AI Ethics & Policy Insights
The AI ethics & policy info we should know - simplified. Week of July 22, 2024. Issue 27.
AI Ethics News
Notable news in the world of AI ethics and responsible AI.
Donald Trump’s “Hands Off” Approach to AI - Republican Presidential Candidate Donald Trump has successfully been able to win the favor of many Silicon Valley names with his proposed “hands off” approach to AI regulation. The Washington Post reported that the America First Policy Institute, a think tank run by Donald Trump’s former chief economic advisor, has put forward a new AI executive order.
👉 Why it matters: If Donald Trump is elected, this new order, or a version of it, would replace the Biden Administration’s AI Executive Order. As seen in the document, this could include: - tech industry self-regulation of safety and security for AI systems - prioritizing safeguarding systems from foreign adversaries - establishing “Manhattan Projects” to create cutting-edge AI for use by the US military The Biden Administration’s AI Executive Order, by contrast, prioritizes ensuring that AI does not inflict harm on citizens of the US, and works to protect the biggest LLMs from being used by bad actors to start a cyberwar.
Lattice Gives AI Workers Rights and Employee Records, Receives Backlash - Lattice AI, a Human Resources platform founded by Jack Altman (yep, Sam’s brother), recently announced that it would “make history” as the first company to give AI agents digital records on their platform. This means that digital employees can be hired and trained on a company’s handbook, given metrics for performance, given performance reviewed, and have their work monitored by their human manager on the platform. The company’s post almost instantly received backlash, and they scrapped the idea shortly after their announcement launched.
👉 Why it matters: This was intended to be a “responsible AI” move by Lattice, as stated in the blog post by their CEO. Unfortunately, it missed the mark with one commenter calling out that this move “disrespects the humanity of your real employee. Worse, it implies that you view humans simply as ‘resources’ to be optimized and measured against machines.” By putting AI and humans on equal ground blurs the line between technology and humanity, and could create a hostile environment in which humans are pitted against AI. In this case, AI could be viewed as more productive than humans while lacking any of the factors that can influence human productivity. Equating the two is harmful, especially in a competitive environment like the workplace.
AI Avatars of Your Deceased Loved Ones - On Demand - Companies in China are offering to create AI avatars of your deceased relatives so you can stay connected after they’ve passed on. Basic avatars are less expensive, while more interactive ones are more expensive, and all avatars rely on a certain amount of personal data from the deceased person in question.
👉 Why it matters: The creation of AI avatars of deceased individuals brings up a number of concerns, chief among them the ability of a person to properly grieve the loss of a loved one, and the potential to create emotional dependence on technology rather than a human. Additionally, human rights advocates often bring up the “right to be forgotten”. While this is primarily spoken about with regard to a person’s right to wipe their image, likeness and presence from digital spaces, it could also violate a person’s right if they are being digitally recreated when it was not their wish.
*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.
OpenAI Leads in Small LLMs with GPT-4o mini
On July 18th, OpenAI launched their latest model. GPT-4o mini is a small language model (SLM) that supports text and vision in the API, and is 60$ less expensive than GPT3.5 Turbo. In the future it will support text, image, video and audio.
👉 Why it matters: This matters because leveraging a LLM for your work or personal projects can be expensive and, in some cases, prohibitive for those trying to get a product off the ground with limited capital. The small-but-mighty model is one way of increasing access to model use, while also stimulating AI innovation and creativity.
AI Policy Beat
A look at what’s happening in the world of AI policy.
Britain Changes Prime Minister, AI Approach
Britain’s new Prime Minister, Keir Starmer, seems to be opting for a cautious approach to AI regulation. Former Primer Minister Rishi Sunak hosted the AI Safety Summit in the UK and welcoming political representatives from all over the world, but it seems the new Prime Minister wants to take a cautious approach to implementing regulation, rather than adopt regulation similar to that of the EU AI Act.
👉 Why it matters: AI continues to proliferate and become more ubiquitous in people’s day-to-day lives. Without an approach to regulation the UK leaves itself open to the perils of AI, including the violation of people’s privacy rights, implementation of biased AI systems, and job loss due to AI adoption.
Meta Won’t Roll Out Meta AI Assistant in the EU
Meta is the latest in the tech company pack to release its multi-modal LLM, Meta AI Assistant, but it won’t be rolling out in the EU any time soon due to updates to its privacy policy which require all users to approve the use of all user data except chat conversations.
👉 Why it matters: Data is critical to powering LLMs, but this situation is unusual. Meta finds itself in a standoff with the EU over the requirements of their GDPR law meant to protect the privacy of users of technology. Meta claims that without user data the product would be “second rate” for users. Several competitors in this market have managed to roll out AI assistants without violating the terms of GDPR. This begs the question - would it be second rate to the other models in market, or second rate to what Meta wants to build to try and get ahead in the AI game?
“Ethical” AI Music Generator Struggles to Produce a Great Song
Jen, an new AI music generator, is considered to have been “ethically trained” as it was only trained on licensed material. This is in contrast to other AI music generators that have been trained on copyrighted materials and are currently in active lawsuits. Unfortunately it’s not getting great reviews from musicians, and the AI struggles to create complete songs with lyrics.
👉 Why it matters: This is another AI company that is making claims about the ethics of their product. While it might be more ethical to generate music with licensed music as the training data, there are still concerns about the use of AI in music and the way this innovation may negatively impact the livelihood of musicians. Ethics aside, lawsuits in the US will drive clarity on materials that may be used to train these models. Poll: Given the article above, can any AI music generators be ethical?
Spotlight on Research
MUSCLE: A Model Update Strategy for Compatible LLM Evolution
In this paper, Jessica Echterhoff, et al., look into how an adapter for finetuned LLMs reduces changes in the behavior when the base model is changed. This research works to solve the challenge of working to address a specific failure with one LLM only to find that the fix doesn’t work when the LLM changes.
Containerizing CS50: Standardizing Students’ Programming Environments
In this paper, David Malan walks through the process of figuring out the optimal way to standardize the technical setup of students in his CS50 in a way that sets them up for success with coding.Content in Crisis: The Rapid Decline of the AI Data Commons
In this paper, Shayne Longpre, et al, audit 14,000 web domains to understand how consent preferences are changing over time, and the long term ramifications of these changes, including “rapidly biasing the diversity, freshness, and scaling laws for general-purpose AI systems.”
Deeper Insights: OpenAI’s Employees Report OpenAI to the SEC for Violating Whistleblower Rights
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.