| |
| Welcome to today's Inside AI! Here are today's top stories: - India and the Philippines face risks as AI chatbots replace call center workers.
- An analysis of LLMs shows a wide range of conversations, including explicit and unsafe topics.
- LinkedIn is rolling out an AI-powered learning coach and other new AI features.
If you find this newsletter useful, please consider forwarding it to your friends and colleagues. Thank you! Beth
p/beth-duckett | |
| 1 | | According to the Washington Post, companies are increasingly turning to AI chatbots for customer service, raising concerns about job displacement. Economists believe this shift could impact global economies, especially in countries like India and the Philippines, where millions are employed in call center jobs. More: - In India, companies are already experimenting by substituting customer service and call center agents with large language models.
- Earlier this year, Philippine Senator Imee Marcos called for an inquiry into tech-driven workforce displacement, particularly in the outsourcing sector.
- Virginia L. Doellgast, a Cornell University labor relations professor, argues that technology will complicate call center jobs. AI chatbots will handle simpler customer service problems, leaving more complex issues for humans.
Zoom out: - The Post report comes after a new study from MIT and Stanford researchers found that customer support agents at a call center saw their productivity rise by 14% when they had access to an AI conversational assistant.
- Over the course of a year, scholars found that agents with the tool were able to resolve 13.8% more issues per hour.
- The least skilled and experienced benefitted the most, finishing their work 35% faster with the tool.
|     | |
| 2 | | An analysis of users' real-world conversations with large language models revealed that while most focused on everyday topics like programming and travel, some also ventured into "unsafe" subjects such as explicit content and toxic behavior. All in all, nearly 10% of LLM conversations involved explicit storytelling and fantasy. 
Background: Researchers from the University of California, Berkeley, UC San Diego, Carnegie Mellon University, Stanford, and Mohamed bin Zayed University of Artificial Intelligence gathered a database of 1 million conversations between people and 25 LLMs. The dataset, compiled from April to August, is the first large-scale real-world LLM conversation dataset. Findings: Out of a random sample of 100,000 conversations, the researchers found that the top 50% of conversations focused on ordinary topics like programming and writing help. However, the other half contained some requests they categorized as "unsafe," including explicit storytelling, sexual fantasies, and discussions of toxic behavior. The authors believe that within the full 1 million conversations, there could be even more harmful content. Why it matters: The researchers found that open-source language models without safeguards tend to generate flagged content more often than proprietary models. However, they still observed cases of "jailbreaking" with models like GPT-4 and Anthropic's Claude. They mention that GPT-4's security measures are broken into about a third of the time. Q: What do you think developers and organizations should do to prevent AI language models from spreading harmful content? Join the discussion here. | | | |
| |
|
| 3 | | Microsoft's LinkedIn partnered with OpenAI to introduce new AI features, including a learning coach and talent-sourcing platform. LinkedIn has already been gradually incorporating AI into its services, including AI-based writing suggestions and AI-created job descriptions. More: - The new offerings include Recruiter 2024, an AI-assisted recruiting experience using generative AI.
- Recruiters type in natural language prompts, and the AI tool will create a project and offer a list of matching candidates.
- LinkedIn is also offering an AI-powered learning coach for LinkedIn Learning, offering tailored advice based on job title and more, as well as an AI tool for marketing called Accelerate.
- AI is also being applied to LinkedIn's Inside Sales for better B2B audience engagement.
| | | |
| 4 | | The National Security Agency (NSA) said it's opening an AI Security Center to defend U.S. AI systems from cyberattacks and other security threats. The center will centralize the NSA's AI security efforts and collaborate with the Pentagon, tech industry, academia, and international partners. More: - Outgoing NSA director General Paul Nakasone announced the move, saying the AI Security Center would be incorporated into the NSA's Cybersecurity Collaboration Center.
- The latter works with private industry and global partners to strengthen the U.S. defense-industrial base against threats, primarily from China and Russia.
