The Data Hunger Games
Your trusted source for insights on the world of responsible AI and AI policy. July 29th, 2024. Issue 28.
A Note from Paige
I have two updates this week:
If you received an email offering 20% off to subscribe to this publication, it was a scam. You can read more here.
On a trial basis, I will be sending AI Ethics & Policy Insights out twice a week. Simply put, there is a lot of important news in this area and things are moving quickly. After careful evaluation I’ve decided that the best option to ensure readers are regularly informed without sending a weekly newsletter the size of the Sunday Times is a twice weekly publication on Mondays and Thursdays. I hope you’ll find it valuable and informative. As always, I’m open to feedback, and thanks for being here!
AI Ethics News
Notable news in the world of AI ethics and responsible AI.
Elon Posts AI-Altered “Parody” of Kamala Harris’s Campaign Video
On July 26th, 2024, Elon Musk posted a “parody” video of Kamala Harris’s campaign video to his more than 100M followers on X. The video includes clips from Harris’s campaign video and the “Harris for President” branding, as well as an AI generated voice over that sounded remarkably like the Vice President. In the video the voice refers to President Joe Biden’s “senility” and claims that Harris is a “diversity hire”.
👉 Why it matters: The upcoming election is like none we’ve ever seen because this time around AI is available broadly and can be used by bad actors to spread mis and disinformation. The sharing of this video highlights several challenges and questions. First, whether it was responsible for Musk to share the video without noting that it’s parody. Second, that we will certainly seen an uptick in AI generated political material coming from citizens. Third, at what point a “parody” AI video becomes mis or disinformation.
Claude’s Curious Crawls
Anthropic’s ClaudeBot web crawler, which gathers training data for the company’s AI models, appears to be…picking favorites? iFixit CEO Kyle Wiens tweeted at Anthropic indicating that their ClaudeBot hit their servers more than a million times in a 24 hour period. Is this the data hunger games?
👉 Why it matters: This violates iFixit’s Terms of Use and consumes devops resources. It highlights the challenge companies face in preventing their content from being used as training data by web crawlers. Anthropic's FAQ states that companies must block the crawler using a robots.txt file, placing the responsibility on companies to protect their content rather than on AI creators to respect terms.
Video Game Actors on Strike
Members of SAG-AFTRA are back in the news as video game voice actors and performers were unable to come to an agreement with major video game companies on protections for the Union against AI.
👉 Why it matters: The major sticking point with AI is that the actors and performers are worried that the large companies will use their work to train the AI that will then put them out of work in the future. This is another voice among the growing concerns about AI being trained by humans and then replacing the humans that trained them.
*Paige works for GitHub, which is owned by Microsoft.
AI in the Wild
NEW! This section will highlight new and interesting uses of AI, so you can stay up-to-date on how the technology is changing.
Meta Champions Open Source with Llama
Meta has released its large language model (LLM) Llama 3.1 for free as a major show of support for the development of open source AI. This move has put the company at the center of the “safe vs. open” AI debate, and it’s invigorating the AI community.
👉 Why it matters: When Meta uses an LLM for Meta AI, it implements safeguards to prevent harm. Openly releasing the model allows others to remove these safeguards but also supports open AI development. Researchers can access and test the model independently, and developers can benefit from Meta's work.
AI Policy Beat
A look at what’s happening in the world of AI policy.
Apple Joins the US AI Agreement
Apple joined the Biden administration’s voluntary artificial intelligence guidelines, joining 15 large companies who have signed onto the agreement since it was revealed in October. This announcement follows the release of “Apple Intelligence”, Apple’s AI platform and announcement to integrate AI into the iPhone and other devices.
👉 Why it matters: In the absence of federal AI regulation, safety of AI is largely determined by the creator the technology. The agreement is voluntary and contains guidance for the safe development of AI, but it also serves to hold companies accountable to their agreement to protect people from AI harms. It’s more symbolic than anything, but it remains a good sign that AI providers continue to agree that safety is critical.
Senators Lean on OpenAI Amidst Safety Concerns
Senators are now leaning on OpenAI to release information about the specifics of its efforts to build safe AI. This follows on the heels of the OpenAI employees who not only notified the federal government that OpenAI was bypassing safety protocols to get products to market, but that OpenAI’s NDA was overly broad and violated whistleblower laws.
👉 Why it matters: OpenAI and other AI providers are under pressure to disclose their safety standards to assure the public that safety, not profit, is prioritized. OpenAI voluntarily signed the Biden administration's AI safety agreement. Failing to meet safety standards could tip the scales toward higher regulation, potentially stifling innovation.
Spotlight on Research
Towards Robust Speech Representation Learning for Thousands of Languages
Researchers created XEUS, a Cross-Lingual Encoder for Universal Speech, to support over 7,000 languages. XEUS consistently outperformed or matched the results of other self-supervised learning models.
Supporting the Digital Autonomy of Elders Through LLM Assistance
Jesse Roberts and colleagues explore how an LLM can support the elderly who are separated by the digital divide or struggle to access internet resources.Akash Wasil and colleagues explore how the federal government can enhance its AI emergency preparedness by detecting and preparing for AI national security threats.
WATCH: Lattice AI’s BIG HR No-No
Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.
Tiktok failed to load.Enable 3rd party cookies or use another browser
Let’s Connect:
Connect with me on LinkedIn.
Looking for a speaker at your next AI event? Email thepaigelord@gmail.com.
Email thepaigelord@gmail.com if you want to connect 1:1.