Here are the articles, guides, and news about AI – week 29. It’s curated so you won’t have to scour the internet yourself.
Before we start!
If you like this topic and you want to support me:
- Comment on the article, or share; This will help us grow 📢
- Connect with me on Linkedin 🔗
- Subscribe to TechTonic Shifts to get your daily dose of tech 💉
There has been a lot of buzz lately about how we should handle AI and make sure it’s safe for everyone. It’s kind of a tug-of-war between letting AI grow and explore, and putting some ground rules in place so things don’t get out of hand.
Some people think that sharing AI openly is a good thing – more people can poke around, find issues, and help make it better. But others worry that it could also make it easier for troublemakers to use AI in not-so-nice ways [previous article: How AI is preventing cyber attacks [including e-book].
There’s also this question of whether some people pushing for AI rules are really just trying to keep competition at bay. It’s a lot to unpack!
A bit of news from the US, and Europe about regulations on AI:
California’s got this proposal called SB 1047 that’s stirring the pot. AI geeks and makers are pretty concerned about what it could mean for them. Oh, and Microsoft? They decided to step back from their spot on OpenAI’s board, which got people talking about the big dogs in tech and their hold on AI. It’s a complex situation with a lot of moving parts, but one thing’s for sure – we’re all trying to figure out the best path forward for AI.
So, Andrew Ng – he’s a pretty big deal in the AI world – he jumped into the conversation too. He’s really worried about California’s SB 1047 bill that’s making its way through the government. Ng thinks the bill is kind of a mess – it’s not clear what it’s asking for and it’s way too complicated. He’s afraid it could put a damper on all the cool AI stuff people are working on, especially the folks who are sharing their work openly for everyone to use and build on.
Instead of these big, confusing rules, Ng thinks we should focus on being more specific. Like, if there’s an AI tool that’s causing problems, let’s make rules for that particular thing rather than throwing a blanket over the whole AI scene. It’s a tricky balance, you know? We want AI to keep growing and getting better, but we also want to make sure it’s not causing harm. Figuring out the right way to handle that is gonna take some work.
Oh boy, talk about bad timing! The U.S. Department of Justice just came out and said they managed to put a stop to some Russian government-backed propaganda that was using AI. It’s like, “Hey, we’re in the middle of trying to figure out AI rules, and now this happens!”
Now, to be fair, it seems like AI-powered propaganda hasn’t been as big of a problem as some people were afraid it might be. And let’s be real, humans have been spreading propaganda without AI for ages. But still, this whole thing is a reminder that AI can be used in some pretty nasty ways if we’re not careful.
The worry is that when stuff like this happens, the people in charge might freak out and slam down even tougher rules on AI. It’s like, “Oh no, AI is being used for evil! Quick, let’s make a bunch of strict laws!” But that could end up causing problems for all the folks who are trying to do good things with AI.
It’s a tough spot to be in. We want to keep people safe from AI being used in bad ways, but we also don’t want to go overboard and make it really hard for the good stuff to happen. I guess that’s why everyone’s so focused on trying to get these AI rules right – it’s a delicate balance. Alright, let’s talk about the elephant in the room – Donald Trump picking Senator J.D. Vance as his running mate. Now, this could be a pretty big deal when it comes to AI rules.
See, Vance has some strong opinions about AI and bias. He’s been pretty vocal about thinking that a lot of AI models out there have a left-wing slant. In his view, this bias could really mess with the way we get and share information, and that’s a problem. Vance sees open-source AI models as a way to fight back against this perceived bias. The idea is that if the models are out there for everyone to see and work on, it’s harder for any one group to push their agenda.
Now, whether you agree with Vance’s take on AI bias or not, his comments do shine a light on the fact that bias in AI is a real concern. We’ve got to make sure these models are fair and not pushing any particular viewpoint over others. With Vance in the mix, there’s a chance we could see a bigger push for open-source AI and more scrutiny on potential biases. It adds another layer to the whole debate around AI regulation.
