-
[chatgtp]
-
[ai-code]
-
[ai-image] processing
-
Rob people of the opportunity to get good
-
Once AI is in place in organizations - it is shadow upgradable - a/b testing +++
-
AI destroys capitalism and the model of 'Work' - Poletics are not able to comprehend this.
-
Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRs 2025
-
We introduce a framework to investigate this question by analyzing four key cognitive behaviors -- verification, backtracking, subgoal setting, and backward chaining -- that both expert human problem solvers and successful language models employ
-
-
What Shapes Do Matrix Multiplications Like? [medium]
- Larger matrix's take less space and become better? what?
- Look at the maths
-
What Shapes Do Matrix Multiplications Like? [medium]
- Maths breakdown of computer and ram for tensor operations
-
TODO AI:Apple-Intelligence - Apples approach to AI
-
The future called ... it said most jobs are now obsolete. AI That can perform tasks by interpreting a computer screen
-
The workers at the frontlines of the AI revolution 2023
- How 4 jobs have been transformed by AI in last year
-
-
For all their brilliance, artificial neural networks remain as inscrutable as ever. As these networks get bigger, their abilities explode, but deciphering their inner workings has always been near impossible. Researchers are constantly looking for any insights they can find into these models.
-
-
The elephant in the room in many education settings.
-
Mass public misunderstanding
-
TODO this section needs fleshing out MASSIVELY!
-
- AI Workbooks, like jupyter notebooks. Combine differnt AI's in a single place
-
- AI Pre Trained Models - Datasets - Workbooks and Tutorials
-
FakeToxicityPrompts: Automatic Red Teaming
- Modern LLMs are disappointingly easy to trick into toxicity
-
- 29 March 2023, the UK government published its whitepaper titled “a pro-innovation approach to AI regulation”, (the Whitepaper) which sets out the UK’s proposed regulatory framework for AI.
- The principles are as follows:
- Safety, security and robustness. AI should function in a robust, secure and safe way throughout its lifecycle and risks should be continually identified, assessed and managed.
- Appropriate transparency and explainability. AI should be appropriately transparent and explainable, i.e., it should provide information on how, when and for what purposes AI is being used, while also making it possible for parties to access, interpret and understand its decision-making processes.
- Fairness. AI should not undermine the legal rights of individuals or organisations, discriminate unfairly against individuals or create unfair market outcomes.
- Accountability and governance. Governance measures should be in place to ensure effective oversight of the supply and use of AI, with clear lines of accountability established across the AI life cycle.
- Contestability and redress. Impacted third parties or users of AI should be able to contest an AI decision or outcome that is harmful or creates material risk of harm.
- https://publiclawproject.org.uk/content/uploads/2023/06/AI-alternative-white-paper-in-template.pdf ‘Alternative AI White Paper’ in June 2023, criticising the approach taken
-
- PhotoPrism® is an AI-Powered Photos App for the Decentralized Web. It makes use of the latest technologies to tag and find pictures automatically without getting in your way. You can run it at home, on a private server, or in the cloud.
-
Calculating Sunflower Oil Production
- A disturbing unravelling of ai
-
MetaGPT: Meta Programming for Multi-Agent Collaborative Framework August 2023 *
-
[llm]
-
Coding Train Beginner's Guide to Machine Learning in JavaScript with ml5.js Youtube Playlist
-
- New algorithms will transform the foundations of computing
- AlphaDev uncovered a faster algorithm for sorting, a method for ordering data.
-
-
A free course designed for people with some coding experience, who want to learn how to apply deep learning and machine learning to practical problems.
-
-
The Little Book of Deep Learning François Fleure (2023?) pdf free
-
type.ai - Launch HN: Type (YC W23) – AI-powered document editor
-
We're building an AI-first document editor that helps you write.
-
-
Understanding Large Language Models 2023
- A Cross-Section of the Most Relevant Literature To Get Up to Speed
-
Understanding ChatGPT - top level diagrams and understanding
-
ChatGPT + Code Interpreter = Magic - using ChatGPT for foundations of games
-
The 70% problem: Hard truths about AI-assisted coding - A field guide and why we need to rethink our expectations
- AI for juniors is a house of cards
-
- A sematic way to identify content that is generated by humans
- Probably quite an important item of metadata on future content
-
- Cheat in remote interviews
- Whisper & GPT-based app for passing remote SWE interviews
-
What is AGI-hard - At this point, knowing what AI can't do is more useful than knowing what it can
-
The Waluigi Effect (mega-post)
-
Definition (half-joking): A large language model is a structural narratologist.
