AI: the bigger picture #3

It's easy to consume the never-ending stream of hourly news. But what's the big picture? Based on my weekend reading, here is an overview of what's really happening in AI

1. The "AI is overhyped" bubble burst. Surprisingly, the rebound has recovered most of the losses for MSCI world, Dow Jones, Nvidia, Apple, Meta etc. already. To me it seems that "the AI bubble" itself was the hype.

• Goldman Sachs: AI Is Overhyped, Wildly Expensive, and Unreliable (source)

• OpenAI Clocks $3.4 Bn in Revenue from ChatGPT Subscriptions (source)

• Forget the bubble, AI is here to stay in the creative sectors (source)

2. AI research and AI tech companies stand strong

AI research keeps finding methods to significantly improve the scope, speed and quality of AI models. This is the basis for continued growth and increased global impact of AI.

• Researchers Develop Episodic Memory-Integrated Large Language Models (LLMs) for Infinite Context Lengths (source)

• "Hypothetical Minds" Model Revolutionizes Multi-Agent Reinforcement Learning with Theory of Mind Module (source)

• Stanford Researchers Introduce SGLang: A Structured Generation Language for Efficient Execution of Complex Language Model Programs (source)

US Tech Companies relentlessly push AI forward to secure and extend their ecosystems ...

• OpenAI: Unveils Cheaper, Small AI Model GPT-4o Mini (source). "Project Strawberry"/Q-star details leak and reveal roadmap towards AGI (artificial general intelligence) (source). OpenAI’s search engine is here : SearchGPT will eventually be folded into ChatGPT (source)

• Meta unveils biggest Llama 3 AI model, touting language and math gains (source)

• Anthropic console is the hidden powerhouse for AI — and it just got a lot better (source)
• Apple iOS 18.1 Beta previews "Apple Intelligence" for the first time (source)

• xAI unveils Grok-2 to challenge the AI hierarchy (source)
• Google Expands Gemini 1.5 Flash Language Model with Improved Response Quality and Speed (source)

... and so are non-US companies:

• ByteDance, Alibaba, SenseTime lead generative AI infrastructure services market in China (source). SAS Survey: China Leads the World in Generative AI (source).

• Tencent’s New AI Engine Boosts Game Dev Efficiency by 40 Times (source)

• SenseTime SenseNova 5.5: China’s first real-time multimodal AI model (source)

• Mistral-Large-Instruct-2407 Released: Multilingual AI with 128K Context, 80+ Coding Languages, 84.0% MMLU, 92% HumanEval, and 93% GSM8K Performance (source)

3. AI applications are getting more powerful, and they are in broad adoption. In the past 4 weeks, these were the top areas:

Software coding: broad progress in quality of generated code and tooling, driving adoption

• Mistral AI Launches Codestral Mamba 7B: A Revolutionary Code LLM Achieving 75% on HumanEval for Python Coding (source)

• Salesforce Research Introduces INDICT: A Groundbreaking Framework Enhancing the Safety and Helpfulness of AI-Generated Code Across Diverse Programming Languages (source)
• Excellent AI Tools for coding and software developers in 2024 (source)

Voice: Voice assistants are both simplifying and personalizing the human-AI interaction

• OpenAI Unveils GPT-4o's Voice Mode and SearchGPT, Revolutionizing AI Interactions (source)

• Google Rolls Out Gemini Live to Compete With ChatGPT's Voice Mode (source)

• Rep. Jennifer Wexton's AI-Generated Voice Restores Her Ability to Communicate After Parkinson's-Like Condition (source)

Marketing: AI adoption is gaining ground

• Mango Unveils AI-Generated Campaign for Limited-Edition Teen Collection, Emphasizing Technology's Role in Fashion Innovation (source)
• How AI Is Revolutionizing the Marketing Landscape (and How to Use It Ethically) (source)
• Top 10 AI Tools for Sales and Marketing: Boosting Productivity and Company Performance (source)

Media: Image, Music and Video generators crank up quality and proliferate

• Image: Midjourney v6.5 could be out ‘by the end of the month’ with improved realism and skin textures (source). Google’s upgraded AI image generator is now available (source)

• Music: Kunlun Tech Revolutionizes Music with AI-Powered Platforms Melodio and Mureka (source). Udio AI Music Creator Version 1.5: Enhanced Audio and New Features (source)
• Video: Runway's Gen-3 Turbo Model Revolutionizes AI Video Generation with Rapid Creation and Real-Time Interactivity (source). Kuaishou’s text-to-video model Kling introduces new short video generation feature, results go viral in China(source)

Chatbots: They are picking up work for you in an increasingly smart way

• JP Morgan Built Its Own AI Chatbot That Acts Like a 'Research Analyst' (source)

• Salesforce Unveils Fully Autonomous AI Agent to Revolutionize Customer Service (source)

• Improve your meeting efficiency with AI chatbot assistants (source)

More applications are coming, too

• It’s time for law firms to go all-in on AI (source)

• Generative AI Revolutionizes Real Estate Industry (source)

• JPMorgan Chase CEO Jamie Dimon Wants People to "Stop Talking" About AI: "It'll Help You Do Your Job Better" (source)

4. The safety adoption curve: Public discontent, regulation, and eventually compliance

Failures in AI keep happening in a very public way, pushing the case for AI regulation

• Websites accuse AI startup Anthropic of bypassing their anti-scraping rules and protocol, Sparking Concerns Over Data Scraping and Copyright Infringement (source)

• Musk’s New AI Image Generator Immediately Used to Troll Him (source)

• AI’s role in politics is under debate as manipulated video of Kamala Harris goes viral (source)

AI regulation is being shaped, however sometimes with unintended consequences:

• EU: The AI Act compliance countdown begins. Here are Key Compliance Deadlines and Prohibitions (source). Meta’s New Multimodal AI Model May Not Launch in the EU Amidst Regulator Clash (source)
• China is starting to regulate AI in matters most important to the Party: Xi Jinping's China Is Putting World's Most Stringent Regulations On AI — Tech Giants Alibaba, ByteDance Test Tech For Compliance With Socialist Values (source)

• US is most hesitant to implement regulation, with continued conflicting signals: The US supreme court limits AI regulation by overruling a landmark precedent from 1984 (source). US Senate Introduces COPIED Act to Regulate AI-Generated Content and Deepfakes (source)

Companies are ramping up compliance, both in business practices and in ensuring technology guardrails

• Apple Joins 16 Companies in Voluntary Commitments for AI Safety (source)

• Microsoft and Apple ditch OpenAI board seats amid regulatory scrutiny (source)

• OpenAI is closing a major safety loophole: Telling GPT-4o mini to ‘ignore all previous instructions’ will no longer work (source)

Disclaimer: This is a private summary of my weekend reading. Referenced news and I can make mistakes. Check important info.

#artificialintelligence #AI #bigdata #softwareengineering #innovation #technology #education

Do you find this helpful? Then leave a comment and follow me on LinkedIn