Of course, this is all happening in the middle of a presidential race, so politics are definitely going to play a role in how this all shakes out. But one thing’s for sure – the questions around AI fairness and transparency aren’t going away anytime soon.
On the other side of the pond, Europe, Ursula von der Leyen, the head honcho at the European Commission, is pretty pumped about AI. She thinks it could be a game-changer if we play our cards right. Like, imagine how much better healthcare could be with AI in the mix! And don’t even get her started on how it could boost productivity across the board.
But here’s the thing – Ursula knows we can’t just dive in headfirst. We’ve got to be smart about this. She wants Europe to lead the way in using AI responsibly, to make sure it’s actually helping people and making society better, not just causing chaos.
On the flip side, António Guterres, the big cheese at the UN, is a bit more wary. He’s worried that if we let AI run wild without any rules, it could spell big trouble. Like, “existential threat” kind of trouble.
António thinks governments and tech companies need to put their heads together ASAP to figure out how to keep AI in check while still reaping the benefits. He also wants to make sure that everyone gets a piece of the AI pie, not just the rich countries. We don’t want to leave developing economies in the dust, right?
Meanwhile, Thierry Breton, the European Commissioner for Internal Market, is all about making Europe the cool kid on the AI regulation block [pun IS intended – Europe and it’s regulations – try to get it people: regulations are NEVER cool] He’s super proud of the EU’s new AI Act, which he sees as a major milestone in setting global standards for how we handle AI. The Act is basically like a rulebook for AI. It says that if an AI system is going to have a big impact, it needs to be transparent, safe, and secure. Thierry believes this will help us keep AI’s risks under control while still letting innovation happen.
So, there you have it
Multiple politicians on both sides of the Western pond, all with different takes on AI. But they all seem to agree on one thing: we’ve got to be smart and careful about how we use this powerful technology. It’s not just about the cool stuff AI can do, but also about making sure it’s actually good for people and society as a whole.
Is this something you should care about?
Alright, let’s break this down. The big question on everyone’s mind is: how are we going to handle AI as it gets bigger and more powerful? The rules we set now could make or break how fast AI grows and who gets to call the shots.
If you’re an AI maker, you’ve got to keep your eye on these rules. Nobody wants to get in trouble with the law, right? But if the rules are too strict or confusing, it could really put a damper on new ideas and tech. And let’s be real, it’s the little guys – the individuals and small startups – who are going to feel the heat the most.
But for the rest of us, this AI governance stuff is a big deal too. We want to make sure AI isn’t being used in shady ways, but we also don’t want to slam the brakes on all the cool, helpful things it could do for society.
The thing is, the choices we make now about AI rules are going to stick with us for a long time. So everyone needs to step up and get involved in these talks – we’ve got to find that sweet spot where AI is safe but also has room to grow and innovate.
The open-source folks, you know, the ones who share their AI work with everyone? They’re at a real crossroads here. What happens next could decide what role they play in the AI game moving forward.
So here’s the deal: let’s all stay in the loop and be proactive. If we work together, we can create a future where AI is both trustworthy and cutting-edge. It’s going to take some effort, but it’s so worth it. The future of AI is in our hands, and it’s up to us to shape it into something awesome!
Hottest AI/ML news
1.FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision
FlashAttention-2 is widely used by most libraries to accelerate Transformer training and inference, already making attention faster on GPUs by minimizing memory reads/writes, but it has yet to take advantage of the latest H100s. FlashAttention-3 achieves a 1.5–2x speedup, reaching up to 740 TFLOPS on FP16 and nearly 1.2 PFLOPS on FP8. This increases GPU utilization to 75% of the theoretical maximum on H100s, up from 35% for FlashAttention-2. We think this is an important step in reducing the cost of the next generation of LLMs and may trigger the start of some major training runs.
2. New Leading 8k Output Tokens Available for Sonnet 3.5 in Claude API
Anthropic doubled the ‘maximum output’ token limit for Claude 3.5 Sonnet from 4096 to 8192 in the Anthropic API — we think this is significant as, despite the huge progress in expanding input token context windows (to 2 million+), output tokens can still be a constraint for many applications (such as translation and conversion tasks). Anthropic also made fine-tuning available for Haiku 3.0 in Amazon Bedrock. The fine-tuning API is currently available in preview.
