Chaque semaine, nous sélectionnons des news Tech, Data et IA pertinentes que nous analysons grâce à l'intelligence artificielle et que nous transformons en podcast (10 minutes) dont les speakers sont des voix générées aussi par IA.

‍

Venez écouter les actus Tech de la semaine ou faire votre veille grâce à ce mini podcast. Inscrivez-vous ici pour ne manquer aucun des prochains podcasts.

Au programme :

⛔️ An #AI Leader Urges Regulation and a Rethink

🧠 💻 Your brain does not process information and it is not a computer

🔥 Mojo is here (@Modular)

🧅 Bun 1.0 is finally here

🕵️ What @OpenAI Really Wants

‍

‍

Les précédents épisodes sont toujours disponibles. 

‍

An AI Leader Urges Regulation and a Rethink

‍

In a spirited call-to-action, tech executive Mr. Suleyman is pushing for increased government regulation on Artificial Intelligence (AI). He proposes that nations should have cabinet-level tech expert ministers and further suggests that the US could use its preeminence in chip technology to establish global standards. He envisions a governance model akin to the Intergovernmental Panel on Climate Change to oversee, and ensure transparency in private companies operating in the AI sector.

These proposals come on the heels of growing global deliberations about how AI ought to be administrated. This week, Senate leader Chuck Schumer has arranged a meeting with top tech tycoons including Elon Musk and Microsoft CEO, Satya Nadella, to confer on these issues.

Mr. Suleyman chose to present his ideas in a book format for "radical accountability", preferring this method as more conclusive evidence of his thoughts, as compared to a series of tweets or blog posts. He describes his book as a "love letter" to the nation-state, a wakeup for policymakers and citizens. He emphasizes the importance of noncommercial checks and balances, which he argue as essential in a system that gravitates away from absolute power and dictatorship, and lean towards a liberal democracy. A firm believer in the nation-state's potential, he urges continuation of efforts to uphold it.

Source 👉 https://www.nytimes.com/2023/09/10/business/dealbook/an-ai-leader-urges-regulation-and-a-rethink.html (Paywalled)

‍

Your brain does not process information and it is not a computer

‍

Despite leading theories in brain science and cognitive psychology, humans do not process information or hold data in the brain the same way computers do. This misconception stems from the 1940’s proliferation of computers, with its impact seen in fields such as psychology, neuroscience and linguistics. This comparison lacks substance, as humans are not born with the design elements allowing computers to process information intelligently, like algorithms or software. Unlike computers, humans do not store and process information using patterns of ones and zeros.

There is also no direct comparison between the human mind and any form of artificial intelligence. People do not download, store and process information from the environment the same way a computer would, and human behaviour is influenced by a vast variety of experiences, which changes brain function. These changes are built on the individual's genetically unique neural structure, developed over a lifetime of unique experiences.

Today, despite major advancements in both computer technology and brain research, and a mainstream tendency to compare human mind to a computer, the mystery of the mind remains largely unsolved. Revolutionary projections like downloading human minds to achieve immortality, are fundamentally flawed and cannot be realised, given the human brain functions nothing like a computer. Even if we could simulate brain neurons in a computer, that pattern would mean nothing outside the body of the brain that produced it.

Moreover, trying to understand the complexity of the brain to maintain human intellect is extremely daunting due to profundity of its neuronal connectivity, strengths and states of proteins that exist at each connection point, as well as variations in everyone's brain structures. Therefore, the approach of treating human brains as computers is counter-productive, and should be discontinued to focus on understanding ourselves.

Source 👉 https://aeon.co/essays/your-brain-does-not-process-information-and-it-is-not-a-computer

‍

Mojo is here (@Modular)

In a significant step forward for the Mojo programming language, it is now available for local download, initially for Linux systems, with Mac and Windows to be added in future updates. Mojo is a programming language specifically for AI developers that aspires to be a superset of Python. It can integrate with arbitrary Python code and aims to offer better performance in performance-critical systems, including AI accelerators like GPUs.

The Mojo platform, introduced in May, has already attracted over 120K developers. The local Mojo toolchain is now providing developers with access to the full set of compiler features and tools which greatly assists in building Mojo applications.

