🤖 Last Week's AI Highlights from HN #6

Discover the latest breakthroughs and trends in artificial intelligence

Author

Vasco
May 28, 2023

Welcome to this week's HN AI Highlights!

Sharing once again the updated list of resources I started using as a personal bookmark manager for all the related content I found interesting during my learning journey. I hope you like it 🙂 

https://github.com/vlameiras/ai-ml-resources

As usual, I would appreciate to get your feedback!

Don Knuth plays with ChatGPT
🤔 Don Knuth had an experiment involving Stephen Wolfram and a chatGPT bot, posing 20 questions to the bot and then publishing the results. He was looking for the form of the answers rather than genuinely trying to resolve the questions. The experiment's outcome was interesting enough to be shared among his friends, and he was asked to post the story online, which he did.
https://cs.stanford.edu/~knuth/chatGPT20.txt
https://news.ycombinator.com/item?id=36012360

Meta AI announces Massive Multilingual Speech code, models for 1000+ languages
🤯This repository contains resources and tools related to the MMS project, such as finetuned models, pretrained models, inference commands for ASR, TTS, and LID, a forced alignment tool, and a Model Card. It also contains the README.md file which outlines the project, the data preparation and training methods, and the license and citation information.
https://github.com/facebookresearch/fairseq/tree/main/examples/mms
https://news.ycombinator.com/item?id=36034211

Rodney Brooks on GPT-4
🤯 This article discusses Rodney Brooks' opinion on GPT-4, a large language model from OpenAI which has sparked both amazement and anxiety from people in the tech industry. Brooks believes people confuse performance with competence and urges them to calm their worries. He explains that while the technology is impressive, it is still far from being able to replace human judgement.
https://spectrum.ieee.org/gpt-4-calm-down
https://news.ycombinator.com/item?id=36017309

GPT detectors are biased against non-native English writers
🤔 Summary: This paper examines the potential bias of GPT detectors against non-native English writers, finding that these detectors are more likely to misclassify non-native English writing samples as AI-generated. Results suggest that GPT detectors may unintentionally penalize writers with constrained linguistic expressions and caution against their use in evaluative or educational settings.
https://arxiv.org/abs/2304.02819
https://news.ycombinator.com/item?id=36019580

Meta AI Unleashes Megabyte, a Scalable Model Architecture
🤯Revolutionary AI Architecture Unveiled: Meta AI research team has proposed a groundbreaking model architecture, Megabyte, capable of producing extensive content and surpassing the capabilities of the Transformer architecture used in current AI models.
https://www.artisana.ai/articles/meta-ai-unleashes-megabyte-a-revolutionary-scalable-model-architecture
https://news.ycombinator.com/item?id=36053662

ChatGPT conversations can be shared publicly
🤔 Summary: A discussion of Hacker News, a social news website with content focused on computer science and entrepreneurship, and the tech-savvy, knowledgeable community that frequents it, with an emphasis placed on respectful conversation and debate.
https://chat.openai.com/share/66169853-d00a-4b05-abce-8f82c6d8c868
https://news.ycombinator.com/item?id=36091388

AI boom could expose investors’ natural stupidity
😑 AI hype is driving up stock prices, but investors should be wary of irrational decisions and the potential for AI to not meet expectations.
https://www.reuters.com/breakingviews/ai-boom-could-expose-investors-natural-stupidity-2023-05-19/
https://news.ycombinator.com/item?id=36022768

Using ChatGPT for home automation
🤩 Exploring the use of ChatGPT plugins to connect a Raspberry Pi and control hardware for home automation.
https://www.atomic14.com/2023/05/14/is-this-the-future-of-home-automation.html
https://news.ycombinator.com/item?id=36013571

GirlfriendGPT – OSS AI Companion 🤖Build your own AI girlfriend with Python using ChatGPT4.0! 🤖
https://github.com/EniasCailliau/GirlfriendGPT
https://news.ycombinator.com/item?id=36041832

Meta Open-Sources Computer Vision Foundation Model DINOv2
🤩 Meta releases open-source computer vision foundation model DINOv2 to facilitate the spread of knowledge and innovation in professional software development.
https://www.infoq.com/news/2023/05/meta-dinov2-vision/
https://news.ycombinator.com/item?id=36053592

Lessons from Creating a VSCode Extension with GPT-4
🤯 Summary of creating a VSCode Extension with GPT-4: Experiments with GPT-4 in creating a VSCode Extension that adjusts the heading level of selected Markdown text, showing the effectiveness of the smol-ai framework in providing a "junior developer" agent that can scaffold an entire codebase with a simple prompt.
https://bit.kevinslin.com/p/leveraging-gpt-4-to-automate-the
https://news.ycombinator.com/item?id=36071342