3. Excitement Growing for Imminent New LLM Releases: LLama 3 405B and New Models in LMSYS Arena
The Information reported that META will release LLama 3 405B on July 23rd. We also saw three more new models appear for testing in LMSYS Arena; ‘upcoming-gpt-mini’, ‘column-u’, and ‘column-r’. This is where GPT4o was first secretly tested shortly before release, but it is unclear which company or companies the new models come from.
4. Microsoft Gives Up Observer Seat on OpenAI Board
Microsoft has stepped down from its observer seat on OpenAI’s board, which OpenAI noted reflected confidence in OpenAI’s trajectory under CEO Sam Altman. The move streamlines Microsoft’s relationship with OpenAI and we think it is likely motivated in part to address and reduce antitrust concerns regarding Microsoft’s influence over the company. OpenAI will not offer future observer roles, preferring direct partnership interactions, as with Microsoft and Apple.
5. OpenAI Secret Project “Strawberry” Aims To Boost AI Reasoning Power
Project Strawberry is OpenAI’s latest effort to improve AI reasoning. While the exact details are kept under wraps, it’s reportedly a significant leap forward in LLM capability. The project aims to enable AI models to plan ahead, understand the world more like humans do, and easily tackle complex multi-step problems. A different source also noted that internal models at OpenAI had scored over 90% on a MATH dataset (championship math problems).
6. OpenAI unveils five-level scale to AGI, aims to reach level 2 soon
OpenAI has created an internal five-level scale to track its large language models’ progress toward AGI. Potentially related to its project “Strawberry” above, it is reportedly on the cusp of achieving Level 2. Level 2 is “Reasoner,” demonstrating human-like problem-solving and characterized by advanced logic and reasoning. Level 3 is AI Agents that work on tasks and actions for days at a time.
7. Meta Researchers Distill System 2 Thinking Into LLMs, Improving Performance on Complex Reasoning
In a new paper, researchers at Meta FAIR present “System 2 distillation,” a technique that teaches LLMs complex tasks without requiring intermediate steps. In this research, Meta integrated System 2’s intricate reasoning methods (such as Chain-of-Thought) into the faster System 1 processes in LLMs.
Very short reads/videos to keep you learning
1. In-Depth Understanding of Vector Search for RAG and Generative AI Applications
This article focuses on vector search in RAG; it discusses why we need a vector search in RAG applications and how vectors and vector databases work. It also explores what makes Azure AI Search a good retrieval system and how it integrates.
This post shows how to build efficient prompts for your applications. It uses Amazon Bedrock playgrounds and Anthropic’s Claude 3 models to demonstrate how to build efficient prompts by applying simple techniques. It also talks about the anatomy of a prompt and presents an in-depth prompt example for Retrieval Augmented Generation.
3. Principles of Reinforcement Learning: An Introduction With Python
This article introduces fundamental principles and offers a beginner-friendly example of reinforcement learning. It explains the key terms in RL, its steps, algorithms, and implementation in Python.
4. Preventing Prompt Injection in OpenAI: A Case Study With Priceline’s OpenAI Tool “Penny”
This article suggests steps to mitigate prompt injections. It also proposes solutions like testing a better model, fully adapting a list of known patterns, running adversarial finetuning, and more.
5. A Data Leader’s Technical Guide to Scaling Gen AI
This article will cover three actions that data and AI leaders can consider to move from gen AI pilots to scaling data solutions. It focuses on how organizations can strengthen the quality and readiness of their data, examines how organizations can use gen AI to build better data products, and explores key data-management considerations.
AI/ML tools
- Storm is an LLM-powered knowledge curation system that researches a topic and generates a full-length report with citations.
- MobileLLM optimizes sub-billion parameter language models for on-device use cases.
- LightRAG is a modular library like PyTorch for building LLM applications like chatbots and code generation, featuring a RAG pipeline.