- Jailbreaking to summon waluigis
-
-
AI Forecasting: One Year In July 2022
- Metrics from AI's ability to work on known datasets
-
Did Google's LaMDA chatbot just become sentient? YouTune 22min 2022
- Nice description of why the google chatbot was thought to be sentient
- Suggestible persona based on volume of data
- Google Engineer on His Sentient AI Claim 10min
- A small handfull of people have control of the way ai should respond to topics such as religion
-
This New A.I. Can Write Anything, Even Code (GPT-3) 15min 2022
- Will humans even write code in 2040 and what would that mean for extreme heterogeneity in computing? 2017
- GPT-3 examples of explaining code and summering documents. Writing small apps
- Generate faces
- Have conversations with simulated dead people
- Just a tool that can write sentences, but it does not understand them
-
Using GPT-3 to explain how code works
- AI can explain code in clear english sentences
- (once again) avoid leading questions "Which is most efficient?" (the real is - "it dosn't matter")
https://experiments.withgoogle.com/collection/ai https://machinelearningforkids.co.uk/ https://teachablemachine.withgoogle.com/
https://www.appsforgood.org/courses/machine-learning
-
MLU-EXPLAIN - Visual explanations of core machine learning concepts (Amazon)
- Train, Test, Validation
- Precision and recall
- Random forest
- Decision trees
- Bias variance trade-off
- Double descent
- visual
- mathematical
-
Starting Your AI Career Has Never Been Easier Andrew Ng, Standford Online
- One of the world’s most popular machine learning courses is getting updated to become even more accessible for beginners.
-
Summary of Envisioning AI for K-12: What should every child know about AI?
radiologists move to differnt hospital - ai cant be transfered
-
Coding Challenge #158: Shape Classifier Neural Network with ml5.js Coding Train 36min
- He trains a network himself with drawing shapes and it recognises shapes drawn on paper!
- This could be an AMAZING activity
- What about training the algorithm with lines - and then trying to recognise a filled circle. Could this be a powerful lesson in the data we tain algorithms with? (like the issues big tech companies had with recognising black faces with login systems)
-
DeepMind: Generally capable agents emerge from open-ended play
- Beautiful visual description of general purpose game playing AI.
- Tasks for the AI to overcome are created procedurally
-
Explaining Machine Learning Predictions - State-of-the-art, Challenges, and Opportunities
-
Why TensorFlow is dying a slow death for Python
- TensorFlow is no longer being developed
- PyTorch is expanding
-
Machine Learning: The Great Stagnation
- (I need to read up on AI and matrix's in general)
- The writer talks about AI research in general stagnating - there are some interesting AI research out there
-
What I Wish Someone Had Told Me About Tensor Computation Libraries
-
Introduction to Reinforcement Learning with David Silver
- Explore the concepts and methods used in modern reinforcement learning research.
- 10 part course
-
Uppestcase and Lowestcase Letters 12min
- Using [ai] to find the uppercase of an uppercase letter and the lowercase of a lowercase letter.
- Some really ugly font output
- Really well described video
-
Ai mid 2021. Self driving car meets reality
- Size of dataset. In one city it works. It wont work across the world with different vehicles and terrain
- AI better than radiologists ... but move it the hospital down the road, with slightly different imaging and it wont work and needs to be retrained from scratch, where as a radiologist can just walk down the road.
-
Large Language Models: A New Moore's Law?
- You only need super large models to train a model
- Use a pre-trained? or use the cloud (you don't need your own)
- [datasets]
Blockgeni Education - AI, Blockchain, DataScience, Crypto engineering
World Models - AI paper to teach 2d car how to drive
- MarIO - AI plays Super Mario World
- learnfun & playfun: A general technique for automating NES games
Open Source Tools & Data for Music Source Separation
- Self-Organising Textures
- Neural Cellular Automata Model of Pattern Formation
- Growing textures
The Elements of Statistical Learning: Data Mining, Inference, and Prediction
AI is made by men (??!?)
-
Data scientist Cathy O’Neil on the cold destructiveness of big data
-
Stanford AI Lab Blog: An Introduction to Knowledge Graphs
- [ai-text-to] #stable-diffusion
- Text2LIVE:: Text-Driven Layered Image and Video Editing
- Splatter Image: Ultra-Fast Single-View 3D Reconstruction
- Rotating 2d images with prediction!
