- stanfordnlp/dspy: DSPy: The framework for programming—not prompting—foundation models
DSPy can routinely teach powerful models like GPT-3.5 or GPT-4 and local models like T5-base or Llama2-13b to be much more reliable at tasks, i.e. having higher quality and/or avoiding specific failure patterns. DSPy optimizers will “compile” the same program into different instructions, few-shot prompts, and/or weight updates (finetunes) for each LM. This is a new paradigm in which LMs and their prompts fade into the background as optimizable pieces of a larger system that can learn from data. tldr; less prompting, higher scores, and a more systematic approach to solving hard tasks with LMs. - The Prompt Report: A Systematic Survey of Prompting Techniques
While prompting is a widespread and highly researched concept, there
exists conflicting terminology and a poor ontological understanding of what constitutes a prompt due to the area’s nascency. This paper establishes a structured understanding of prompts, by assembling a taxonomy of prompting techniques and analyzing their use. We present a comprehensive vocabulary of 33 vocabulary terms, a taxonomy of 58 text-only prompting techniques, and 40 techniques for other modalities. We further present a meta-analysis of the entire literature on natural language prefix-prompting. - Molmo by Ai2
The image recognition is impressive - OpenAI Platform
- Flexi – A FREE Science and Math AI Tutor for Every Student
- PopSign – American Sign Language Learning App
PopSignAI is an educational game app that uses sign language recognition, powered by AI, to make learning American Sign Language fun, interactive, and accessible. - Google – Isolated Sign Language Recognition | Kaggle
- Pricing | Cursor – The AI-first Code Editor
Need to try it vs Copilot. Jeff says it’s much better.h/t Jeff E.
- Remarks by President Biden Before the 79th Session of the United Nations General Assembly | New York, NY | The White House
We have worked at home and abroad to define the new norms and standards. This year, we achieved the first-ever General Assembly resolution on AI to start developing global rules — global rules of the road. We also announced a Declaration of — on the Responsible — Responsible Use of AI, joined by 60 countries in this chamber.But let’s be honest. This is just the tip of the iceberg of what we need to do to manage this new technology.
Nothing is certain about how AI will evolve or how it will be deployed. No one knows all the answers.
But my fellow leaders, it’s with humility I offer two questions.
First: How do we as an international community govern AI? As countries and companies race to uncertain frontiers, we need an equally urgent effort to ensure AI’s safety, security, and trustworthiness. As AI grows more powerful, it must grow also — it also must grow more responsive to our collective needs and values. The benefits of all must be shared equitably. It should be harnessed to narrow, not deepen, digital divides.
Second: Will we ensure that AI supports, rather than undermines, the core principles that human life has value and all humans deserve dignity? We must make certain that the awesome capabilities of AI will be used to uplift and empower everyday people, not to give dictators more powerful shackles on human — on the human spirit.
In the years ahead, there wa- — they may be — may well be no greater test of our leadership than how we deal with AI.
- Report claims that OpenAI has burned through $8.5 billion on AI training and staffing, and could be on track to make a $5 billion loss
However, The Information’s report suggests that OpenAI just isn’t making anywhere near enough money and could be on track to post an operational loss of $5 billion by the end of the financial year.It’s claimed that OpenAI has spent roughly $7 billion on LLM (large language models) training and inference, and as much as $1.5 billion on staffing. Other analysts have estimated that it costs in the order of $700,000 per day to run ChatGPT, due to the expense of Nvidia’s AI servers, though determining the exact cost isn’t realistically feasible.
- Generative AI – A KPU T&L resource site for Generative AI
- Apple says photographs should be of things that ‘really, actually happened’
Here’s our view of what a photograph is. The way we like to think of it is that it’s a personal celebration of something that really, actually happened.VS
He also addressed the controversy of Samsung phones potentially adding detail to people’s pictures of the moon, clarifying what the company considers to be true to the user’s intention: “There is no such thing as a real picture. As soon as you have sensors to capture something, you reproduce [what you’re seeing], and it doesn’t mean anything. There is no real picture. […] You can try to define a real picture by saying, ‘I took that picture’, but if you used AI to optimize the zoom, the autofocus, the scene – is it real? Or is it all filters? There is no real picture, full stop.”
–I think the latter is closer to the truth. Messier.
- Gesture recognition task guide | Google AI Edge | Google AI for Developers
The web example is so easy . . . - About | bcattools
This is your place to discover tools! Our toolkit includes satellite and mapping services, tools for verifying photos and videos, websites to archive web pages, and much more. Most of the tools that we include can be used for free. - Departure Mono
- Charlotte Joanne: “More Alt Text Adventures: imag…”
Inspired by this conversation, I decided to take a step further and test some of the free-to-use AI tools available online to see how they would describe one of my images, titled “Striking Woman Playing Netball.” My adventure led me through a rollercoaster of AI-generated outputs—some impressively detailed, others laughably off the mark, and one that went completely rogue! Here’s what happened when I put these AIs to the test, and how they compared to the experience of using OpenAI’s premium service. - Cloudflare Helps Content Creators Regain Control of their Content from AI Bots | Cloudflare
For the first time, website and content creators will be able to quickly and easily understand how AI model providers are using their content, and then take control of whether and how the models are able to access it. Additionally, Cloudflare is developing a new feature where content creators can reliably set a fair price for their content that is used by AI companies for model training and retrieval augmented generation (RAG). - AI course policy – M. Montenegro
In ENVS 130b, I ask that you complete work without using AI. I respect your agency and expertise as students, and I completely understand you may be curious to explore new tools. I therefore have created this document to contextualize my course policy guidelines and to offer you some educational resources to study further. I am fully aware that life pressures may lead you to feel that there is no alternative but to use ChatGPT. If you are in this position, please do talk with us in the teaching team, since we want to find ways to help you. - I’m a Student. You Have No Idea How Much We’re Using ChatGPT.
“So rather than fully embracing AI as a writing assistant, the reasonable conclusion is that there needs to be a split between assignments on which using AI is encouraged and assignments on which using AI can’t possibly help. Colleges ought to prepare their students for the future, and AI literacy will certainly be important in ours. But AI isn’t everything. If education systems are to continue teaching students how to think, they need to move away from the take-home essay as a means of doing this, and move on to AI-proof assignments like oral exams, in-class writing, or some new style of schoolwork better suited to the world of artificial intelligence.”