AI Today and Everyday — April 15, 2024
What’s new in the world of AI
Big upgrade for ChatGPT users
- OpenAI has announced that paying ChatGPT Plus, Team, or Enterprise subscribers will now have access to its upgraded and improved ChatGPT-4 Turbo model.
- The model improves ChatGPT’s writing, math, logical reasoning, and code generation. It also helps ChatGPT understand longer prompts and delivers more conversational responses.
- It also offers more recent responses, as it’s been trained on information up to December 2023 (compared to the previous version, which was trained on data until April 2023).
Musk launches “Vision” Grok to rival ChatGPT-4
Our Report: Following the release of Grok-1.5 LLM (Large Language Model) last month, Musk’s AI company — xAI — has launched Grok-1.5 Vision (Grok-1.5V), an AI chatbot that can process text and visual information (which the original Grok-1.5, couldn’t do).
Key Points:
- Grok-1.5V reportedly performed better than its rivals — including GPT-4, Claude, and Gemini Pro — in RealWorldQA benchmark tests that measure real-world understanding.
- It can process visual information — including documents, diagrams, screenshots, and photographs — translating diagrams into code and drawings into stories.
- Grok-1.5V isn’t available yet, but it is “coming soon” to early testers and existing Grok users, as they improve its capabilities to process images, audio, and video.
Why you should care: Musk wants Grok-1.5V to write tweets for X (formerly Twitter) Premium users, despite reports that Grok has recently created and promoted dangerous fake stories, raising concerns among its developers.
Apple Macs getting AI overhaul
Image source: Bloomberg
The Rundown: Apple is reportedly gearing up to overhaul its entire lineup of Macs with an upcoming M4 chip family — which will place a heavy emphasis on AI capabilities.
The details:
- The M4 chips are nearing production and will come in at least three main varieties: entry-level Donan, high-performance Brava, and top-end Hidra.
- Apple plans to update every Mac model with M4 chips, with releases starting in late 2024 and extending into early 2025.
- Initial M4 refreshes will include new iMacs, MacBook Pros, and Mac Minis, followed by MacBook Airs, Mac Studio, and Mac Pro.
- More details and features are expected at the company’s developer conference on June 10.
Why it matters: With slumping Mac sales and Apple seen as lagging behind in AI, a rapid M4 transition with next-gen capabilities could inject some serious energy into the device lines. Expect some fireworks in June — with Apple likely itching to publicly announce its entrance into the AI frenzy.
Create mind maps in seconds with ChatGPT
The Rundown: In this tutorial, you’ll learn how to transform complex ideas into visual and informative mind maps for free using ChatGPT.
Step-by-step:
- Head over to ChatGPT. You can either use GPT-3.5 (free version) or GPT-4 (paid version).
- Write the following prompt: “Create a mind map of [Your Topic]. List topics as central ideas, main branches, and sub-branches.”
- Once ChatGPT generates the initial mind map outline, ask for a markdown format: “Create this same mind map in markdown format.”
- Paste the markdown into Markmap and watch your mind map come to life.
Bonus tip: You can customize your mind map’s appearance and then download it as interactive HTML or static SVG.
OpenAI fires two researchers for leaking info
Image source: The Information
The Rundown: According to a new report from The Information, OpenAI has terminated researchers Leopold Aschenbrenner and Pavel Izmailov for allegedly leaking information outside of the company.
The details:
- Leopold Aschenbrenner was part of OpenAI’s “superalignment” team and an ally of chief scientist Ilya Sutskever, with Pavel Izmailov also spending time on the AI safety team.
- The nature of the leaked information is unknown, though rumors of an AI breakthrough called Q* surfaced in November during OpenAI’s board turmoil.
- Sutskever was one of the primary figures in the 2023 firing of Sam Altman, and his role has been unclear since Altman’s reinstatement.
Why it matters: Although the OpenAI internal tensions have been quieter in 2024, these firings indicate there is still plenty of drama left to unfold. While the famous question of ‘What did Ilya see?’ has been refuted by Altman, firing a reported ally of Sutskever will only add more fuel to the fire.
The “first AI software engineer” was exposed for fraud.
Image credit: AI Secret
Devin, an AI once hailed as the “first AI software engineer,” was involved in a scandal when its proclaimed abilities were discredited by YouTuber “Internet of Bugs.” This revelation stirred debates on AI ethics and reliability in software engineering, underscoring the risk of AI exaggerating capabilities.
The Details:
- Unrealistic Demonstrations: Devin was shown solving tasks that were either non-existent or unnecessary, indicating a significant discrepancy between its advertised and actual capabilities.
- Creating and Solving Own Bugs: Devin was caught creating bugs only to fix them later, a process that does not align with genuine software engineering practices.
- Overcomplicated Solutions: The solutions provided by Devin often involved outdated or overly complex methods not suitable for modern programming practices.
Why It Matters:
- Ethics and Trust: The incident underscores the importance of transparency and honesty in demonstrating AI capabilities, as misleading demonstrations can erode trust in emerging technologies.
- AI’s Role in Software Engineering: It highlights the limitations of current AI technologies in understanding and performing complex tasks independently, challenging the notion that AI can soon replace human software engineers.
- Impact on Industry Perception: Such controversies can influence the perception of AI’s reliability and effectiveness in practical applications, affecting funding, research priorities, and public acceptance.
Google DeepMind robots play soccer
Image source: Google DeepMind
The Rundown: Researchers at Google DeepMind and the University of Cambridge just taught miniature humanoid robots to play soccer against each other, showcasing complex skill learning and agile adaptability.
The details:
- The researchers first taught basic skills like walking and getting up, then had robots practice playing against gradually improving versions of themselves.
- Compared to the baseline robots, the AI-trained versions walked almost 3x as fast, turned around 5x quicker, and got up from falls 63% faster.
- Researchers found the AI helped learn clever strategies, like taking short, quick steps when defending without being explicitly told to do so.
Why it matters: While these clumsy but adorable robots won’t be heading to the World Cup any time soon, the research demonstrates the power of AI in enabling complex skill learning and adaptability — even leading to developing unique behaviors and tactics on their own.
Quick AI News
- Apple AI — Apple’s upcoming AI unveil on June 10 is expected to operate entirely on device, with no cloud processing component.
- Light-Powered Chiplet — China’s new photonic chiplet “Taichi” promises to power future AGI with its superior scalability and energy efficiency.
- AI in Academics — A recent study reveals that ChatGPT frequently concocts scientific citations that appear believable but don’t refer to actual academic work.
- AI in Film Restoration — New film restoration processes using AI have raised objections amid accusations of creating an “unnatural” look.
Trending AI Tools
- Strut- Collaborative workspace for all writing needs
- Musho.ai- Prompts to dev-ready UI design
- mnml.ai- One click sketch to architecture render
- Sonauto- Music editor transforming lyrics and melodies into full songs
- Persana- All-in-one AI prospecting platform
- Bliki AI- Plan eco-friendly trips with smart itinerary suggestions
