• AI Time to Impact
  • Posts
  • . . AI: AI Agents get to vote and the FCC ban's AI robocalls (2.9.24)

. . AI: AI Agents get to vote and the FCC ban's AI robocalls (2.9.24)

Plus your top 3 links of the week

Author

Marshall Kirkpatrick
February 09, 2024

Friends, today we’ve stories about voting: by both humans and machines. From FCC actions against AI voter robocalls to advancements in LLMs through Multi-Agent voting. In an election year, this mix of innovation, regulation, and risks promises to be interesting.

Each day we’re tracking the top stories discussed by the AI community, but what about the part of the AI community that is you, our readers (please bring your friends), what are you reading?

Top 3 Links Readers Clicked This Week

And now today’s news…

-Marshall Kirkpatrick, Editor

First impacted: AI developers, AI researchers
Time to impact: Short

This paper shows that the sampling-and-voting methodology presents a straightforward yet effective strategy for significantly enhancing the performance of Large Language Models (LLMs) across a variety of tasks. Leveraging multiple agents and aggregating their decisions through voting improves the accuracy of LLMs; interestingly, the method's impact rises and then falls as it is used on progressively more difficult tasks. [More Agents Is All You Need] Explore more of our coverage of: LLM Performance, Sampling-and-Voting Method, Task Difficulty. Share this story by email

First impacted: Fraud Specialists, Voters
Time to impact: Short

The Federal Communications Commission (FCC) has ruled against robocalls that use AI-generated voices, in response to recent misuse during elections. Experts warn that voters should prepare for personalized spam targeting them across various platforms, as the "true dark hats" continue to push boundaries and exploit technology to create disinformation. The FCC can now impose immediate fines on firms that use AI voices, opening the possibility for both recipients and state attorneys general to sue those firms, with potential fines reaching up to $23,000 per call. I'll believe it when I stop seeing it though... [AI-generated voices in robocalls can deceive voters. The FCC just made them illegal] Explore more of our coverage of: FCC Rulings, AI-Generated Robocalls, Consumer Protection Act. Share this story by email

First impacted: AI firms, Developers
Time to impact: Short

Developer platform Vercel has launched nine AI integrations, as well as a new model playground. The playground will allow developers to try dozens of models instantly to generate text, images and audio in dashboards. [Introducing AI Integrations on Vercel – Vercel] Explore more of our coverage of: Vercel AI Integration, AI Model Playground, Frontend Applications. Share this story by email

First impacted: Data analysts, AI researchers
Time to impact: Short

This paper explores the potential of LLMs for automating table-related tasks. The study details evolving strategies in LLM usage, such as instruction-tuning and prompting, while outlining challenges in private deployment, efficient inference, and the development of benchmarks for table manipulation and advanced data analysis. Worth a read if you're looking to leverage LLMs for enhanced data processing and analytics. [Large Language Model for Table Processing: A Survey] Explore more of our coverage of: LLMs Automation, Table-Related Tasks, Advanced Data Analysis. Share this story by email

Yann LeCun Recognized in TIME100 Impact Awards

First impacted: AI developers, AI researchers
Time to impact: Short

Yann LeCun, Meta's Chief AI Scientist, has been recognized by the TIME100 Impact Awards for his role in open-sourcing Meta's AI models. LeCun, who spearheaded the release of models such as the Llama-2, says that open-source AI is crucial for cultural diversity and democracy. [The TIME100 Impact Awards] Explore more of our coverage of: Meta AI, Open-Source Models, AI Diversity. Share this story by email

That’s it! More AI news on Monday!