OpenAI launches GPT-4.5 💡, Amazon’s Alexa+ 🎙, Figure home robots 🦾

Unlock the full potential of your workday with cutting-edge AI strategies and actionable insights, empowering you to achieve unparalleled excellence in the future of work. Download the free guide today!

💬 OpenAI Developments

OpenAI announces GPT-4.5, warns it’s not a frontier AI model

OpenAI has launched GPT-4.5, its newest AI model, available first for ChatGPT Pro users. While it improves on GPT-4 with better writing, pattern recognition, and reduced hallucinations, OpenAI clarifies that GPT-4.5 is not a frontier model and may underperform compared to o1 or o3-mini. Trained with synthetic data and new supervision techniques, GPT-4.5 aims to feel more intuitive and emotionally nuanced in conversations. It will roll out to more users in the coming weeks and is also available on Azure AI Foundry. OpenAI is expected to launch GPT-5 by late May, integrating its o3 reasoning model for a more advanced AI system.

OpenAI unveils GPT-4.5 ‘Orion,’ its largest AI model yet

OpenAI has launched GPT-4.5 "Orion," its largest AI model yet, available first for ChatGPT Pro users and paid API developers. While it improves on GPT-4 with better factual accuracy, reduced hallucinations, and improved creativity, it falls short of leading AI reasoning models like DeepSeek’s R1 and Claude 3.7 Sonnet on complex benchmarks. Despite being expensive to run, OpenAI sees GPT-4.5 as a stepping stone toward GPT-5, which will integrate its "o" reasoning series for more advanced capabilities.

OpenAI GPT-4.5 System Card

The GPT-4.5 System Card by OpenAI outlines the advancements, capabilities, and safety measures of their latest AI model. GPT-4.5 improves on GPT-4o with enhanced general-purpose reasoning, broader knowledge, better alignment with user intent, and reduced hallucinations. It incorporates new alignment techniques and reinforcement learning to enhance steerability, emotional intelligence, and creativity. Extensive safety evaluations indicate no significant increase in risks, with measures taken to prevent harmful content generation, bias, and adversarial attacks. GPT-4.5 also undergoes rigorous cybersecurity, biological, and radiological threat assessments, showing medium to low risk in sensitive areas. OpenAI continues monitoring and refining its models under a structured Preparedness Framework.

OpenAI’s GPT-4.5 is better at convincing other AIs to give it money

OpenAI's GPT-4.5 has shown high persuasiveness, excelling in internal tests where it convinced another AI (GPT-4o) to donate virtual money and reveal secret codewords. The model used a strategy of requesting small donations, making its attempts more successful than previous models. Despite this, OpenAI claims GPT-4.5 does not meet the threshold for high-risk persuasion but acknowledges growing concerns over AI's role in spreading misleading information and social engineering attacks. The company is revising its risk assessment methods to address these potential threats.

🌐 AI in Various Sectors

Healthcare: Innovaccer introduced AI-powered agents to automate scheduling, patient intake, and referrals, reducing administrative burdens for healthcare providers. The company, which recently raised $275M in Series F funding, integrates AI tools with EHRs to improve care coordination and minimize errors. Other startups like VoiceCare AI, Infinitus Systems, and Hyro AI are also leveraging AI and voice agents to streamline healthcare administration.

Robotics: Figure plans to begin alpha testing its humanoid robot, Figure 02, in homes by late 2025, leveraging its Helix Vision-Language-Action (VLA) model to accelerate learning. This shift follows Figure’s break from OpenAI to develop its own AI models. While industrial applications, like its BMW factory pilot, remain a priority, the company aims to explore household tasks and elderly assistance. However, challenges like home variability, pricing, and safety make widespread adoption a long-term goal, with alpha testing marking an early exploratory phase.

