AI safety seems hard to measure

FromCold Takes Audio

Start listening View podcast show

AI safety seems hard to measure

FromCold Takes Audio

ratings:

Length:

22 minutes

Released:

Dec 8, 2022

Format:

Podcast episode

Description

Four analogies for why "We don't see any misbehavior by this AI" isn't enough.https://www.cold-takes.com/ai-safety-seems-hard-to-measure/

Released:

Dec 8, 2022

Format:

Podcast episode

Titles in the series (50)

Amateur read-throughs of blog posts on Cold-Takes.com, for those who prefer listening to reading. Available on Apple, Spotify, Google Podcasts, and anywhere else you listen to podcasts by searching Cold Takes Audio.

Skip carousel

All Possible Views About Humanity's Future Are Wild by Cold Takes Audio
15 min listen
Digital People Part 1 - intro by Cold Takes Audio
11 min listen
Summary of history (empowerment and well-being lens) by Cold Takes Audio
18 min listen
Are we "trending toward" transformative AI? (How would we know?) by Cold Takes Audio
16 min listen
How to make the best of the most important century? by Cold Takes Audio
22 min listen
Digital People Part 3 - Implications by Cold Takes Audio
13 min listen
The Duplicator by Cold Takes Audio
15 min listen
Forecasting Transformative AI, Part 1: What Kind of AI? by Cold Takes Audio
18 min listen
Digital People Part 2 (optional) - FAQ by Cold Takes Audio
25 min listen
This Can't Go On by Cold Takes Audio
16 min listen
AI Timelines: Where the Arguments, and the ”Experts,” Stand by Cold Takes Audio
19 min listen
Has Life Gotten Better? by Cold Takes Audio
9 min listen
Forecasting transformative AI: the ”biological anchors” method in a nutshell by Cold Takes Audio
29 min listen
Why AI alignment could be hard with modern deep learning (guest post by Ajeya Cotra) by Cold Takes Audio
29 min listen
Call to Vigilance by Cold Takes Audio
6 min listen
Forecasting transformative AI: what's the burden of proof? by Cold Takes Audio
29 min listen
Rowing, Steering, Anchoring, Equity, Mutiny by Cold Takes Audio
22 min listen
Unraveling the evidence about violence among very early humans by Cold Takes Audio
24 min listen
Falling everyday violence, bigger wars and atrocities: how do they net out? by Cold Takes Audio
14 min listen
Did life get better during the pre-industrial era? (Ehhhh) by Cold Takes Audio
20 min listen
Has Life Gotten Better?: the Post-Industrial Era by Cold Takes Audio
14 min listen
Pre-agriculture gender relations seem bad by Cold Takes Audio
20 min listen
Visualizing Utopia by Cold Takes Audio
21 min listen
Was life better in hunter-gatherer times? by Cold Takes Audio
22 min listen
The Bayesian Mindset by Cold Takes Audio
32 min listen
Why Describing Utopia Goes Badly by Cold Takes Audio
18 min listen
Future-Proof Ethics by Cold Takes Audio
27 min listen
Minimal-trust investigations by Cold Takes Audio
20 min listen
Debating myself on whether “extra lives lived” are as good as “deaths prevented” by Cold Takes Audio
20 min listen
Stakeholder Management and Cost Disease by Cold Takes Audio
18 min listen

More Episodes from Cold Takes Audio

Skip carousel

What AI companies can do today to help with the most important century by Cold Takes Audio
18 min listen
Jobs that can help with the most important century by Cold Takes Audio
31 min listen
Spreading messages to help with the most important century by Cold Takes Audio
20 min listen
How we could stumble into AI catastrophe by Cold Takes Audio
29 min listen
Transformative AI issues (not just misalignment): an overview by Cold Takes Audio
25 min listen
Racing Through a Minefield: the AI Deployment Problem by Cold Takes Audio
21 min listen
High-level hopes for AI aligment by Cold Takes Audio
24 min listen
Why Would AI "Aim" To Defeat Humanity? by Cold Takes Audio
46 min listen
The Track Record of Futurists Seems ... Fine by Cold Takes Audio
21 min listen
Nonprofit Boards are Weird by Cold Takes Audio
25 min listen
AI Could Defeat All Of Us Combined by Cold Takes Audio
24 min listen
Useful Vices for Wicked Problems by Cold Takes Audio
25 min listen
Ideal governance (for companies, countries and more) by Cold Takes Audio
18 min listen
Debating myself on whether “extra lives lived” are as good as “deaths prevented” by Cold Takes Audio
20 min listen
The Wicked Problem Experience by Cold Takes Audio
15 min listen
Learning By Writing by Cold Takes Audio
16 min listen
Defending One-Dimensional Ethics by Cold Takes Audio
26 min listen
Future-Proof Ethics by Cold Takes Audio
27 min listen
Stakeholder Management and Cost Disease by Cold Takes Audio
18 min listen
Reader reactions and update on "Where's Today's Beethoven" by Cold Takes Audio
18 min listen
Why it matters if "ideas get harder to find" by Cold Takes Audio
11 min listen
Where's Today's Beethoven? by Cold Takes Audio
37 min listen
The Bayesian Mindset by Cold Takes Audio
32 min listen
Visualizing Utopia by Cold Takes Audio
21 min listen
Why Describing Utopia Goes Badly by Cold Takes Audio
18 min listen
Minimal-trust investigations by Cold Takes Audio
20 min listen
Did life get better during the pre-industrial era? (Ehhhh) by Cold Takes Audio
20 min listen
Falling everyday violence, bigger wars and atrocities: how do they net out? by Cold Takes Audio
14 min listen
Rowing, Steering, Anchoring, Equity, Mutiny by Cold Takes Audio
22 min listen
Unraveling the evidence about violence among very early humans by Cold Takes Audio
24 min listen

