Real estate
- PropHero Raises Record Breaking Series a Round: PropHero, a Sydney-based company leveraging AI and big data for property investments, has raised a record-breaking 25 million in a Series A round led by Bailador Technology Investments. Founded in 2021 by Mickael Roger and Pablo Gil Brusola, PropHero aims to become a global real estate investment platform, serving over 2,000 paid customers and offering free tracking services to 8,000 users worldwide. The funds will enhance their AI platform, expand operations, and target a doubling of revenue to 50 million while focusing on transparency and trust.
- Henry AI Raises 4M to Automate Commercial Real Estate Transactions: Henry AI, a pioneering AI analyst for the commercial real estate sector, has secured 4.3 million in seed funding to revolutionize this 20 trillion industry by automating its repetitive manual tasks. By swiftly generating professional-grade marketing materials and detailed financial analyses, Henry AI significantly reduces preparation time, enabling professionals to focus on building relationships and closing deals. Backed by several national brokerages, the startup enhances efficiency and creates new opportunities in the dynamic market.
AI
- A New Generation of AIs. Claude 3.7 and Grok 3: Ethan Mollick discusses the advancements of the new Gen3 AI models, Claude 3.7 and Grok 3, highlighting their significant improvements over previous generations like GPT-4. These Gen3 models are trained with considerably more computing power, enabling them to perform complex tasks like coding, creating interactive visualizations, and conducting PhD-level research. The new Scaling Laws suggest AI models improve both through increased computing power during training and allowing more computational time for problem-solving. Despite their imperfections, these models represent a potential shift from AI’s role in mere task automation to capability augmentation, urging organizations to explore new possibilities in AI integration.
- Introducing GPT-4.5 | OpenAI: OpenAI has introduced GPT-4.5, a research preview of its most advanced chat model. It enhances pre-training and post-training scalability, improving pattern recognition, connection drawing, and creative insights. Early tests show improved natural interaction, a broader knowledge base, and better user intent following, with fewer hallucinations. GPT-4.5, developed on Microsoft Azure AI supercomputers, exemplifies unsupervised learning scale, complimenting reasoning models which enhance complex problem-solving.
- NVIDIA Earth-2 Features First Gen AI to Power Weather Super-Resolution for Continental US: NVIDIA has introduced Earth-2 featuring CorrDiff, a generative AI model, to improve weather forecasts for the continental US. CorrDiff enhances weather predictions from a 25-kilometer resolution to 2 kilometers using AI techniques, significantly increasing forecast precision. This model, integrated into the NVIDIA Earth-2 platform, aids weather agencies, researchers, and companies in managing risks from extreme weather by supporting effective planning and infrastructure development. Trained on detailed weather simulations, CorrDiff offers high-speed, energy-efficient forecasts that assist in regional forecasting and disaster management. Available as an NVIDIA NIM microservice, it enables rapid, scalable weather predictions, highlighting advancements in forecasting technology.
- Securing the Conversational Frontier: Advanced Red Team Testing Techniques for Chatbots: The article by Deep Dhillon discusses advanced techniques for red team testing to enhance the security of chatbots. It outlines key methods such as Ground Truth-driven efficacy testing, which provides insights into security performance by establishing expected outcomes for various scenarios. Input fuzzing tests resilience against unexpected inputs, while adversarial simulations defend against manipulation. API and backend testing ensure robust infrastructure security. The article emphasizes creating scenarios to test vulnerabilities, including contextual ambiguities and impersonation risks, and recommends tools for both white-box and black-box testing to safeguard chatbots against technical and social engineering threats, aligning with best practices such as the OWASP Top Ten.
- What Is Red Teaming for Generative AI?: Red teaming, a concept originating from military exercises during the Cold War, has evolved from probing computer systems for weaknesses to stress-testing generative AI for potential harms. In this context, red teaming involves intentionally provoking AI models to reveal vulnerabilities, such as producing inappropriate content or exposing biases. This practice helps improve AI safety and security by realigning models with new instruction data. IBM and MIT have advanced this technique by using “red team” language models to generate a variety of prompts, surfacing both common and rare vulnerabilities. IBM researchers have demonstrated that even state-of-the-art models like GPT-3.5 Turbo can be compromised, emphasizing the ongoing necessity of red teaming as generative AI develops. IBM’s efforts also include creating datasets like AttaQ and SocialStigmaQA to provoke and analyze detrimental AI behavior, aiming to foster safer AI deployments amid evolving threats and regulatory landscapes.
Technology
- You Are Completely Unprepared for What This Humanoid Servant Robot Looks Like: Norwegian robotics startup 1X has unveiled its latest humanoid robot, Neo Gamma, designed for household tasks like preparing tea and vacuuming. Wrapped in a KnitSuit for dynamic movement, Neo Gamma has a face likened to a blend of pop culture characters for an eerie effect. Despite CEO Bernt Børnich’s vision of robots in every home, commercial availability is uncertain. While 1X and others like Tesla and Figure vie in the humanoid market, hefty engineering and affordability challenges remain.
Others
- Distraction Starts From Within: Ben Putano’s article, “Distraction Starts From Within,” emphasizes that distraction is driven by internal triggers rather than external factors like social media or chores. Putano discusses Nir Eyal’s journey from promoting attention-grabbing products to advocating against distraction through his book Indistractable. Eyal suggests that recognizing and addressing internal triggers, such as anxiety or boredom, is crucial to overcoming distraction. Techniques like journaling and meditation can help manage these triggers, allowing individuals to regain focus.