🔥 Hottest News: GPT-4o Launched: OpenAI has released GPT-4o, which takes AI interaction to new heights with multimodal capabilities for text, audio, and images, simultaneously! Faster, more human-like interactions, better reasoning, and improvements in cost and speed.
ChatGPT Enhancements: Boosting trust in AI-generated media with new watermarking technologies and AI-based detection classifiers to verify the authenticity of digital content.
Corporate AI Developments: Microsoft has introduced MAI1, focusing on ethical AI deployment. Meanwhile, rumours are circulating about Apple potentially integrating ChatGPT into iOS 18, promising to elevate its AI offerings.
AI Innovations Spotlight: OpenAI's Media Manager allows content creators to manage their intellectual property with AI, while Atlassian's Rovo tool integrates advanced AI to improve enterprise decision-making and knowledge management.
AI in Business Applications: AI is making significant inroads in various sectors—helping hotels manage customer complaints, aiding doctors with patient documentation, and being tested by California’s government to enhance public services.
Responsible AI Practices: OpenAI released guidelines on desired behaviours for AI models, emphasising compliance, harm reduction, fairness, and privacy.
🔍 This Week's Spotlight: Dive into the insights from The Generative AI Dossier by Deloitte AI Institute, where we look into the historical and impact of generative AI across across multiple industries.
Ethan Mollick’s Insights: Practical advice from his latest book "Co-Intelligence"
👇 Interested in these updates? Read the full newsletter for a detailed exploration of these topics and more insights.
🔥Hottest News: Introducing GPT-4o: A Leap Towards Natural AI Interaction
Source: ChatGPT
OpenAI has launched GPT-4o, its latest AI mode. This new model can process and understand inputs from all three modalities simultaneously, providing a comprehensive and contextually rich interactive experience. GPT-4o enhances real-time reasoning and accuracy, making it suitable for complex applications such as virtual assistants, educational tools, and accessibility devices. Announcement
Top 10 Key Features of GPT-4o
Multimodal Capabilities: Accepts and generates text, audio, and image inputs and outputs.
Human-like Response Time: Responds to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, similar to human conversation response times.
Performance: Matches GPT-4 Turbo in text (English) and code, with significant improvements in text for non-English languages. Superior vision and audio understanding.
Cost: 50% cheaper in the API than GPT-4 Turbo.
Speed: Twice as fast as GPT-4 Turbo, providing quicker responses for real-time applications.
Rate Limits: Higher rate limits, allowing up to 10 million tokens per minute, five times higher than GPT-4 Turbo.
Real-time Interaction: Supports simultaneous processing of audio, visual, and text inputs, enhancing interactive experiences.
Improved Reasoning and Accuracy: Significant advancements in reasoning and factual accuracy, making it more reliable for complex tasks and decision-making.
Availability: Available in ChatGPT Free, Plus, and Team plans, with Enterprise access coming soon. Free tier users have limited access with a cap on messages.
Advanced Tools and Vision Capabilities: Free users have limited access to advanced tools like data analysis, file uploads, browsing, and enhanced vision functionalities that improve image understanding.
Voice Mode in GPT-4o: Key Improvements
Previous Voice Mode:
Structure: Used separate models for audio-to-text, text processing (GPT-3.5 or GPT-4), and text-to-audio.
Limitations: Lacked the ability to understand tone, multiple speakers, or background noises and couldn't output laughter, singing, or express emotions naturally.
Latency: Average response times were 2.8 seconds (GPT-3.5) and 5.4 seconds (GPT-4).
New Voice Mode (GPT-4o):
Unified Model: Integrates text, vision, and audio in one neural network.
Capabilities: Better understanding of tone, multiple speakers, and background noises; can generate laughter, singing, and emotional expressions.
Performance: Faster response times and more accurate, contextually rich interactions.