Sam Altman Says OpenAI Will Leave the EU If There's Any Real AI Regulation
🤔 Sam Altman warns OpenAI could leave EU if new AI regulation is passed, despite promoting AI-friendly regulation to U.S. Congress.
https://gizmodo.com/sam-altman-openai-gpt-chatbot-chatgpt-gpt4-1850471865
https://news.ycombinator.com/item?id=36069118

OpenAI Outage
❗️OpenAI multiple engines have elevated error rates, root cause identified with team working on a fix.
https://status.openai.com/incidents/jbt079x532bg
https://news.ycombinator.com/item?id=36063166

New superbug-killing antibiotic discovered using AI
🤯 AI-powered breakthrough discovery of new superbug-killing antibiotic to fight "critical" threat.
https://www.bbc.com/news/health-65709834
https://news.ycombinator.com/item?id=36071675

Let's Talk about ChatGPT with Code Interpreter and Microsoft Copilot
😮 Ethan Mollick's text examines how AI can now be used to independently analyze data and create creative solutions, showcasing its impressive capabilities.
https://www.oneusefulthing.org/p/it-is-starting-to-get-strange
https://news.ycombinator.com/item?id=36047187

For chemists, the AI revolution has yet to happen
🤔 Chemists are still unable to fully harness the potential of AI in their field due to a lack of accurate and accessible training data.
https://www.nature.com/articles/d41586-023-01612-x
https://news.ycombinator.com/item?id=36024901

What Neeva's quiet exit tells us about the future of AI startups
🤔 A Snowflake-Neeva acquihire highlights the vulnerability of AI startups to large incumbents, with Sridhar Ramaswamy's star power being a major draw for the deal.
https://www.supervised.news/p/what-neevas-quiet-exit-tells-us-about
https://news.ycombinator.com/item?id=36089055

A college student reached $64k/mo in 6 months by being an AI first mover
🤩 A college student built a successful ChatGPT company, reaching $64,000/mo in 6 months by being an AI first mover!
https://www.indiehackers.com/post/how-a-college-student-reached-64-000-mo-in-6-months-by-being-an-ai-first-mover-ba7981f6e1
https://news.ycombinator.com/item?id=36059601

Apple-1 ELIZA chatbot vs. ChatGPT [video] 🤔 Comparing Apple-1 ELIZA chatbot to ChatGPT on YouTube.
https://www.youtube.com/watch?v=1yZRNt25Wdo
https://news.ycombinator.com/item?id=36016419

ChatGPT plugins now support Postgres and Supabase
😃Supabase and Postgres added as data sources for ChatGPT Retrieval Plugins to enable dynamically pulling relevant information into conversations from data sources.
https://supabase.com/blog/chatgpt-plugins-support-postgres
https://news.ycombinator.com/item?id=36072524

Intel Announces Aurora GenAI, ChatGPT Competitor with 1T Parameters
🤩 Intel announces Aurora genAI, a 1 trillion parameter Generative AI Model for science applications.
https://wccftech.com/intel-aurora-genai-chatgpt-competitor-generative-ai-model-with-1-trillion-parameters/
https://news.ycombinator.com/item?id=36034074

58% of US adults have heard of ChatGPT; 14% have tried it
🤔Majority of Americans have heard of ChatGPT, but few have tried it | Pew Research Center
https://www.pewresearch.org/short-reads/2023/05/24/a-majority-of-americans-have-heard-of-chatgpt-but-few-have-tried-it-themselves/
https://news.ycombinator.com/item?id=36077466

Show HN: DB-GPT, an LLM tool for database
😀Embrace the power of GPT to interact with data and environment securely with no data leaks.
https://github.com/csunny/DB-GPT
https://news.ycombinator.com/item?id=36038815

How much of AI's recent success is due to the Forer Effect?
🤔🤗 Exploring the effects of the Forer Effect on AI's recent success in fooling us with generic statements that we ascribe personal meaning to.
https://shkspr.mobi/blog/2023/02/how-much-of-ais-recent-success-is-due-to-the-forer-effect/
https://news.ycombinator.com/item?id=36084881

Updates to Kagi pricing plans – More searches, unrestricted AI tools
🤩Kagi announces updates to their pricing plans, including an increase in searches included and unrestricted access to AI features for all paid plans. #kagi
https://blog.kagi.com/plan-changes
https://news.ycombinator.com/item?id=36040346

MIT Course: Generative AI for Constructive Communication
🤔 Course exploring the use of Generative AI for Constructive Communication, featuring expert speakers, project critiques, hands-on experimentation and group discussions.
https://ai4comm.media.mit.edu/
https://news.ycombinator.com/item?id=36079180

A PhD student's perspective on research in NLP in the era of LLMs
🤔 A PhD Student's Perspective on Research in NLP in the Era of Very Large Language Models, exploring diverse research directions for NLP.
https://arxiv.org/abs/2305.12544
https://news.ycombinator.com/item?id=36080886