Social Media and Interaction: TikTok is shutting down its Creator Marketplace on April 1 and replacing it with TikTok One, a broader platform with AI-powered tools to help brands and creators collaborate on ads. TikTok One will offer trend insights, AI-assisted scriptwriting, video creation, and digital avatars through Symphony Creative Studio. The platform will also integrate AI-driven ad generation and optimization in TikTok Ads Manager. Businesses and creators are being encouraged to migrate their data before the shutdown, as TikTok shifts towards a more AI-centric advertising ecosystem.

🚀 Tech Industry Moves

Microsoft has launched a dedicated Copilot app for macOS, giving Mac users direct access to its AI-powered chatbot for tasks like drafting emails, summarizing documents, and generating images with DALL-E 3. Previously only available via browsers and mobile apps, this release enhances Mac user experience but requires macOS 14.0+ and an M1 chip or later. Additionally, Microsoft announced free, unlimited access to Copilot’s Voice and Think Deeper features, enabling language practice and advanced reasoning with OpenAI’s o1 model.

Meta is reportedly planning to launch a stand-alone AI chatbot app for Meta AI by Q2 2025, aiming to compete with ChatGPT and Google Gemini. The assistant, currently embedded in Facebook, WhatsApp, and Instagram, may also introduce a paid subscription tier with additional features. With 700 million active monthly users, Meta is aggressively expanding its AI presence, including its Llama open models and the upcoming LlamaCon developer conference in April.

Google Sheets is getting a Gemini-powered upgrade to help users analyze data faster and generate AI-driven insights like trends, correlations, and outliers. The update, now available for Workspace business users, also enables advanced visualizations like heatmaps. Users can ask Gemini to predict trends, identify anomalies, or analyze top-performing channels, with the AI generating insights using Python code or formulas. To use the feature, users can click the Gemini icon in Sheets and input queries for automated analysis.

Amazon has introduced Alexa+, a next-generation AI assistant powered by generative AI, offering smarter, more conversational, and action-oriented capabilities. Alexa+ integrates with smart home devices, music services, online shopping, and even appointment scheduling, leveraging a network of AI "experts" for seamless automation. It personalizes experiences by remembering user preferences, managing homes, and providing deep knowledge across various topics. Available via Echo devices, mobile apps, and browsers, Alexa+ is free for Prime members ($19.99/month otherwise) and will begin rolling out in the U.S. in the coming weeks.

💡 Emerging AI Technologies and Companies

Snowflake grows startup accelerator with $200M in new capital

Snowflake is expanding its startup accelerator with $200 million in new funding, focusing on AI-driven, industry-specific products built on its platform. The accelerator offers startups technical support, co-marketing opportunities, and AWS credits, with notable alumni like Coalesce, LandingAI, and Twelve Labs. Snowflake is also launching a 30,000-square-foot AI hub and a $20 million AI upskilling program while strengthening partnerships with Microsoft, OpenAI, and Anthropic. The company's AI investments are paying off, as it recently beat Wall Street expectations with $987 million in Q4 2024 revenue.

Workhelix taps years of research to help enterprises figure out where to apply AI

Workhelix, a startup co-founded by Stanford and MIT researchers, helps enterprises identify where AI can best be applied by breaking down roles into tasks and scoring them for AI suitability. The company, which has clients like Accenture, Wayfair, and Coursera, recently raised a $15M Series A led by AIX Ventures, with backing from Andrew Ng, Reid Hoffman, and Mira Murati. Unlike pure software solutions, Workhelix combines AI-powered analytics with expert guidance, aiming to bridge the gap between AI potential and real business value in enterprise automation.

🎨 AI in Creative Industries

Akool, a startup doing AI-driven avatars, has enhanced its AI-driven 3D human avatars, enabling real-time interactions by connecting them to large language models (LLMs). The company offers talking avatars for scripted messages and streaming avatars for dynamic conversations, making them ideal for customer support, education, and AI agents. Users can customize avatars with voice cloning and integrate responses from OpenAI or other LLMs. Akool's low-latency, full-body motion avatars create realistic, engaging experiences across platforms like websites, kiosks, and mobile apps, with future plans to improve motion realism.

If you want more updates related to AI, subscribe to our Newsletter