Significance: Represents a significant step towards more natural and intuitive human-computer interactions
When Will we Get Access to GPT-4o:
Immediate Rollout: GPT-4o's text and image capabilities are starting to roll out today in ChatGPT. Available to free tier users, with extended access to Plus users, who get up to 5x higher message limits.
Upcoming Features: A new version of Voice Mode with GPT-4o in alpha will be available within ChatGPT Plus in the coming weeks.
Developer Access: Developers can access GPT-4o via the API as a text and vision model. GPT-4o is 2x faster, half the price, and has 5x higher rate limits compared to GPT-4 Turbo. Support for GPT-4o’s new audio and video capabilities will be launched to a small group of trusted partners in the API in the coming weeks.
Red Team Access: Extended red team access starts today to further test and refine the model.
Around the Web
ChatGPT Enhancements:
OpenAI introduces new technologies to verify the authenticity of digital content, including watermarking and AI-based detection classifiers. As part of the Coalition for Content Provenance and Authenticity, OpenAI is setting industry standards to ensure transparency in content creation. These efforts include a societal resilience fund and tools that signal the origin of digital content, bolstering trust and security in AI-generated media. Announcement.
Corporate AI Developments:
Microsoft is launching MAI1, a new AI language model developed in-house, featuring 500 billion parameters. Led by Mustafa Suliman, the project builds on technology from the acquired startup Inflection. Although smaller than OpenAI's GPT-4, MAI1 represents a significant step in Microsoft's AI strategy, emphasizing enhanced in-house capabilities and focusing on safety and ethical AI development. The model is expected to bolster Microsoft's competitiveness in natural language processing and other AI-driven applications. Video.
Apple is finalising an agreement with OpenAI to embed ChatGPT into the upcoming iOS 18, aiming to boost its suite of AI features. Simultaneously, Apple's negotiations with Google to incorporate the Gemini chatbot continue without a finalised deal. Article.
AI Innovations Spotlight
OpenAI announces the development of Media Manager, a new tool designed to empower creators and content owners. This innovation allows individuals to specify the inclusion or exclusion of their works in AI research and training. Future updates will expand the available choices and features, further aligning AI development with creator preferences and ethical standards. Announcement.
Atlassian introduces Rovo, a powerful tool leveraging generative AI to enhance enterprise decision-making and information management. Integrated within Atlassian’s ecosystem, Rovo streamlines search and knowledge management across both internal tools like Jira and Confluence, and third-party applications such as Google Drive and Microsoft SharePoint. With advanced features like knowledge cards and AI chat functions, Rovo not only simplifies data discovery but also facilitates deeper insights and automated task completion, offering a significant productivity boost in the IT industry. Article.
AI in Business Applications:
In the hospitality industry, crafting responses to customer complaints can be challenging and time-consuming. With the rise of generative AI, hotels and travel companies are turning to ChatGPT for help. While AI-generated responses offer efficiency and neutrality, some industry professionals remain wary of losing the personal touch. Article.
California's government, led by Governor Gavin Newsom, is set to integrate generative AI technology to improve public services. This initiative involves partnering with five companies to test AI tools in reducing customer service wait times and enhancing traffic safety among other uses. The state is conducting a six-month internal trial, assessing the technology's effectiveness before potential broader deployment. Concerns about job loss, misinformation, and privacy are also being addressed with planned safeguards and continuous testing. Article.
AI technology is transforming medical practices by reducing the documentation burden on doctors and enhancing patient interactions. Dr. Rebecca Mishuris of Mass General Brigham, along with NYU Langone Health, is pioneering the use of AI to transcribe and summarise patient visits, significantly cutting down post-visit documentation time. Studies show doctors spend about two hours on desk activities for every hour of patient interaction, a ratio that AI aims to improve. This shift allows doctors to focus more on patients during visits, potentially decreasing burnout. Article.
