Your weekly briefing on all things AI

Nevada is set to tackle their backlog of unemployment requests with ai that takes 5 minutes to determine what took humans an average of 3 hours. Artificial Intelligence is rapidly reshaping decision-making across industries, but the question remains: how do we keep it from going off the rails? Especially in irreversible situations. How humans and AI interface will be critical to winning tools in this next wave of development.

Let’s dive in:

> Google AI Is Turning Hours of Government Work Into Minutes

> AI Tool MILTON Detects Early Signs of Over 1,000 Diseases

> AI Can Nudge Users To Get The Most Value From Your Offerings

> Prompt Tip: How to Guide ChatGPT Responses

Unemployment Claims Now in the Hands of Google's AI – Fair or Frightening?

Nevada is about to launch a new AI system from Google’s Vertex AI Studio that could soon help decide whether you get unemployment benefits by reviewing hearing transcripts and making recommendations to human referees. The aim is to cut decision times from hours to minutes and clear the backlog of claims that built up during the pandemic. Although this technology offers significant efficiency improvements, concerns about its accuracy and potential biases remain. To address these, human referees will carefully review all AI recommendations, with strict monitoring and privacy protections in place. This initiative reflects a broader trend of AI taking on significant roles in decision-making. Hebrew U professor, Yuval Noah Harari, has warned that in the next decade, millions of AI bureaucrats may control decisions in jobs, finance, and government, potentially shifting power from humans to these 'alien intelligences.' As AI continues to evolve, balancing innovation with transparency and ethical oversight will be crucial to ensuring it benefits the public effectively and fairly.

More Stories

• OpenAI's New O1 Model Delivers Human-Like Reasoning. (Link)

• Google's Ambitious 7-Year Journey to Integrate AI with Robotics. (Link)

• NotebookLM by Google Now Explains Complex Topics Aloud. (Link)

• OpenAI in Discussions to Secure Funding at a $150B Valuation. (Link)

• Google Cofounder, Sergey Brin, Reveals Daily AI Work at Google. (Link)

• Y Combinator's Latest Batch Sees Three-Quarters Focused on AI Startups. (Link)

• Amazon’s Audible Narrators Can Now Clone Themselves Using AI. (Link)

• LLM Startups Struggle to Compete with Big Tech's Spending Surge. (Link)

• Cybever Launches AI-Driven 3D World Creation Platform. (Link)

• ServiceNow Unveils Customizable AI Agents for Enterprise Workflows. (Link)

• Figure's New Humanoid Robot Takes the Spotlight. (Link)

• Intel's Foundry to Develop Custom Chip for Amazon, Shares Surge. (Link)

|| Healthcare

AstraZeneca’s MILTON Detects Early Signs of Over 1,000 Diseases

AstraZeneca's AI tool, MILTON, has shown remarkable ability to identify early signs of more than 1,000 diseases from routine patient data, long before symptoms appear. By analyzing results from standard tests and extensive biomarkers, MILTON can predict disease risk with impressive accuracy. Utilizing data from 500,000 participants in the UK Biobank, MILTON demonstrated "exceptional" performance in predicting 121 diseases and "highly predictive" results for an additional 1,091. This advancement offers promising prospects for early intervention and targeted treatments. However, experts emphasize the need for strict ethical guidelines to prevent misuse by insurers or employers, ensuring the tool's benefits are applied responsibly. (Link)

|| Real Estate

AI Innovations Poised to Revitalize U.S. Housing Market

The U.S. housing market has been mired in stagnation, with homeowners hesitant to sell due to favorable mortgage rates and buyers facing high costs. Drew Uher, founder and CEO of HomeLight, suggests that artificial intelligence (AI) could be the key to breaking this gridlock. HomeLight’s AI-powered Buy Before You Sell program aims to address the market's rigidity by offering loans that allow homeowners to purchase a new home before selling their current one. As a result, this innovative solution has the potential to reinvigorate the market and offer fresh opportunities for both buyers and sellers. (Link)

|| Finance

FMA Releases AI Research for New Zealand Financial Sector

The Financial Markets Authority (FMA) has released a report on Artificial Intelligence (AI) in New Zealand’s financial services. The study, part of the FMA’s Occasional Paper series, examines current AI use, future applications, and regulatory concerns. Covering insights from 30 financial entities, including deposit takers and insurers, the report highlights issues such as algorithmic bias, data privacy, and cybersecurity. Despite a 40% response rate, the findings offer valuable insights and will inform ongoing discussions about AI governance and compliance. A roundtable on 1 October 2024 will further explore these topics with industry stakeholders. For more information, please contact our experts. (Link)

