self-reflection… all key steps to get real intelligence (to be clear, smartest ppl working on it)
SAFETY
-
Inflection AI Emphasizes Pi Safety Against AI Attacks
By
–
At @inflectionAI we work very hard to make safety our top priority. Our AI, http://
Pi.ai, is not vulnerable to any of these attacks. Rather than provide a stock safety phrase, Pi will push back on the user in a polite but very clear way. @CadeMetz @kevinroose -
MIT Framework Explains Robot Failures Using Counterfactual Explanations
By
–
What happens when you purchase a robot for household tasks, but the bot fails? MIT framework helps non-technical users learn why using counterfactual explanations, then fine-tunes an ML algorithm so the robot can perform the task correctly: http://
bit.ly/3Q019Iw -
Three Cybercrime Predictions In The Age Of ChatGPT
By
–
Three Cybercrime Predictions In The Age Of ChatGPT
#AI #AIio #BigData #ML #NLU #Futureofwork @TopCyberNews @SpirosMargaris @MarshaCollier @MHiesboeck
@MHcommunicate @Fisher85M @MikeQuindazzi @NealSchaffer
http://
ow.ly/SGVI30swnpT -
AI leaders publicly acknowledge artificial intelligence dangers statement
By
–
AI leaders sign a statement to openly acknowledge the dangers of #AI https://
zd.net/42xbqyu #leadership #ethics -
Tech Leaders Call for Critical Thinking to Counter AI Deepfakes
By
–
Rules and technology can only go so far toward mitigating misuse of #AI in the information space. As technology evolves, people will need to learn skills to help manage AI.
@vilasdhar
@BradSmi
@ArvindKrishna
@SenSchumer
https://
wsj.com/articles/pro-t
ake-tech-executives-call-for-critical-thinking-to-counter-deepfakes-disinformation-11a008f7
… -
Trusting AI Companies That Warn Users About Product Risks
By
–
The more I stick around, the more I see the value of businesses that ask users to employ caution on their own products. For some reason, I just trust them more.
-
Avoiding Ethical Nightmares of Emerging Technology
By
–
How to Avoid the Ethical Nightmares of Emerging #Technology https://
bit.ly/44fLUPs via @HarvardBiz #AI #ethics -
LLM Text Adversarial Problems Easier Than Vision Recognition
By
–
I do agree that trying to solve this within LLM will be hard. But the problem seems easier to solve in practice than vision, unless people come up with a way to make an adversarial string that reads like natural language
-
Adversarial Examples in Vision vs Language Models Vulnerability
By
–
Adversarial examples have been hard to solve in vision mainly because it was imperceptible to humans and added adv noise is hard to remove. Looking at the demo, it seems relatively easy to build a filter to remove adversarial suffix before feeding the query into LLM…