Welcome back to Daily Zaps, your regularly-scheduled dose of AI news ⚡️ 

Here’s what we got for ya today:

• 🎥 Early leaked access to OpenAI Sora

• 🤖 DOE wants to use AI to reduce electric grid projects timelines

• 🧞‍♂️ California moves to regulate AI in employment decisions

• 📈 Has AI scaling hit a limit?

Let’s get right into it!

Sign Up | Advertise | Tools

STARTUPS

OpenAI shuts down Sora alpha program due to leak

OpenAI's text-to-video AI model, Sora, has faced controversy after some beta testers leaked early access to protest what they call "unpaid R&D and PR." These artists, invited to test and provide feedback, claim OpenAI is exploiting their labor to enhance Sora while controlling its outputs through strict content approval.

The group argues that OpenAI, valued at $150 billion, should support artists more meaningfully rather than using them to boost the credibility of corporate products. OpenAI denied the allegations, stating participation in Sora’s alpha program is voluntary and intended to balance creativity with safety. While Sora’s public release was anticipated by the end of the year, OpenAI has delayed it to address concerns like safety, compute scaling, and potential misuse.

GOVERNMENT

DOE wants to use AI to reduce electric grid project timelines

The US Department of Energy (DOE) is offering $30 million through its AI4IX program to speed up connecting renewable energy projects like solar and wind farms to the power grid. Right now, over 2,600 gigawatts of clean energy projects are stuck in “interconnection queues,” waiting up to seven years for studies to determine necessary grid upgrades and costs. These delays stem from an outdated system designed for fossil fuels.

The DOE aims to use AI to identify and fix application issues faster, cutting down delays and helping more renewable energy projects get online. With electricity demand rising due to technologies like AI, this initiative is critical for modernizing the grid and meeting future energy needs. Funding proposals are open until January 2025, with winners announced in winter 2025.

CAREERS

California moves to regulate AI in employment decisions

The California Privacy Protection Agency (CPPA) has voted to advance new regulations on automated decision-making technology (ADMT), which uses personal data and computation to assist or replace human decisions. If enacted, these rules would impose strict requirements on employers using AI tools for key employment decisions like hiring, promotions, and terminations.

Key provisions include notifying employees, contractors, and applicants about ADMT use and their rights to opt out; conducting bias reviews to ensure non-discrimination; providing human review options for certain decisions; and performing annual cybersecurity audits and risk assessments to safeguard personal data.

OPINION

Has AI scaling hit a limit?

For years, AI progress has been driven by the belief that bigger models, more data, and more computing power automatically make AI smarter. This scaling approach brought us breakthroughs like ChatGPT and massive investments in AI development. However, recent results suggest this strategy is reaching its limits. Models like OpenAI’s Orion and Google’s Gemini are showing diminishing improvements despite consuming enormous resources.

Challenges such as a shortage of high-quality training data, massive energy demands, and the limitations of current architectures like transformers are slowing progress. Experts argue that the future of AI requires new approaches, such as focusing on smarter reasoning during use or building "world models" that understand causality and the physical world.