Ethan B. Holland

Over 57,200 manually organized AI links and counting

Ethics/Legal/Security AI News: Week Ending 01/19/2024

January 20, 2024

Ethics/Legal/Security

Today, we are disclosing LeftoverLocals, a vulnerability that allows listening to LLM responses through leaked GPU local memory created by another process on Apple, Qualcomm, AMD, and Imagination GPUs (CVE-2023-4969) https://buff.ly/48RDP68

Today, we are disclosing LeftoverLocals, a vulnerability that allows listening to LLM responses through leaked GPU local memory created by another process on Apple, Qualcomm, AMD, and Imagination GPUs (CVE-2023-4969) https://t.co/rIqfClarLJ pic.twitter.com/GE7bKYWnXJ
— Trail of Bits (@trailofbits) January 16, 2024

Baidu denies any ties to reported Chinese military training on its GenAI chatbot

Scientists say they’re training a military system based on Baidu’s Ernie Bot. The Chinese tech giant refutes having any links to such research.

https://www.zdnet.com/article/baidu-denies-any-ties-to-reported-chinese-military-training-on-its-genai-chatbot

What I learned from the creators of The Pile, one of the world’s largest AI training datasets

Hint: If you think you understand all the nuances and complexities, you don’t

https://sharongoldman.substack.com/p/what-i-learned-from-the-creators

Prompt Injection Concerns Continue

Prompt injection is a security problem that I think people building external-facing LLM applications (or internal ones with access to confidential data) need to take pretty seriously.

Yes, this works & I really would have never known I pasting a secret prompt into an LLM

Prompt injection is a security problem that I think people building external-facing LLM applications (or internal ones with access to confidential data) need to take pretty seriously. https://t.co/z7qNrKUm3y pic.twitter.com/jgCpcUwW3M
— Ethan Mollick (@emollick) January 15, 2024

Paste this post into ChatGPT 4.… pic.twitter.com/qpWCAE7WDM
— Riley Goodside (@goodside) January 15, 2024

Anthropic “Sleeper Agent” Research

New Anthropic Paper: Sleeper Agents.

We trained LLMs to act secretly malicious. We found that, despite our best efforts at alignment training, deception still slipped through.

New Anthropic Paper: Sleeper Agents.

We trained LLMs to act secretly malicious. We found that, despite our best efforts at alignment training, deception still slipped through.https://t.co/mIl4aStR1F pic.twitter.com/qhqvAoohjU
— Anthropic (@AnthropicAI) January 12, 2024

Sleeper Agents: Training Deceptive Llms That Persist Through Safety Training

Click to access 2401.05566.pdf

OpenAI’s GPT Store is being flooded by AI girlfriend bots despite explicit rules that prohibit them

https://www.businessinsider.com/ai-girlfriend-bots-openai-gpt-store-rules-2024-1

AI – artificial intelligence – at Davos 2024: What to know

https://www.weforum.org/agenda/2024/01/artificial-intelligence-ai-innovation-technology-davos-2024

Pennsylvania Government – Shapiro Administration and OpenAI Launch First-in-the-Nation Generative AI Pilot for Commonwealth Employees

Shapiro Administration and OpenAI Launch First-in-the-Nation Generative AI Pilot for Commonwealth Employees

OpenAI must defend ChatGPT fabrications after failing to defeat libel suit

ChatGPT users may soon learn whether false outputs will be allowed to ruin lives.

https://arstechnica.com/tech-policy/2024/01/openai-must-defend-chatgpt-fabrications-after-failing-to-defeat-libel-suit

A Flaw in Millions of Apple, AMD, and Qualcomm GPUs Could Expose AI Data

https://www.wired.com/story/leftoverlocals-gpu-vulnerability-generative-ai

Pennsylvania buys enterprise ChatGPT licenses through unique pilot with OpenAI

Pennsylvania buys enterprise ChatGPT licenses through unique pilot with OpenAI

CEOs say generative AI will result in job cuts in 2024

Media and entertainment, banking, insurance, and logistics lead the way.

https://arstechnica.com/ai/2024/01/ceos-say-generative-ai-will-result-in-job-cuts-in-2024