- base64.ai
- Extract text, photos, and signatures from all document types
- No more manual data entry
- [automation]
- AI wins state fair art contest, annoys humans - Stealth win for AI-generated art inspires heated ethics debate on social media.
- The computers rejecting your job application
- Frankly, it was a little stressful to know that my application was being judged by a computer and not by a human being.
- A professional journalist, I had recently applied for a new job, and for the first part of the recruitment process the publisher made me play a number of simple online games from the comfort of my own home. These included having to quickly count the number of dots in two boxes, inflating a balloon before it burst to win money, and matching emotions to facial expressions.
- Then an artificial intelligence (AI) software system assessed my personality, and either passed or failed me. No human had a look-in.
- I wondered: is it fair for a computer alone to accept or reject your job application?
-
Building games and apps entirely through natural language using OpenAI’s code-davinci model
-
OpenAI has a new code generating model that’s improved in a number of ways and can handle nearly two times as much text (4,000 tokens.) I built several small games and applications without touching a single line of code. There are limitations, and coding purely by simple text instructions can stretch your imagination, but it’s a huge leap forward and a fun experiment. All the demos can be played with here:
- [game]
-
-
OpenAI Five Beats World Champion DOTA2 Team 2-0
- it's not a competition
-
- From an MSpaint style image, create photorealistic landscapes
-
It’s Not Just Size That Matters: Small Language Models Are Also Few-Shot Learners
-
... enormous amounts of compute are required for training and applying such big models, resulting in a large carbon footprint and making it difficult for researchers and practitioners to use them. We show that performance similar to GPT-3 can be obtained with language models that are much “greener” in that their parameter count is several orders of magnitude smaller. This is achieved by converting textual inputs into cloze questions that contain a task description, combined with gradient-based optimization; exploiting unlabeled data gives further improvements. ...
-
-
An argument for the impossibility of machine intelligence 2021, Landgrebe, Smith
-
Consciousness is not computation
-
computer simulations can never produce the entity they are simulating since the “simulation” is not an independent system. It is only defined in terms of its relation to an external observer. A simulation of a brain cannot produce consciousness any more than a simulation of the weather can produce rain.
-
Unfortunately in arguing that consciousness is not computation, we end up struggling to imagine how the mind can compute anything at all.
-
There is something special going on in the cerebral cortex, and only the cerebral cortex, that seems to produce consciousness
-
-
Powerful ‘Machine Scientists’ Distill the Laws of Physics From Raw Data
- [data]
-
🦾 codex_py2cpp 🤖 - OpenAI conversion from [python] to [c]++
-
Experiment 1: Creating a landing page using AI tools + No code
-
FlexGen Running large language models on a single GPU for throughput-oriented scenarios.
- Large models are too big for single system GPU's. We can split it into 'chunks' and process lots of AI steps in increments in batches. This enables single GPU system to process bulk AI requests at the expense of latency (overnight?)
-
Superhuman: What can AI do in 30 minutes?
- Whole draft marketing campaign
-
90% of My Skills Are Now Worth $0 ...but the other 10% are worth 1000x
- AI can now put words in order - writing is not that hard anymore
-
Generative Agents: Interactive Simulacra of Human Behavior
- The sims +++
- Could be used to simulate ideas and help integrate people socially
-
GPT makes learning fun again 2023
-
Recently, I tried learning about a new subject using two approaches: 1. Google searching, 2. talking to GPT about it. I found, for the eighth time this month, that talking to GPT was far better than searching. The contrast was so striking that I thought it's worth dissecting the differences to see why the GPT experience was so much better.
-
-
The Dual LLM pattern for building AI assistants that can resist prompt injection
- The interesting problem with AI assistants
-
-
Blotter combines generative AI with song recognition technology to churn out audio-reactive visuals 24/7, all in real time. All songs licensed via Soundstripe.
-
-
I’m an ER doctor. Here’s how I’m already using ChatGPT to help treat patients.
- Explaining things to distressed people takes time - ChatGTP can create empathic responses to free up doctors time
-
Since this incident, I’ve taken to using ChatGPT to help empathically explain specific medical scenarios to patients and their loved ones. It’s become an invaluable resource for the frequent situations where my ER ward is too busy or short-staffed for explaining complex medical diagnoses in a way that is accurate but easy to understand
-
ArtPrompt: ASCII Art-based Jailbreak Attacks against Aligned LLMs
- Ascii art bypass's the filters people try to put in place