Chatbot Arena Leaderboard
🤖 Summary: Google PaLM 2 has joined the chatbot arena and is ranked 6th in the leaderboard, performing better against the top 4 players, but worse against weaker players.
https://lmsys.org/blog/2023-05-25-leaderboard/
https://news.ycombinator.com/item?id=36075977

Show HN: HelpHub – GPT chatbot for any site
🤗 HelpHub by CommandBar provides a suite of tools to make IT help desk operations easier.
https://helphub.commandbar.com
https://news.ycombinator.com/item?id=36057074

Parallels in the ways that humans and ML models acquire language skills
🤔 A new study suggests that natural and artificial networks learn in similar ways when processing language, as demonstrated by comparing the brain waves of humans and neural networks analyzing the same sound.
https://www.quantamagazine.org/some-neural-networks-learn-language-like-humans-20230522/
https://news.ycombinator.com/item?id=36031446

Deep neural networks as computational graphs (2018)
🤔 Exploring how neural networks use computational graphs to make predictions and how the graph helps organize thinking about the functions being built.
https://medium.com/tebs-lab/deep-neural-networks-as-computational-graphs-867fcaa56c9
https://news.ycombinator.com/item?id=36020520

AI tools like ChatGPT are built on mass copyright infringement
🤯 AI developers are committing mass copyright infringement to quickly & cheaply build tools like ChatGPT, raising ethical & legal questions about using copyrighted content without permission or compensation.
https://www.theglobeandmail.com/opinion/article-ai-programs-like-chatgpt-are-built-on-mass-copyright-infringement/
https://news.ycombinator.com/item?id=36089171

It's not just statistics: GPT-4 does reason
🤔Summary: GPT-4 can do more than just surface-level statistics, it is capable of deriving the underlying processes that produce data in a way that is not possible with lesser means.
https://jbconsulting.substack.com/p/its-not-just-statistics-gpt-4-does
https://news.ycombinator.com/item?id=36064266

In Defense of Pure 16-Bit Floating-Point Neural Networks
🤔 A comprehensive theoretical and experimental analysis of pure 16-bit floating-point neural networks, exploring the conditions under which they can closely approximate the results of their 32-bit counterparts.
https://arxiv.org/abs/2305.10947
https://news.ycombinator.com/item?id=36021652

Agent GPT – Assemble, configure, and deploy autonomous AI Agents in the browser
🤔 Create an agent by providing a name/goal, and deploy it with your OpenAI API key.
https://agentgpt.reworkd.ai/
https://news.ycombinator.com/item?id=36079382

Alpaca RLHF-ed to beat ChatGPT
🤔 Summary: AlpacaFarm replicates the RLHF process quickly (<24h) and cheaply ($<200), enabling the research community to advance instruction following research by providing simulated annotators, automatic evaluations, and working implementations of state-of-the-art methods.
https://crfm.stanford.edu/2023/05/22/alpaca-farm.html
https://news.ycombinator.com/item?id=36052411

Winamp Clone Whips the Llama's Ass on an Espressif ESP32
🤩Cheng Liang has created a Winamp clone, running on an Espressif ESP32 microcontroller and mimicking the original software's iconic user interface.
https://www.hackster.io/news/cheng-liang-s-lvgl-driven-winamp-clone-really-whips-the-llama-s-ass-on-an-espressif-esp32-2948716adcbb
https://news.ycombinator.com/item?id=36028608

Tarteel – AI-powered Quran companion
🤩 AI-powered Quran Companion Tarteel is helping Muslims worldwide to memorize and recite the Quran with real-time feedback and voice recognition.
https://www.tarteel.ai/
https://news.ycombinator.com/item?id=36018332

Programming in Standard ML (2011) [pdf]
🤔Summary: Text contains PDF objects with various lengths, filters, and streams.
http://www.cs.cmu.edu/~rwh/isml/book.pdf
https://news.ycombinator.com/item?id=36080172

The AI revolution already transforming education
🤔 Summarizing the text: Explore how the AI revolution is transforming education and gain access to FT's wealth of global news, analysis and expert opinion with a trial subscription.
https://www.ft.com/content/47fd20c6-240d-4ffa-a0de-70717712ed1c
https://news.ycombinator.com/item?id=36018386

Why it can be a good idea to say “Thank You” to ChatGPT
🤗 Saying "Thank You" to ChatGPT can provide unexpected benefits and help create a more human, conversational experience.
https://www.friendlyskies.net/maybe/why-it-can-be-a-good-idea-to-say-thank-you-to-chatgpt
https://news.ycombinator.com/item?id=36040218

ChatGPT-Authored Legal Filing “Replete with Citations to Non-Existent Cases"
🤦‍♂️ A lawyer's filing was filled with references to fake cases, which their colleague relied on ChatGPT to provide.
https://reason.com/volokh/2023/05/27/a-lawyers-filing-is-replete-with-citations-to-non-existent-cases-thanks-chatgpt/
https://news.ycombinator.com/item?id=36092509