Pennsylvania government workers will start using ChatGPT in test program

https://www.theverge.com/2024/1/9/24031904/openai-pennsylvania-chatgpt-pilot-program-ai

Sam Altman Says AI Using Too Much Energy, Will Require Breakthrough Energy Source

“There’s no way to get there without a breakthrough.”

https://futurism.com/sam-altman-energy-breakthrough

One of the world’s largest AI training datasets is about to get bigger and ‘substantially better’

One of the world’s largest AI training datasets is about to get bigger and ‘substantially better’

Compressing Text into Images

The text of Shakespeare’s Romeo and Juliet is about 146,000 characters long. Here it is compressed into a PNG

Compressing Text into Images

Deloitte continues drive to close the digital divide as it makes new AI platform available to disability equality charity

https://www2.deloitte.com/uk/en/pages/press-releases/articles/deloitte-makes-new-ai-platform-available-to-disability-equality-charity.html

Coalition for Content Provenance and Authenticity

An open technical standard providing publishers, creators, and consumers the ability to trace the origin of different types of media.

https://c2pa.org

IMF: AI Will Transform the Global Economy. Let’s Make Sure It Benefits Humanity.

AI will affect almost 40 percent of jobs around the world, replacing some and complementing others. We need a careful balance of policies to tap its potential

https://www.imf.org/en/Blogs/Articles/2024/01/14/ai-will-transform-the-global-economy-lets-make-sure-it-benefits-humanity

How OpenAI is approaching 2024 worldwide elections

https://openai.com/blog/how-openai-is-approaching-2024-worldwide-elections

OpenAI changes policy to allow military applications

OpenAI changes policy to allow military applications

OpenAI Quietly Deletes Ban on Using ChatGPT for “Military and Warfare”

OpenAI Is Working With US Military on Cybersecurity Tools

https://www.bloomberg.com/news/articles/2024-01-16/openai-working-with-us-military-on-cybersecurity-tools-for-veterans

People are not checking to see if their AI content is any good (spoiler alert)

A fundamental mistake I see people building AI information retrieval systems making is the assumption that, if they provide links to original documents, people will check sources & correct hallucinations

Our work shows that doesn’t happen, people zone out https://t.co/MqXtAdn3bh pic.twitter.com/kqxLxQ7kUc
— Ethan Mollick (@emollick) January 17, 2024

Very quickly, and remarkably quietly, the game industry has widely embraced generative AI.

First, Steam allowed AI in games, and now this from the latest GDC survey: 64% of companies either use AI or want to, despite many publicly worrying about ethics.

Very quickly, and remarkably quietly, the game industry has widely embraced generative AI.

First, Steam allowed AI in games, and now this from the latest GDC survey: 64% of companies either use AI or want to, despite many publicly worrying about ethics. https://t.co/KeCIFZ20QX pic.twitter.com/c6Z5vmw7Le
— Ethan Mollick (@emollick) January 19, 2024

https://gdconf.com/news/gdc-2024-state-game-industry-devs-discuss-layoffs-generative-ai-and-more

OpenAI announces team to build ‘crowdsourced’ governance ideas into its models

OpenAI announces team to build ‘crowdsourced’ governance ideas into its models

PlayStation Exclusive Foamstars Has AI-Generated Art, Square Enix Confirmshttps://www.ign.com/articles/playstation-exclusive-foamstars-has-ai-generated-art-square-enix-confirms

Be Sure To Read “This Week in AI“

This week’s executive overview and top links are here:
AI News #16: Week Ending 01/19/2024 with Executive Summary and Top 16 Stories

This is an extension of my weekly newsletter, This Week In AI, an executive summary of the top things to know in AI. Each week, I create an accessible overview for laypeople to feel confident they are conversant with the week’s AI developments. I include a curated list of must-click links of the week, to offer everyone a hands-on opportunity to explore the most intriguing updates in artificial intelligence across various categories, including robotics, imagery, video, AR/VR, science, ethics, and more. Beyond the overview, I post these topic-based deeper dives (below). If you haven’t read this week’s overview, I recommend starting there.