Related podcast episodes

Skip carousel

AI & Hype & Security (Oh My!) - Caleb Sima - ASW #284: A lot of AI security has nothing to do with AI -- things like data privacy, access controls, and identity are concerns for any new software and in many cases AI concerns look more like old-school API concerns. But...there are still important aspects... by Application Security Weekly (Video)
33 min listen
AI & Hype & Security (Oh My!) - Caleb Sima - ASW #284: A lot of AI security has nothing to do with AI -- things like data privacy, access controls, and identity are concerns for any new software and in many cases AI concerns look more like old-school API concerns. But...there are still important aspects... by Security Weekly Podcast Network (Video)
33 min listen
AI & Hype & Security (Oh My!) & Hacking AI Bias - Caleb Sima, Keith Hoodlet - ASW #284: A lot of AI security has nothing to do with AI -- things like data privacy, access controls, and identity are concerns for any new software and in many cases AI concerns look more like old-school API concerns. But...there are still important aspects... by Security Weekly Podcast Network (Audio)
65 min listen
Anthropic Researchers Uncover "Sleeper Agent" Capabilities in AI Models by AI Chat: ChatGPT & AI News, Artificial Intelligence, OpenAI, Machine Learning
10 min listen
7 Falsehoods About Outsourcing Your Call Center by Advice from a Call Center Geek!
18 min listen
HAR files, Okta breach, EO on AI, Ransomware, Solarwinds CISO charged, and Bagels! - ESW #338: Oh, the HARror! Sanitizing HAR files is not as easy as some might lead you to believe. CISA funds Cyber.org for K-12 cyber education and ORNL creates a Center for AI Security Research (CAISER). Cloudflare creates a tool out of spite, and CISA creates... by Security Weekly Podcast Network (Video)
71 min listen
HAR files, Okta breach, EO on AI, Ransomware, Solarwinds CISO charged, and Bagels! - ESW #338: Oh, the HARror! Sanitizing HAR files is not as easy as some might lead you to believe. CISA funds Cyber.org for K-12 cyber education and ORNL creates a Center for AI Security Research (CAISER). Cloudflare creates a tool out of spite, and CISA creates... by Enterprise Security Weekly (Video)
71 min listen
ISACA Industry Spotlight | Ali Pabrai: There is no denying the passion that ecfirst's CEO, Ali Pabrai has for cybersecurity. In this ISACA Podcast, Ali tells ISACA's Hollee Mangrum-Willis that after all his years in the industry, he is still more excited than a two-year-old at the entranc... by ISACA Podcast
41 min listen
Exploitable by Default: Vulnerabilities in GPT-4 APIs and “Superhuman” Go AIs with Adam Gleave of Far.ai: In this episode, Nathan sits down with Adam Gleave, founder of Far AI, for a masterclass on AI exploitability. by "The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis
104 min listen
0025: The Concerning AI Summer is Over: Some things we talked about: Companies developing narrow AI, not giving one thought about AI safety, because just getting the thing to work at all is really hard. Self-driving cars and how fast they’re progressing Difference between Open AI and MIRI in approach Weaponized offense and defense Eliezer’s thought about that missing the point http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0025-2016-09-04.mp3 by Concerning AI | Existential Risk From Artificial Intelligence
20 min listen
Ep. 441 - Government/AI Truth Manipulating Tools by Dangerous World Podcast
34 min listen
The big problems with Apple’s photo scanning system: Apple’s controversial new child safety tool and we look at the rise of office tribes by The WIRED Podcast
43 min listen
#40: Joel Beal CEO & Co-Founder of Alloy by Leaders in Value Chain
59 min listen
Cyber & Crypto Podcast - Episode 107: Episode 107 by Cybersecurity & Cryptocurrency Podcast with Eric English
31 min listen
How GenAI Can Improve SecOps - Ely Kahn - ESW #359: We've talked about generative AI in a general sense on our podcast for years, but we haven't done many deep dives into specific security use cases. That ends with this interview, as we discuss how generative AI can improve SecOps with Ely Kahn. Some... by Security Weekly Podcast Network (Video)