A recent study published in npj Digital Medicine evaluates the efficiency of ChatGPT in extracting structured data from unstructured clinical notes. Utilising the ChatGPT 3.50-turbo model, the research demonstrated an impressive ability to accurately classify pathological data and staging information from extensive datasets of lung tumour and pathology reports. With accuracy rates reaching up to 98.6%, ChatGPT outperformed traditional keyword search algorithms and deep learning-based approaches. Article.
ChatGPT's integration into Colorado's K-12 education systemis reshaping the educational landscape. Teachers, like Amber Wilson at Denver’s Thomas Jefferson High School, are adapting their methods, shifting towards more in-class activities and oral defences of student thinking in response to the generative AI's capabilities. While some districts, like Denver Public Schools, restrict ChatGPT's use due to data privacy concerns, the technology's influence is undeniable. Educators are rethinking assignments to better utilise AI's potential while addressing the challenges of academic integrity and ensuring equitable access to technology. Article.
Responsible AI Practices:
On May 8, 2024, OpenAI released a comprehensive draft detailing desired and undesired behaviours for AI models like ChatGPT. This Model Specification emphasises legal compliance, harm reduction, fairness, and privacy, aiming to refine AI development in line with ethical standards. This guide helps users understand AI responses and supports alignment with OpenAI's values. Announcement.
🔍This Week's Spotlight:The Generative AI Dossier (By Deloitte AI Institute)
Source: The Generative AI Dossier Deloitte
In this edition, we have reviewed and analysed the 146-page paper from Deloitte AI Institute, titled "The Generative AI Dossier: A selection of high-impact use cases across six major industries." This comprehensive document provides an extensive look at the transformative effects of Generative AI across various sectors. In our analysis, we explore how Generative AI builds upon and significantly enhances existing technologies. We focus on specific roles that Generative AI fulfils across different industries, highlighting its impact on efficiency, personalisation, and business intelligence:
1. Consumer: Marketing Content Assistant:
Historical Context: Automated content generation tools have traditionally been used for creating simple, formulaic content like social media posts and basic advertisements.
Generative AI Impact: Advances the capabilities to produce more complex and creative content, such as video scripts, personalised articles, and engaging multimedia content that dynamically adapts to user feedback and trends.
Historical Context: Early automated customer support systems relied heavily on predefined scripts and were limited to handling very routine queries.
Generative AI Impact: Enhances these systems by enabling more sophisticated understanding and generation of responses, offering personalised and contextually relevant interactions that closely mimic human customer service representatives.
3. Consumer: Virtual Try-On
Historical Context: Virtual try-on technologies have been available in rudimentary forms, using simple overlay techniques without much personalisation.
Generative AI Impact: Improves realism and accuracy, allowing for highly personalised fitting experiences that consider individual body dimensions and preferences, significantly enhancing the online shopping experience.
4. Energy, Resources, and Industrial (ER&I): Asset Maintenance Planner
Historical Context: Predictive maintenance has been employed using sensor data and simple predictive algorithms to forecast equipment failure.
Generative AI Impact: Incorporates more complex data sources, including operational data, environmental conditions, and more intricate machine learning models, to predict failures more accurately and optimize maintenance schedules.
5. Energy, Resources, and Industrial (ER&I): Materials Designer
Historical Context: The use of computational chemistry and materials science to predict material properties has been an ongoing development but often required extensive trial and error.
Generative AI Impact: Automates and accelerates the discovery process, enabling the simulation and testing of thousands of materials combinations to innovate faster and more efficiently.
Historical Context: Traditional fraud detection systems were rule-based with static parameters, which fraudsters could learn and circumvent.
Generative AI Impact: Employs advanced machine learning models that continuously learn from transaction data, adapt to new fraudulent tactics, and detect complex fraud patterns with higher accuracy.
Historical Context: Risk management used statistical models and historical data for analysis, which often lagged behind real-time events.
Generative AI Impact: Utilises dynamic modelling techniques that incorporate real-time data streaming, predictive analytics, and scenario simulation to offer more accurate and timely risk assessments.