A notable benefit of mojo is its ability to perform at a level close to C++ and CUDA, traditionally considered superior in instances requiring high performance or specialized hardware. Moreover, it provides access to the full Python ecosystem, leading to seamless integration with Python libraries while also providing performance advantages.

Mojo is also designed for integration with the Modular AI Engine, facilitating the extension of AI workloads with customizable operations, high-performance mathematical algorithms among others. Frequently, with only minor changes to existing Python code, substantial speed boosts are achievable.

In addition to being available for download to your computer, Mojo also provides a powerful function beyond being merely a compiler. It provides essential tools such as a VS Code Extension, supporting syntax highlighting and code completion among others, a shell via the mojo driver, and a Jupyter kernel for building and running Mojo notebooks. The initial SDK release supports X86/Linux systems only, with more systems and features set for future updates.

Community engagement is encouraged, with over 19K developers actively discussing Mojo across platforms such as Discord and GitHub. There are plans to commence open-sourcing sections of Mojo later within the year, in what promises to be continuous language improvements in the future.

Source 👉 https://www.modular.com/blog/mojo-its-finally-here

‍

Bun 1.0 is finally here.

Bun, a fast and all-inclusive toolkit designed to manage JavaScript and TypeScript development, is finally stable and ready for production use. The toolkit aims to streamline the process of running, building, testing, and debugging, offering a solution to the slow and complex JavaScript tooling that has evolved over the years. Bun is pitched as a practical replacement for numerous existing tools and functions, aiming to enhance process without significantly altering the developer's workflow.

One of its major strengths is being a drop-in replacement for Node.js, making tools such as ‘node’, ‘npx’, and ‘nodemon’ superfluous. Bun can manage different types of files, thereby replacing transpilers like ‘tsc’ and ‘babel’. It is also an efficient JavaScript bundler and an npm-compatible package manager, eliminating the need for tools like ‘esbuild’, ‘webpack’, and ‘parcel’. Further, Bun functions as a Jest-compatible test runner thus replacing ‘jest’ and other testing libraries.

Built using Apple's WebKit engine rather than Google's V8 engine as in Node.js, Bun provides a faster and more efficient runtime experience. Its support for both the CommonJS and ES modules enhances its compatibility. Additionally, it has built-in support for Web standard APIs and offers hot reloading to boost productivity.

Bun offers advantages such as increased speed, customizable features, and effective management of full-stack applications. With promised improvements in API design and developer experience, Bun aims to make software development faster, more efficient, and enjoyable. With its debut, the JavaScript community can anticipate a smoother and more straightforward development process.

Source 👉 https://bun.sh/blog/bun-v1.0

‍

What OpenAI Really Wants

‍
Sam Altman, the 38-year-old entrepreneur and co-founder of OpenAI, is currently on a worldwide tour that has already covered 25 cities across six continents. He recently visited London, where he traveled from one event to another, engaging with various audiences eager to gain insight into the rapidly evolving world of artificial intelligence (AI).

Despite not actively participating in the research or development phases of OpenAI's groundbreaking products like the Chabot, ChatGPT or its advanced counterpart GPT-4, Altman has been recognized as a central figure in this AI revolution. His photo has often been used as the symbol of new-age challenges that AI poses to humanity in various prominent news articles. He also serves as the representative figure for OpenAI's visual AI product, Dall-E.

Altman fielded numerous questions from different spheres, be it from the government, academia, the industry or just interested bystanders. The questions revolved around the potential dangers of AI, the possibilities of its regulation and China's role in this field.

His schedule in London was tight, starting with an off-the-record session with the Round Table, a diverse group of professionals from different sectors. Later, he had a conversation at the Londoner Hotel with 600 Oxford Guild members, followed by a Q&A session with 100 engineers and entrepreneurs. He rounded off the day with an on-stage talk at University College London that drew a varied crowd of academics, tech enthusiasts, and journalists. No matter the audience, Altman managed to answer all queries comprehensively while constantly staying tuned to the world online. His AI endeavors promise to be a game-changer, and people worldwide are eager to grasp what this means for the future of humankind.

Source 👉 https://www.wired.com/story/what-openai-really-wants/

‍

‍