30 min listen
How GenAI Can Improve SecOps - Ely Kahn - ESW #359: We've talked about generative AI in a general sense on our podcast for years, but we haven't done many deep dives into specific security use cases. That ends with this interview, as we discuss how generative AI can improve SecOps with Ely Kahn. Some... by Enterprise Security Weekly (Video)
30 min listen
Wed Nov-29-2023: AI-generated Work, Chief AI Officers, and Misinformation by Business of Tech
13 min listen
Israel said to have tipped the US off concerning Kaspersky risks. Accenture databases exposed. Deloitte breach may be worse than initially thought.: Israel said to have tipped the US off concerning Kaspersky risks. Accenture databases exposed. Deloitte breach may be worse than initially thought. by CyberWire Daily
18 min listen
We Take Privacy, Not Our CISO, Seriously: All pictures and links for this episode can be found on CISO Series () We're looking for the one company brave enough to say they don't care about privacy on the latest episode of CISO/Security Vendor Relationship Podcast. This episode was recorded... by CISO Series Podcast
46 min listen
11| AI, Risk, Fairness & Responsibility — John Zerilli by MULTIVERSES
100 min listen
AI Revolutionizing Workforce: IBM Study Shows 40% Skill Obsolescence in 3 Years by The Mark Cuban Podcast
12 min listen
Amazon Rings up the Police - DTNS SPECIAL: Chris Ashley from the SMR Podcast chats with Tom about Amazon's cooperation with Police on promoting Ring doorbell video sharing, how that might feed into Amazon's imperfect Recognition facial recognition system and what we need to do to make... by Daily Tech News Show
24 min listen
Defensive Security Podcast Episode 222: https://www.csoonline.com/article/3285982/data-protection/4-reasons-why-cisos-must-think-like-developers-to-build-cybersecurity-platforms.html https://www.csoonline.com/article/3287655/phishing/stop-training-your-employees-to-fall-for-phishing-attacks.... by Defensive Security Podcast - Malware, Hacking, Cyber Security & Infosec
52 min listen
lynsey wolf, conducting insider threat investigations, CASB and UEBA utlization to good use.: Show Topic Summary (less than 300 words) Insider threat still exists, Lynsey Wolf talks with us about HR’s role in insider threat, how prevalent investigations are in the post-pandemic work from home environment. Questions... by BrakeSec Education Podcast
94 min listen
Junior Cyber People: All links and images for this episode can be found on CISO Series () There are so few jobs available for junior cybersecurity professionals. Are these cyber beginners not valued? Or are we as managers not creating the right roles for them to improve... by Defense in Depth
29 min listen
Episode 73 Jane Lee by Fraudish
38 min listen
Rethinking the CISO Model, Edge Ecosystem Insights - Nathan Case, Theresa Lanowitz - ESW #325: The traditional concept of the CISO may literally be 'too much', according to Nathan Case. It's based on systems of control and unrealistic assumptions that don't survive contact with real life. In this conversation, we'll discuss what the top... by Security Weekly Podcast Network (Audio)
154 min listen
7 Questions, 7 Mistakes, and a CISO Checklist - BSW #253: In the Leadership and Communications section, 7 Pressing Cybersecurity Questions Boards Need to Ask, 7 mistakes CISOs make when presenting to the board (Let's see if those align), CISO Checklist for Offboarding Security Staff, and more! Visit ... by Security Weekly Podcast Network (Video)
29 min listen
373: We’re a Happiness Podcast by Grumpy Old Geeks
60 min listen
If A.I. Is Not Regulated, 'We Are Going to Be Sorry': Tom Siebel, CEO of C3.ai, thinks business needs to adopt artificial intelligence at scale, but with restrictions. by Leadership Next
26 min listen

Discover this podcast and so much more

AI safety seems hard to measure

AI safety seems hard to measure

Description

Titles in the series (50)

More Episodes from Cold Takes Audio

Related podcast episodes