8. Government & Public Services (GPS): Public Engagement Facilitator
Historical Context: Automated systems for handling public inquiries typically provided generic responses and were limited to simple queries.
Generative AI Impact: Improves these systems significantly, allowing for nuanced conversations, handling complex queries, and providing tailored information, thereby enhancing citizen engagement.
9. Government & Public Services (GPS): Process Automation Specialist
Historical Context: Process automation in government services often involved basic tasks like form processing and data entry.
Generative AI Impact: Extends automation to more complex governmental processes, including decision-making aids for policy development and compliance monitoring, making operations more efficient and reducing human errors.
These analyses highlight how Generative AI not only builds on but significantly enhances the capabilities of existing technologies, introducing new levels of efficiency, personalisation, and intelligence into various business roles. Read full document here.
Insights from Ethan Mollick's Principles of Co-Intelligence
As artificial intelligence (AI) continues to advance, Ethan Mollick's book offers crucial insights into enhancing our interactions with AI technologies. Here, I'll explore the four foundational principles he outlines, reflecting on my personal takeaways and how they can revolutionise our approach to AI in both personal and professional settings:
Principle 1: Always Invite AI to the Table
Mollick advocates for a proactive inclusion of AI in all aspects of decision-making, except where restricted by legal or ethical concerns.
This principle is not just about leveraging AI for assistance but about understanding its capabilities and limitations through direct engagement.
For individuals, the cost of experimenting with AI is minimal compared to the potential innovations it can drive.
As AI evolves, becoming a familiar companion in our thought processes will be crucial for those aiming to integrate technology seamlessly into their daily lives and work.
Principle 2: Be the Human in the Loop
The concept of keeping a 'human in the loop' underscores the importance of human oversight in AI-driven systems.
Mollick emphasises the critical role of human judgment, especially given that AI systems, in their eagerness to satisfy, might not always prioritise accuracy.
Humans are needed to apply ethical considerations, critical thinking, and context that AI, by its nature, might miss.
This principle reminds us that while AI can process information at remarkable speeds, it often requires human direction to navigate complex moral landscapes and avoid potential 'hallucinations' or errors.
Principle 3: Treat AI like a Person (But Tell It What Kind of Person It Is)
This intriguing principle suggests treating AI as an infinitely capable, yet potentially misleading intern.
AI systems perform better when given specific roles or personas. This approach not only improves the quality of the AI's output but also helps in managing our expectations of what AI can and cannot do.
By defining the AI's persona, users can steer the technology toward more useful and ethical outputs, effectively tailoring AI interactions to suit specific needs and scenarios.
Principle 4: Assume This Is the Worst AI You Will Ever Use
Mollick concludes with a principle that encourages a forward-looking perspective on AI's development.
By treating current AI as the worst we will encounter, we remain open to advancements and maintain a critical eye on its present capabilities.
This mindset ensures that we are constantly evaluating AI technology critically, pushing for improvements and better integrations in future iterations.
Ethan Mollick's principles offer a roadmap for harnessing AI's potential responsibly and effectively. From inviting AI into everyday decision-making to treating it as a dynamic tool that requires careful handling and specific directionAs we look to integrate AI more deeply into our lives, understanding these principles can empower us to use technology not just as a tool, but as a partner in our continuous quest for growth and innovation.
🗣️ We'd Love to Hear from You!
Did you find this edition insightful? Like, share your thoughts, and ask us anything.
Stay curious, and until next time, enjoy exploring the vast possibilities AI offers.
As we explore the captivating world of AI technologies, it's important to note that the content of this article is a product of human effort, enhanced by AI tools and additional resources. We have made sure to properly attribute and link all referenced articles to their original sources.
Links with this icon
were created by LinkedIn and links without it were added by the author.
AI Experts - Join our Network of AI Speakers, Consultants and AI Solution Providers. Message me for info.
7moExciting updates in the AI landscape. Can't wait to read more.