- Nakasone said the AI Security Center would be the NSA's central hub for guiding AI security and promoting secure AI development within national security systems and the defense industry.
Zoom out: - The new center comes as the U.S. government has expanded its use of AI systems in defense and intelligence.
- Meanwhile, U.S. adversaries have stepped up their use of AI in influence operations. Last month, Microsoft said it had found that Chinese government-aligned campaigns are increasingly using AI-generated images to target U.S. political candidates and impersonate voters.
| | | |
| A message from our sponsor, Symon.AI. |  | | Fuel Your Startup’s Growth with No-Code AI Data Analytics! Symon.AI brings a no-code platform with AI analytics to your fingertips. We help propel your startup’s growth trajectory. Expand your growth with Symon.AI: - Dive deep into datasets & unlock growth-driven insights.
- Harness the power of machine learning, no data expert is required.
- Collaborate, share, and translate data into actionable strategies.
- Elevate your startup’s growth story, trusted by leaders in North America & Europe.
Eager to witness the transformation? See why companies like HID and KPN Ventures use Symon AI to level up their businesses.
Explore Symon.AI’s intuitive tools & experience growth like never before!
Get Started for Free | |
|
| 5 | | Visa announced plans to invest $100M in companies working on generative AI technologies in the commerce and payments space. Visa Ventures, the company's investment arm, will oversee the investments. More: - David Rolf, Visa Ventures' head, told TechCrunch there will be flexibility in terms of investment volume and check size.
- The fund will likely invest a few million each in mostly early-stage companies, though Visa Ventures will still be able to "make larger investments where there is a strong rationale to do so," Rolf said.
- Visa is seeking to back companies applying GenAI to solve problems in payments, commerce, and fintech.
- This includes areas like B2B payment processes, infrastructure affecting commerce, data organization for GenAI, and user experiences, according to Rolf.
| | | |
| |
| 7 | | Quick Hits: - It's been 150 years since the last great leap in engine technology. Invest in LiquidPiston to be part of the next one.*
- OpenAI's DALL-E 3 image creator is now accessible to all Bing Chat and Bing Image Creator users.
- Zelda Williams, Robin Williams' daughter, called the use of AI to recreate her father's voice "disturbing," noting the ethical concerns surrounding voice replication without the user's consent.
- Senate Intelligence Committee Chairman Mark Warner, D-Va., has called on Congress to prioritize AI regulation in U.S. elections, saying that deepfakes in the 2016 elections could look like "child's play" compared to 2024.
- Google is set to unveil the Pixel 8 and Pixel 8 Pro smartphones, which boast AI features, during its hardware event tomorrow, Oct. 4. The event will livestream starting at 7 a.m. PT on YouTube and the Google Store site.
- Even the best engineers can run into challenges. Straight from the cloud experts at DoiT, here are 5 core strategies for cloud innovation.*
*This is a sponsored post. | | | |
| Upcoming Events | | OCT
12 | | Panel discussion on cybersecurity insights for financial services companies* | | | | | | | OCT
16 | | Meet Our Fund 5, Building and Investing in AI | | | | | | * This is a sponsored event | | | |
Term of the Day Large language models: Large language models (LLMs) are a statistical method that helps predict the following outcome in a given sequence of input. Question of the Week Do you share details about your salary with your coworkers? Join the conversation |
INSIDE AI LEADERBOARD (7 DAYS) |
 | AI and technology writer | | Beth is an editor and writer of Inside's AI and Tech newsletters, with a career spanning journalism since 2007. She has written for publications including The Arizona Republic and USA Today and authored a book on the solar industry in 2016. Reach her at Beth.Duckett@yahoo.com. | | This newsletter was edited by Aaron Crutchfield | |
| |
 | See the power of the Wiz platform. | |
|
 | Symon.AI empowers you with predictive and prescriptive insights using a no-code, data prep and AutoML platform. | |
|
| |
Comments
Post a Comment