• Smartcat Lands $43M for its AI-Driven Translation Platform. (Link)

• Huma Finance Attracts $38M to Expand into Solana and Stellar’s Soroban. (Link)

• 11x.ai Raises $24M Led by Benchmark to Build AI Digital Employees (Link)

• Fal.ai Gets $23M From a16z for Hosting Media-Generating AI Models (Link)

• Fathom Note Taker Raises $17M -$2M From Crowd Funding (Link)

• Landbase Raises $12.5M to Revolutionize Go-To-Market Strategies with AI. (Link)

• PhoreMost Adds $12M to Series B for Drug Development. (Link)

• Operant AI Secures $10M to Enhance Runtime Application Protection. (Link)

• CardiaTec Raises $6.5M to Transform Heart Disease Treatment with AI. (Link)

• Realm Security Kicks Off with $5M to Combat Cybersecurity Data Overload. (Link)

• Syantra Garners $4.9M in Series A-1 Funding For AI Breast Cancer Testing. (Link)

• Purple Transform Nabs £4.5M for AI-Powered Rail Data Analytics. (Link)

• Optalitix Attracts $3M to Advance Insurance Software Innovation. (Link)

CommandAI’s Nudges: User Onboarding That Customers Can Enjoy

One major reason for early customer churn is the learning curve required to get value from your platform/tool. Nudges leverages AI to create and optimize your platform’s onboarding and useability. It tracks the kinds of interactions your users are having with your platform, and how they interact with the nudges (carefully governed popups) in order to help you improve your customer’s experience.

More tools to try:

• Fellow App: centralizes recordings, transcriptions, and summaries for a unified meeting experience.

• Phonescreen AI: recruitment tool for parsing job posts, answering candidate queries, generating screening questions, and scoring responses.

• Command AI: provides seamless, non-intrusive AI-powered user assistance for enhancing product, marketing, and customer support teams.

• Genkin: simplifies tracking your spending with interactive charts and AI-powered chat for instant financial insights.

• Hoop: uses AI to organize and prioritize tasks across all your teams, integrating effortlessly from meetings to Slack.

• Scripe.io: creates viral LinkedIn posts 10x faster, perfect for personal branding, teams, and businesses.

• CX Genie: automates and personalizes customer support with chatbots, ticket management, and help desk features.

• Recast Studio: helps marketing teams effortlessly edit and repurpose video and audio content without needing editing skills.

Using Examples to Guide ChatGPT Responses

Describing a particular style or tone can be difficult, but showing is often easier than telling. By providing clear examples, you allow ChatGPT to learn from the structure and style you’ve demonstrated. This makes it easier to produce content—whether it's for marketing emails, customer support responses, or team communications— that aligns with your brand voice without needing to explain every little detail. This technique is known as "few-shot" prompting.

Example Use Case: Responding to Customer Complaints

Let’s say you own a service-based business, and you often get complaints. You want ChatGPT to handle these complaints in a tone that’s empathetic but also solution-oriented.

Prompt to copy:

[whenever I ask you to respond to customer complaints, please reply in the style of the following examples:

1. Example 1:
   Customer Complaint: "I’ve been waiting two weeks for my delivery. This is ridiculous!"
   Response Style: "I’m so sorry for the delay, [Customer’s Name]. I completely understand how frustrating this is. Let me check on your order right now and make sure we get this resolved as quickly as possible. I’ll keep you updated every step of the way."

2. Example 2:
   Customer Complaint: "The item I received is damaged, and I’m really upset!"
   Response Style: "I sincerely apologize for the condition in which your item arrived, [Customer’s Name]. I know how disappointing this must be. I’ll arrange a replacement immediately, and we’ll get it to you as fast as possible. Please let me know if there’s anything else I can assist you with in the meantime."

New Customer Complaint: <insert complaint here>]

Why This Works

Despite taking a bit more work upfront, by providing concrete examples, you're giving ChatGPT a clearer understanding of what you're looking for. The model learns from these examples and adapts its responses to match the desired style, saving you time on editing and fine-tuning responses. This method is especially helpful when you need the AI to replicate a tone that's difficult to describe in abstract terms.