- Tabby is a self-hosted AI coding assistant offering an open-source and on-premises alternative to GitHub Copilot.
Top papers of the week
This paper introduces the Comprehensive RAG Benchmark (CRAG), a factual question-answering benchmark of 4,409 question-answer pairs, and mock APIs to simulate web and Knowledge Graph (KG) search. It contains diverse questions across five domains and eight question categories. It reflects varied entity popularity from popular to long-tail and temporal dynamisms ranging from years to seconds.
2. MambaVision: A Hybrid Mamba-Transformer Vision Backbone
This paper proposes a novel hybrid Mamba-transformer backbone. The work redesigns the Mamba formulation to enhance its capability for efficient modeling of visual features. The MambaVision models achieve a new State-of-the-Art (SOTA) performance in terms of Top-1 accuracy and image throughput.
3. MJ-Bench: Is Your Multimodal Reward Model Really a Good Judge for Text-to-Image Generation?
MJ-Bench is a new benchmark for evaluating multimodal reward models that provide feedback on text-to-image generation technologies, such as DALLE-3 and Stable Diffusion. It tests models on criteria such as alignment, safety, image quality, and bias. Notably, the benchmark found that closed-source VLMs like GPT-4o excel in providing effective feedback.
4. Distilling System 2 into System 1
This work examines the integration of System 2’s intricate reasoning methods (such as Chain-of-Thought) into the faster System 1 processes in LLMs. By employing self-supervised learning, the authors have improved System 1 performance and lowered computation costs by embedding System 2’s reasoning capabilities into System 1, suggesting a more efficient approach to handling complex reasoning in AI.
Quick links
1. Towards AI recently tested Launchpad by Latitude.sh, a container-based GPU cloud for inference and fine-tuning. Launchpad’s notable feature is its advanced, high-level dedicated container-based GPUs, capable of handling the significant computational demands of AI workloads.
2. Our friends at Mira have come out of stealth and announced their $9m seed raise. Mira is building decentralized AI infrastructure. They abstract AI infrastructure into “Flows”, a new AI building block that combines models, data & compute into a specific instruction set. Developers leverage Flows to minimize overhead & contributors publish diverse Flows on Mira, creating an ecosystem of AI resources. Mira already has over a dozen teams leveraging their Flow Market, contributing complex AI products across various sectors. Get early access to their platform here!
3. Amazon’s AI-powered shopping assistant, Rufus, is now available for all U.S. customers in the Amazon mobile app. The AI chatbot has been trained on Amazon’s product catalog, customer reviews, community Q&As, and other public information.
4. AWS launched App Studio to build internal enterprise applications from a written prompt. Amazon defines enterprise apps as having multiple UI pages that can pull from various data sources, perform complex operations like joins and filters, and embed business logic.
5. Patronus AI unveiled Lynx, an open-source model designed to detect and mitigate hallucinations in LLMs. Lynx outperforms industry giants like OpenAI’s GPT-4 and Anthropic’s Claude 3 in hallucination detection tasks, representing a significant leap forward in AI trustworthiness.
6. Intel Capital Backs AI Construction Startup That Could Boost Intel’s Own Manufacturing Prospects Intel Capital is leading a $15 million investment into Buildots, a company that uses AI and computer vision to create a digital twin of construction sites. Buildots, which uses AI and computer vision to create digital twins of construction sites, has now raised $121 million.
Who can help my brilliant student!
Who can help? I am mentoring a last year, ambitious, student Preetham Dundigalla, an ML- and full-stack engineer, president of the student council, who will graduate early next year from the VIT University. He is a gifted person, and he wants to start already with ML or engineering projects to gain experience. This is his resume.
Well, that’s a wrap for today. Tomorrow, I’ll have a fresh episode of TechTonic Shifts for you. If you enjoy my writing and want to support my work, feel free to buy me a coffee ♨️
Think a friend would enjoy this too? Share the newsletter and let them join the conversation. LinkedIn appreciates your likes by making my articles available to more readers.
Leave a Reply